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
47 Description: shutdown vs evict race
48 Details : client_disconnect_export vs connect request race.
49 if client will evicted at this time - we start invalidate
50 thread without referece to import and import can be freed
54 Frequency : rare, need acl's on inode.
56 Description: client can't handle ost additional correctly
57 Details : if ost was added after client connected to mds client can have
58 hit lnet_try_match_md ... to big messages to wide striped files.
59 in this case need teach client to handle config events about add
60 lov target and update client max ea size at that event.
62 Severity : enhancement
64 Description: Changelogs
65 Details : Changelogs are a lightweight mechanism to track filesystem
66 metadata and namespace changes. The changelog is recorded
67 permanently on the MDTs, and is periodically "consumed" / purged
68 when records are no longer needed.
70 Severity : enhancement
72 Description: compact fld format with extents
73 Details : Store range of seq rather than every seq in FLD. Seq
74 controller update FLD rather than clients. In Case of CMD, mdt0
75 has FLD, all other metadata server act as non persistent proxy
76 for FLD queries and cache fld entries in fld cache.
81 Description: don't skip ost target if they assigned to file
82 Details : Drop slow OSCs if we can, but not for requested start idx.
83 This means "if OSC is slow and it is not the requested
84 start OST, then it can be skipped, otherwise skip it only
85 if it is inactive/recovering/out-of-space.
89 Description: more cleanup in mds_lov
90 Details : not send LOV EA under replay, we can't know about they size at this
91 time. Don't allow client connect to mds before any ost connected,
92 for avoid problems with LOV EA size and returning EIO to client.
94 Severity : enhancement
96 Description: Interoperability at server side (Disk interoperability)
98 Severity : enhancement
100 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
102 Severity : enhancement
104 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
106 Severity : enhancement
108 Description: New client IO stack (CLIO).
110 Severity : enhancement
112 Description: Commit on sharing. Eliminate inter-client dependencies between
113 uncommitted transactions by doing transaction commits.
114 Thereby clients may recovery independently.
117 Frequency : Create a symlink file with a very long name
119 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
120 Details : If there is no extra space in the request for early cancels,
121 ldlm_req_handles_avail() returns 0 instead of a negative value.
123 Severity : enhancement
125 Description: Add /proc entry for import status
126 Details : The mdc, osc, and mgc import directories now have
127 an import directory that contains useful import data for debugging
130 Severity : enhancement
132 Description: Re-disable certain /proc logging
133 Details : Enable and disable client's offset_stats, extents_stats and
134 extents_stats_per_process stats logging on the fly.
137 Frequency : Only on FC kernels 2.6.22+
139 Description: oops in statahead
140 Details : Do not drop reference count for the dentry from VFS when lookup,
141 VFS will do that by itself.
143 Severity : enhancement
145 Description: Generic /proc file permissions
146 Details : Set /Proc file permissions in a more generic way to enable non-
147 root users operate on some /proc files.
151 Description: Hitting mdc_commit_close() ASSERTION
152 Details : Properly handle request reference release in
153 ll_release_openhandle().
157 Description: quota recovery deadlock during mds failover
158 Details : This patch includes att18982, att18236, att18237 in bz14840.
160 1. fix osts hang when mds does failover with quotaon
161 2. prevent watchdog storm when osts threads wait for the
166 Frequency : only patchless client
167 Description: add workaround for race between add/remove dentry from hash
169 Severity : enhancement
171 Description: Allow OST glimpses to return PW locks
175 Description: LBUG when llog conf file is full
176 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
180 Description: Prevent import from entering FULL state when server in recovery
184 Description: service mount cannot take device name with ":"
185 Details : Only when device name contains ":/" will mount treat it as
191 Description: replace ptlrpcd with the statahead thread to interpret the async
192 statahead RPC callback
196 Frequency : on recovery
197 Description: I/O failures after umount during fail back
198 Details : if client reconnected to restarted server we need join to recovery
199 instead of find server handler is changed and process self
200 eviction with cancel all locks.
202 Severity : enhancement
204 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
206 Severity : enhancement
208 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
210 Severity : enhancement
212 Description: Upcall on Lustre log has been dumped
213 Details : Allow for a user mode script to be called once a Lustre log has
214 been dumped. It passes the filename of the dumped log to the
215 script, the location of the script can be specified via
216 /proc/sys/lnet/debug_log_upcall.
221 Description: avoid idr_remove called for id which is not allocated.
222 Details : Move assigment s_dev for clustered nfs to end of initialization,
223 to avoid problem with error handling.
228 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
229 Details : When connection is reused this not moved from CONN_UNUSED_HASH
230 into CONN_USED_HASH and this prodice warning when put connection
231 again in unused hash.
234 Severity : enhancement
236 Description: File striping can now be set to use an arbitrary pool of OSTs.
238 Severity : enhancement
240 Description: Export bytes_read/bytes_write count on OSC/OST.
244 Description: Early reply size mismatch, MGC loses connection
245 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
246 the connect flags are properly negotiated.
251 Description: quotas are not honored with O_DIRECT
252 Details : all writes with the flag O_DIRECT will use grants which leads to
253 this problem. Now using OBD_BRW_SYNC to guard this.
257 Description: add quota statistics
258 Details : 1. sort out quota proc entries and proc code.
259 2. add quota statistics
261 Severity : enhancement
263 Description: enable quota support for HEAD.
267 Description: Properly propagate oinfo flags from lov to osc for statfs
268 Details : restore missing copy oi_flags to lov requests.
270 Severity : enhancement
272 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
273 : l_getgroups(8), lst(8), routerstat(8)
274 Details : included man pages for llobdstat(8), llstat(8),
275 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
277 Severity : enhancement
279 Description: configure's --enable-quota should check the
280 : kernel .config for CONFIG_QUOTA
281 Details : configure is terminated if --enable-quota is passed but
282 : no quota support is in kernel
286 Description: Remove portals compatibility
287 Details : Remove portals compatibility, not interoperable with releases
292 Description: Resolve device initialization race
293 Details : Prevent proc handler from accessing devices added to the
294 obd_devs array but yet be intialized.
296 Severity : enhancement
298 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
300 Severity : enhancement
302 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
307 Description: Using wrong pointer in osc_brw_prep_request
308 Details : Access to array[-1] can produce panic if kernel compiled with
309 CONFIG_PAGE_ALLOC enabled
313 Description: Client runs out of low memory
314 Details : Consider only lowmem when counting initial number of llap pages
318 Description: *optional* service tags registration
319 Details : if the "service tags" package is installed on a Lustre node
320 When the filesystem is mounted, a local-node service tag will
321 be created. See http://inventory.sun.com/ for more information
322 about the Service Tags asset management system.
326 Description: Kernel BUG tries to release flock
327 Details : Lustre does not destroy flock lock before last reference goes
328 away. So always drop flock locks when client is evicted and
329 perform unlock regardless of successfulness of speaking to MDS.
333 Description: add refcount protection for osc callbacks, avoid panic on shutdown
337 Description: sanity test 65a fails if stripecount of -1 is set
338 Details : handle -1 striping on filesystem in ll_dirstripe_verify
343 Description: ASSERTION(CheckWriteback(page,cmd)) failed
344 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
347 Severity : enhancement
349 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
353 Description: do not process already freed flock
354 Details : flock can possibly be freed by another thread before it reaches
355 to ldlm_flock_completion_ast.
359 Description: LBUG during stress test
360 Details : Need properly lock accesses the flock deadlock detection list.
364 Description: oops in page fault handler
365 Details : kernel page fault handler can return two special 'pages' in error
366 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
370 Description: timeout with invalidate import.
371 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
372 should be handled by ptlrpcd. This produce long age waiting and
373 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
375 Severity : enhancement
377 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
381 Description: enable MGS and MDT services start separately
382 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
383 a co-located MGS without starting the MGS, which is a complement
384 to 'nosvc' mount option.
388 Frequency : after recovery
389 Description: precreate to many object's after del orphan.
390 Details : del orphan st in oscc last_id == next_id and this triger growing
391 count of precreated objects. Set flag LOW to skip increase count
392 of precreated objects.
396 Frequency : rare, on clear nid stats
397 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
398 Details : when clean nid stats sometimes try destroy live entry,
399 and this produce panic in free.
403 Description: Stack overflow during MDS log replay
404 ease stack pressure by using a thread dealing llog_process.
408 Description: wait until IO finished before start new when do lock cancel.
409 Details : VM protocol want old IO finished before start new, in this case
410 need wait until PG_writeback is cleared until check dirty flag and
411 call writepages in lock cancel callback.
413 Severity : enhancement
415 Description: using special macro for print time and cleanup in includes.
419 Description: mds_mfd_close() ASSERTION(rc == 0)
420 Details : In mds_mfd_close(), we need protect inode's writecount change
421 within its orphan write semaphore to prevent possible races.
425 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
429 Frequency : rare, on shutdown ost
430 Description: don't hit live lock with umount ost.
431 Details : shrink_dcache_parent can be in long loop with destroy dentries,
432 use shrink_dcache_sb instead.
436 Description: don't panic with use echo client
437 Details : echo client pass NULL as client nid pointer and this produce null
442 Description: fix build on ppc32
443 Details : compile code with -m64 flag produce wrong object file for ppc32.
447 Description: add message levels for liblustreapi
451 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
452 Details : Move the 'good_osts' check before the 'total_bavail' check. This
453 will result in an -EAGAIN and in the exit call path we call
454 alloc_rr() which will with increasing aggressiveness attempt to
455 aquire precreated objects on the minimum number of required OSCs.
459 Description: Use old size assignment to avoid deadlock
460 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
461 the scheduling while holding a spinlock. We do not need locking
462 for size in ll_update_inode() because size is only updated from
463 the MDS for directories or files without objects, so there is no
464 other place to do the update, and concurrent access to such inodes
465 are protected by the inode lock.
469 Description: resolve "_IOWR redefined" build error on SLES10
473 Description: dump the memory debugging after all modules are unloaded to
474 suppress false negative in conf_sanity test 39
476 Severity : enhancement
478 Description: build kernel-ib packages for OFED 1.3 in our release cycle
483 Description: fix SLES kernel versioning
484 Details : the kernel version for our SLES 10 kernel did not include a "-"
485 before the "smp" at the end. while this was not a problem in
486 general, it did mean that software trying to use the kernel
487 version to try to detect a vendor specific kernel would fail.
488 this was most evident by the OFED build scripts.
492 Description: Don't update lov_desc members until making sure they are valid
493 Details : When updating lov_desc members via proc fs, need fix their
494 validities before doing the real update.
498 Description: don't put request into delay list while invalidate in flight.
499 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
500 import in flight. this produce timeout for invalidate and sometimes
501 can cause stale data.
505 Frequency : on ppc only
506 Description: not convert ost objects for directory because it's not exist.
507 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
511 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
513 Severity : enhancement
515 Description: Update to support 2.6.22.14 vanilla kernel.
519 Frequency : rare, on recovery
520 Description: read procfs can produce deadlock in some situation
521 Details : Holding lprocfs lock which send rpc can produce block for destroy
522 obd objects and this also block reconnect with -EALREADY. This
523 isn't fix all lprocfs bugs - but make it rare.
525 Severity : enhancement
527 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
530 Frequency : frequent on X2 node
532 Description: mdc_set_open_replay_data LBUG
533 Details : Set replay data for requests that are eligible for replay.
537 Description: lustre_mgs: operation 101 on unconnected MGS
538 Details : When MGC is disconnected from MGS long enough, MGS will evict the
539 MGC, and late on MGC cannot successfully connect to MGS and a lot
540 of the error messages complaining that MGS is not connected.
544 Frequency : on network error
545 Description: panic with double free request if network error
546 Details : mdc_finish_enqueue is finish request if any network error ocuring,
547 but it's true only for synchronus enqueue, for async enqueue
548 (via ptlrpcd) this incorrect and ptlrpcd want finish request
551 Severity : enhancement
553 Description: client-side metadata stat-ahead during readdir(directory readahead)
554 Details : perform client-side metadata stat-ahead when the client detects
555 readdir and sequential stat of dir entries therein
558 Frequency : on start mds
560 Description: Implement get_info(last_id) in obdfilter.
563 Frequency : occasional
565 Description: Correctly check stale fid, not start epoch if ost not support SOM
566 Details : open with flag O_CREATE need set old fid in op_fid3 because
567 op_fid2 was overwritten with new generated fid, but mds can answer
568 with one of these two fids and both is not stale. Setattr
569 incorrectly started an epoch and assume will be called
570 done_writing, but without SOM done_writing ever being called.
573 Frequency : rare, depends on device drivers and load
575 Description: MDS or OSS nodes crash due to stack overflow
576 Details : Code changes in 1.8.0 increased the stack usage of some functions.
577 In some cases, in conjunction with device drivers that use a lot
578 of stack the MDS (or possibly OSS) service threads could overflow
579 the stack. One change which was identified to consume additional
580 stack has been reworked to avoid the extra stack usage.
583 Frequency : occasional
585 Description: Do not fail import if osc_interpret_create gets -EAGAIN
586 Details : If osc_interpret_create got -EAGAIN it immediately exits and
587 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
588 oscc_has_objects and see OSC has no objests and call
589 oscc_internal_create to resend create request.
591 Severity : enhancement
593 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
595 Severity : enhancement
597 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
600 Frequency : very rare
602 Description: Fix replay if there is an un-replied request and open
603 Details : In some cases, older replay request will revert the
604 mcd->mcd_last_xid on MDS which is used to record the client's
607 Severity : enhancement
609 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
611 Severity : enhancement
613 Description: Add rhel5 support to HEAD.
615 Serverity : enhancement
617 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
622 Description: Don't allow skipping OSTs if index has been specified.
623 Details : Don't allow skipping OSTs if index has been specified, make
624 locking in internal create lots better.
628 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
629 Details : access to bitfield in structure is always rounded to long
630 and this produce problem with not atomic change any bit.
634 Description: Lustre make rpms failed.
635 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
636 builds Lustre rpms from tarball.
639 Frequency : rare on shutdown ost
641 Description: If llog cancel was not send before clean_exports phase, this can
642 produce deadlock in llog code.
643 Details : If llog thread has last reference to obd and call class_import_put
644 this produce deadlock because llog_cleanup_commit_master wait when
645 last llog_commit_thread exited, but this never success because was
646 called from llog_commit_thread.
650 Description: allow userland application know is lost one of stripes.
651 Details : fill lvb_blocks with error code on ost and return it to
652 application if error flag found.
656 Description: NULL lov_tgts causing MDS oops
657 Details : more safe checks for NULL lov_tgts for avoid oops.
659 Severity : enhancement
661 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
665 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
666 Details : If an OST is down the MDS will hang indefinitely in
667 obd_statfs_rqset() waiting for the statfs data. While for
668 MDS QOS usage of statfs, it should not stuck in waiting.
670 Severity : enhancement
672 Description: remote_acl support
673 Details : Support ACL-based permission check for remote user.
674 Support setfacl/getfacl for remote user with the utils
675 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
676 the system "{s,g}etfacl" utils.
678 Severity : enhancement
680 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
682 Severity : enhancement
684 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
688 Description: make mgs_setparam() handle fsname containing dash
689 Details : fsname containing a dash does not work with lctl conf_param
691 Severity : enhancement
693 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
695 Severity : enhancement
697 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
701 Frequency : rare, at shutdown
702 Description: access already free / zero obd_namespace.
703 Details : if client_disconnect_export was called without force flag set,
704 and exist connect request in flight, this can produce access to
705 NULL pointer (or already free pointer) when connect_interpret
706 store ocd flags in obd_namespace.
710 Frequency : only at startup
711 Description: not alloc memory with spinlock held.
712 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
713 if any spinlock held.
715 Severity : enhancement
717 Description: make lustre randomly fail allocating memory
718 Details : Make lustre randomly failed allocating memory for testing purpose.
720 Severity : enhancement
722 Description: lost problems with lov objid file
723 Details : Fixes some scability and access to not inited memory problems
724 in work with lov objdid file.
729 Description: lfs find does not continue on file error
730 Details : Continue other files processing when a file/dir is absent.
734 Description: Inconsistent usage of lustre_pack_reply()
735 Details : Standardize the usage of lustre_pack_reply() such that it
736 always generate a CERROR on failure.
739 Frequency : occasional
741 Description: MDS hang or stay in waiting lock
742 Details : If client receive lock with CBPENDING flag ldlm need send lock
743 cancel as separate rpc, to avoid situation when cancel request
744 can't processed due all i/o threads stay in wait lock.
748 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
752 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
756 Description: Update to SLES9 kernel-2.6.5-7.287.3.
760 Description: Update to RHEL4 latest kernel.
762 Severity : enhancement
764 Description: Build SLES10 patchless client fails
765 Details : The configure was broken by run ./configure with
766 --with-linux-obj=.... argument for patchless client. When the
767 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
768 can't search header adequately. Use absolute path such as
769 -I($LINUX)/include instead.
773 Description: interrupt oig_wait produce painc on resend.
774 Details : brw_redo_request can be used for resend requests from ptlrpcd and
775 private set, and this produce situation when rq_ptlrpcd_data not
776 copyed to new allocated request and triggered LBUG on assert
777 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
778 ptlrpcd set if request is changed and can be safety changed to use
783 Description: LASSERT_{REQ,REP}SWAB macros are buggy
784 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
785 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
786 these macros and replace them with their difinitions instead.
790 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
791 Details : Update which_patch & target file for SLES10 latest kernel.
794 Description: Remove client patches from SLES 10 kernel.
795 Details : This causes SLES 10 clients to behave as patchless clients
796 even on a Lustre-patched (server) kernel.
798 Severity : enhancement
800 Description: self-adjustable client's lru lists
801 Details : use adaptive algorithm for managing client cached locks lru
802 lists according to current server load, other client's work
803 pattern, memory activities, etc. Both, server and client
804 side namespaces provide number of proc tunables for controlling
807 Severity : enhancement
809 Description: light-weight GSS support
810 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
811 and reduce performance overhead.
813 Severity : enhancement
815 Description: Linux keyring support
816 Details : Support using service of Linux keyring for Lustre GSS internal
817 context refresh/cache mechanism.
821 Description: Fix errors in lfs documentation
822 Details : Fixes man pages
826 Description: don't use GFP_* in generic Lustre code.
827 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
831 Description: obdclass is limited by single OBD_ALLOC(idarray)
832 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
836 Description: warnings with build patchless client with vanila 2.6.19 and up
837 Details : change old ctl_table style and replace ctl_table/ctl_table_header
838 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
842 Frequency : Only for SLES
843 Description: sanity_quota fail test_1
844 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
845 file, which makes TSTID[2] inunique.
849 Frequency : Only for SLES9
850 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
851 Details : Update target/ChangeLog/which_patch .
855 Frequency : Only for test_10 in sanity-quota.sh
856 Description: a bug in quota test code
857 Details : checking a test flag in a wrong place
861 Frequency : Only for test_13 in sanity-quota.sh
862 Description: a bug in quota test script
863 Details : shouldn't check overfull allocated quota
867 Description: jbd statistics
868 Details : Port older jbd statistics patch for sles10
870 Severity : enhancement
872 Description: Add printing inode info into message about error in writepage.
877 Description: reply_lock_interpret crash due to race with it and lock cancel.
878 Details : Do not replay locks that are being cancelled. Do not reference
879 locks by their address during replay, just by their handle.
883 Frequency : When flocks are used.
884 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
885 Details : Flock locks might destroy just granted lock if it could be merged
886 with another existing flock, this is done in completion handler,
887 so teach ldlm_cli_enquque_fini that this is a valid case for
893 Description: Oops in read and write path when failing to allocate lock.
894 Details : Check if lock allocation failed and return error back.
898 Description: lstripe command fails for valid OST index
899 Details : The stripe offset is compared to lov->desc.ld_tgt_count
900 instead of lov->desc.ld_active_tgt_count.
904 Description: sanity.sh failed test 103
905 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
910 Description: df doesn't work properly if diskfs blocksize != 4K
911 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
915 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
916 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
918 Severity : enhancement
920 Description: Adaptive timeouts
921 Details : RPC timeouts adapt to changing server load and network
922 conditions to reduce resend attempts and improve recovery time.
926 Description: llapi_file_create() does not allow some changes
927 Details : add llapi_file_open() that allows specifying the mode and
928 open flags, and also returns an open file handle.
932 Description: merge and cleanup kernel patches.
933 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
937 Description: Add journal checksum support.(Kernel part)
938 Details : The journal checksum feature adds two new flags i.e
939 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
940 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
941 indicates that the commit block contains the checksum for
942 the blocks described by the descriptor blocks. Now commit
943 record can be sent to disk without waiting for descriptor
944 blocks to be written to disk. This behavior is controlled
945 using JBD2_FEATURE_ASYNC_COMMIT flag.
949 Description: OSS needs mutliple precreate threads
950 Details : Add ability to start more than one create thread per OSS.
954 Description: Sanity.sh test_65k failed: setstripe should have succeeded
955 Details : execute lfs setstripe on client
959 Description: mds_obd_create error creating tmp object
960 Details : When the user sets quota on root, llog will be affected and can't
961 create files and write files.
964 Frequency : Always on ia64 patchless client, and possibly others.
966 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
967 Details : This allows the patchless client to be loaded on architectures
972 Description: RedHat Update kernel for RHEL5
973 Details : Add the kernel config file for RHEL5.
977 Description: RedHat Update kernel for RHEL5
978 Details : Modify the kernel config file more closer RHEL5.
982 Description: Build failure against Centos5 (RHEL5)
983 Details : Define PAGE_SIZE when it isn't present.
987 Description: test_7 on sanity-quota.sh can't be used on separate machines
988 Details : add facet to handle it
992 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
993 Details : with NFS, the anon dentry's parent was set to itself in
994 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
995 corresponding dentry other than use rec->ur_name.
997 Severity : enhancement
999 Description: lfs setstripe enhancement
1000 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1004 Description: enable data checksumming by default
1005 Details : enable checksum by default, allow --disable-checksum
1006 configure option and "-o nochecksum" mount option. Checksums
1007 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1011 Description: lustre support for RHEL5
1012 Details : Add support for RHEL5.
1016 Description: Client eviction due to failover config
1017 Details : after a connection loss, the lustre client should attempt to
1018 reconnect to the last active server first before trying the
1019 other potential connections.
1023 Description: when mds and osts use different quota unit(32bit and 64bit),
1024 quota will be released repeatly.
1025 Details : void sending multiple quota reqs to mds, which will keep the
1026 status between the reqs.
1030 Description: osts not allocated evenly to files
1031 Details : change the condition to increase offset_idx
1034 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1036 Description: Data corruption for OSTs that are formatted larger than 2TB
1038 Details : When generating the bio request for lustre file writes the
1039 sector number would overflow a temporary variable before being
1040 used for the IO. The data reads correctly from Lustre (which
1041 will overflow in a similar manner) but other file data or
1042 filesystem metadata may be corrupted in some cases.
1046 Description: Tune the kernel for good SCSI performance.
1047 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1048 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1053 Description: rewrite ext2-derived code in obdclass/uuid.c
1054 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1055 from scratch preserving functionality.
1059 Description: "lfs find -obd UUID" prints directories
1060 Details : "lfs find -obd UUID" will return all directory names instead
1061 of just file names. It is incorrect because the directories
1062 do not reside on the OSTs.
1066 Description: use i_size_read and i_size_write in 2.6 port
1067 Details : replace inode->i_size access with i_size_read/write()
1070 Frequency : when removing large files
1072 Description: scheduling issue during removal of large Lustre files
1073 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1074 It causes scheduling issues when removing large files (17TB in the
1078 Frequency : only with liblustre clients on XT3
1080 Description: evictions taking too long
1081 Details : allow llrd to evict clients directly on OSTs
1084 Frequency : only on ppc
1086 Description: /proc/fs/lustre/devices broken on ppc
1087 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1088 We should make sure the type of 'index' is loff_t and then cast
1089 explicitly as needed below. Do not assign an explicitly cast
1093 Frequency : only for rhel5
1095 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1096 Details : Modify the target file & which_kernel.
1098 Severity : enhancement
1100 Description: omit set fsid for export NFS
1101 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1102 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1106 Frequency : Always, for kernels after 2.6.16
1107 Description: Fix warning idr_remove called for id=.. which is not allocated.
1108 Details : Last kernels save old s_dev before kill super and not allow
1109 to restore from callback - restore it before call kill_anon_super.
1113 Description: buffer overruns could theoretically occur
1114 Details : llapi_semantic_traverse() modifies the "path" argument by
1115 appending values to the end of the origin string, and a
1116 overrun may occur. Adding buffer overrun check in liblustreapi.
1120 Description: Fix error on 'ls .' at the top of the Lustre mount.
1121 Details : Don't revalidate dentry if it is a root dentry.
1125 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1126 Details : Modify vm-tunables-rhel4.patch.
1130 Description: Kernel config for 2.6.18-vanilla.
1131 Details : Modify targets/2.6-vanilla.target.in.
1132 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1133 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1134 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1135 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1139 Description: improve handling recoverable errors
1140 Details : if request processig with error which can be recoverable on server
1141 request should be resend, otherwise page released from cache and
1146 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1147 Details : Modify target file & which_patch.
1148 A flaw was found in the IA32 system call emulation provided
1149 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1150 value could be stored in the %RAX register, which could trigger an
1151 out-of-bounds system call table access. An untrusted local user
1152 could exploit this flaw to run code in the kernel
1153 (ie a root privilege escalation). (CVE-2007-4573).
1157 Description: O_DIRECT bypasses client statistics.
1158 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1159 read_ahead_stats were updated. copied the stats section from
1160 osc_send_oap_rpc() into async_internal().
1164 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1168 Description: change order of libsysio includes
1169 Details : '#include sysio.h' should always come before '#include xtio.h'
1173 Description: handle "serious error: objid * already exists" more gracefully
1174 Details : If LAST_ID value on disk is smaller than the objects existing in
1175 the O/0/d* directories, it indicates disk corruption and causes an
1176 LBUG(). If the object is 0-length, then we should use the existing
1177 object. This will help to avoid a full fsck in most cases.
1179 Severity : enhancement
1181 Description: adapt the lustre_config script to support the upgrade case
1182 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1183 server targets to 1.6 in parallel.
1187 Description: To avoid grant space > avaible space when the disk is almost
1188 full. Without this patch you might see the error "grant XXXX >
1189 available" or some LBUG about grant, when the disk is almost
1191 Details : In filter_check_grant, for non_grant cache write, we should
1192 check the left space by if (*left > ungranted + bytes), instead
1193 of (*left > ungranted), because only we are sure the left space
1194 is enough for another "bytes", then the ungrant space should be
1195 increase. In client, we should update cl_avail_grant only there
1196 is OBD_MD_FLGRANT in the reply.
1200 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1201 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1202 described in CVE-2007-4573. This problem could allow untrusted
1203 local users to gain root access.
1206 Frequency : when using O_DIRECT and quotas
1208 Description: Incorrect file ownership on O_DIRECT output files
1209 Details : block usage reported by 'lfs quota' does not take into account
1210 files that have been written with O_DIRECT.
1215 Description: touch file failed when fs is not full
1216 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1217 otherwise we can get ENOSP while fs is not full.
1221 Description: parallel lock callbacks
1222 Details : Instead of sending blocking and completion callbacks as separated
1223 requests, adding them to a set and sending in parallel.
1226 Frequency : only for Cray XT3
1227 Bugzilla : 12829/13455
1228 Description: Changing primary group doesn't change the group lustre assigns to
1230 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1231 overridden with the primary group provided by the group upcall,
1232 whereas the supplied fsgid can be trusted if it is in the list of
1233 supplementary groups returned by the group upcall.
1235 Severity : enhancement
1237 Description: Allow masking D_WARNING, D_ERROR messages from console
1238 Details : Console messages can now be disabled via lnet.printk.
1242 Description: User code with malformed file open parameter crashes client node
1243 Details : Before packing join_file req, all the related reference should be
1244 checked carefully in case some malformed flags cause fake
1245 join_file req on client.
1249 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1250 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1251 mode is equal to the granted mode. However, it should also check
1252 that LDLM_FL_CP_REQD is not set.
1256 Description: Heavy nfs access might result in deadlocks
1257 Details : After ELC code landed, it is now improper to enqueue any mds
1258 locks under och_sem, because enqueue might want to decide to
1259 cancel open locks for same inode we are holding och_sem for.
1263 Description: Client eviction while running blogbench
1264 Details : A lot of unlink operations with concurrent I/O can lead to a
1265 deadlock causing evictions. To address the problem, the number of
1266 oustanding OST_DESTROY requests is now throttled to
1267 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1268 callbacks are processed in priority.
1272 Description: enable ACLs on MDS by default
1273 Details : ACLs must be enabled on MDS by default.
1276 Frequency : PPC/PPC64 only
1278 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1279 Details : fix duplicated definitions between asm-ppc64/types.h and
1280 lustre_types.h on PPC.
1283 Frequency : PPC/PPC64 only
1285 Description: asm-ppc/segment.h does not exist
1286 Details : fix compile issue on PPC.
1290 Description: better handle error messages in extents code
1293 Frequency : RHEL4 only
1295 Description: mkfs is very slow on IA64/RHEL4
1296 Details : A performance regression has been discovered in the MPT Fusion
1297 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1298 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1299 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1301 Severity : enhancement
1303 Description: SNMP support enhancement
1304 Details : Adding total number of sampled request for an MDS node in snmp
1307 Severity : enhancement
1309 Description: Optimize ldlm waiting list processing for PR extent locks
1310 Details : When processing waiting list for read extent lock and meeting read
1311 lock that is same or wider to it that is not contended, skip
1312 processing rest of the list and immediatelly return current
1313 status of conflictness, since we are guaranteed there are no
1314 conflicting locks in the rest of the list.
1318 Description: Time out and refuse to reconnect
1319 Details : When the failover node is the primary node, it is possible
1320 to have two identical connections in imp_conn_list. We must
1321 compare not conn's pointers but NIDs, otherwise we can defeat
1322 connection throttling.
1326 Description: port llog fixes from b1_6 into HEAD
1327 Details : Port llog reference couting and some llog cleanups from b1_6
1328 (bug 10800) into HEAD, for protect from panic and access to already
1329 free llog structures.
1333 Description: Detect stride IO mode in read-ahead
1334 Details : When a client does stride read, read-ahead should detect that and
1335 read-ahead pages according to the detected stride pattern.
1339 Description: data checksumming impacts single node performance
1340 Details : add support for several checksum algorithm. Currently, only CRC32
1341 and Adler-32 are supported. The checksum type can be changed on
1342 the fly via /proc/fs/lustre/osc/*/checksum_type.
1346 Description: use adler32 for page checksums
1347 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1352 Description: build for x2 fails
1353 Details : fix compile issue on Cray systems.
1357 Description: Properly match for duplicate locks
1358 Details : Due to different lock order from skiplists code, we need to
1359 traverse entire list for now
1362 Frequency : only on PPC/SLES10
1364 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1365 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1366 be defined. Add a hack in mkfs_lustre.c to work around this compile
1371 Description: LASSERT on MDS when client holding flock lock dies
1372 Details : ldlm pool logic depends on number of granted locks equal to
1373 number of released locks which is not true for flock locks, so
1374 just exclude such locks from consideration.
1378 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1379 Details : Use fsfilt_sync() for both the whole filesystem sync and
1380 individual file sync to eliminate dangerous inode locking
1381 with I_LOCK that can lead to a deadlock.
1385 Description: performance in 1.6.3
1386 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1387 sure that 1MB requests are not fragmented by the block layer.
1389 Severity : enhancement
1391 Description: organize the server-side client stats on per-nid basis
1392 Details : Change the structure of stats under obdfilter and mds to
1400 The "uuid"s file would list the uuids of _active_ exports.
1401 And the clear entry is to clear all stats and stale nids.
1403 Severity : enhancement
1405 Description: eliminate client locks in face of contention
1406 Details : file contention detection and lockless i/o implementation
1407 for contended files.
1411 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1414 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1416 Description: MDS may lose file striping (and hence file data) in some cases
1417 Details : If there are additional extended attributes stored on the MDS,
1418 in particular ACLs, SELinux, or user attributes (if user_xattr
1419 is specified for the client mount options) then there is a risk
1420 of attribute loss. Additionally, the Lustre file striping
1421 needs to be larger than default (e.g. striped over all OSTs),
1422 and an additional attribute must be stored initially in the
1423 inode and then increase in size enough to be moved to the
1424 external attribute block (e.g. ACL growing in size) for file
1429 Description: skiplist implementation simplification
1430 Details : skiplists are used to group compatible locks on granted list
1431 that was implemented as tracking first and last lock of each
1432 lock group the patch changes that to using doubly linked lists
1436 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1437 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1438 because LASSERT on both the data supplied by a client, and the
1439 data on disk is dangerous and incorrect.
1441 Severity : enhancement
1443 Description: Slow truncate/writes to huge files at high offsets.
1444 Details : Directly associate cached pages to lock that protect those pages,
1445 this allows us to quickly find what pages to write and remove
1446 once lock callback is received.
1450 Description: more ldlm soft lockups
1451 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1452 starve other threads from the resource lock for a long time in
1453 case of long waiting queue, so change the debug level from
1454 D_OTHER to the less frequently used D_INFO.
1456 Severity : enhancement
1458 Description: add -gid, -group, -uid, -user options to lfs find
1462 Description: Hung threads in invalidate_inode_pages2_range
1463 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1464 once one is completed. As a result, some RPCs are stuck in the
1465 queue and are never sent.
1469 Description: filter threads hungs on waiting journal commit
1470 Details : Cleanup filter group llog code, then only filter group llog will
1471 be only created in the MDS/OST syncing process.
1475 Description: Procfs and llog threads access destoryed import sometimes.
1476 Details : Sync the import destoryed process with procfs and llog threads by
1477 the import refcount and semaphore.
1479 Severity : enhancement
1481 Description: openlock cache of b1_6 port to HEAD
1486 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1487 Details : If insertion of an extent fails, then discard the inode
1488 preallocation and free data blocks else it can lead to duplicate
1493 Description: don't always update ctime in ext3_xattr_set_handle()
1494 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1495 In some cases the ctime should not be updated, for example for
1496 2.0->1.8 compatibility it is necessary to delete an xattr and it
1497 should not update the ctime.
1501 Bugzilla : 15713/16362
1502 Description: Assertion in iopen_connect_dentry in 1.6.3
1503 Details : looking up an inode via iopen with the wrong generation number can
1504 populate the dcache with a disconneced dentry while the inode
1505 number is in the process of being reallocated. This causes an
1506 assertion failure in iopen since the inode's dentry list contains
1507 both a connected and disconnected dentry.
1511 Description: assertion failure in ldlm_handle2lock()
1512 Details : fix a race between class_handle_unhash() and class_handle2object()
1513 introduced in lustre 1.6.5 by bug 13622.
1518 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1519 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1524 Description: oops in sd_iostats_seq_show()
1525 Details : unloading/reloading the scsi low level driver triggers a kernel
1526 bug when trying to access the sd iostat file.
1531 Description: Kernel panics during QLogic driver reload
1532 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1533 patch, causing memory corruption.
1538 Description: journal_dev option does not work in b1_6
1539 Details : pass mount option during pre-mount.
1541 Severity : enhancement
1543 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1544 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1545 extent information of a file. It can be used to map logical blocks
1546 in a file to physical blocks in the block device.
1550 Description: LDLM soft lockups - improvement
1551 Details : It is be possible to send the lock handle along with each read
1552 or write request because the client is already doing a lock match
1553 itself so there isn't any reason the OST should have to re-do that
1557 Frequency : only on Cray X2
1559 Description: X2 build failures
1560 Details : fix build failures on Cray X2.
1564 Description: xid & resent requests
1565 Details : Initialize RPC XID from clock at startup (randomly if clock is
1568 Severity : enhancement
1570 Description: Add lustre_start utility to start or stop multiple Lustre servers
1575 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1576 Details : In case of memory pressure, list_del() can be called twice on
1577 req->rq_history_list, causing a kernel oops.
1581 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1582 Details : fix stack overflow in the distributed lock manager by defering
1583 export eviction after a failed AST to the elt thread instead of
1584 handling it in the dlm interpret routine.
1588 Description: Convert some comments to new format.
1589 Details : Update documenting comments to match doxygen conventions.
1593 Description: Grammar fixes.
1594 Details : A couple of trivial spelling fixes.
1598 Description: OSD_COUNTERS-mandatory
1599 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1600 to simplify the code.
1604 Description: simplify cmm_device freeing logic.
1605 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1609 Description: Add lockdep support to dt_object_operations locking interface.
1610 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1611 parameter indicating lock ordering. Update mdd code to use new
1616 Description: Introduce failloc constants for lockless IO tests.
1617 Details : Add two new failloc constants to test lockless IO. Only one of
1618 them in implemented---another is checked in yet to be landed
1623 Description: Add lockdep support for inode mutex.
1624 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1625 in the situations where Lustre has to lock more than one inode
1630 Description: Add optional invariants checking support.
1631 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1632 configure switch. This macro is to be used for consistency and
1633 sanity checks that are too expensive to be left in `production'
1638 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1639 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1644 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1645 Details : Use spin_lock_nested() in (the only) situation where more than
1646 one ldlm_lock is locked simultaneously. Also, fix possible
1647 dead-lock in ldlm_lock_change_resource() by enforcing particular
1652 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1653 Details : Instead of specifying each ldlm_lock call-back through separate
1654 parameter, wrap them into struct ldlm_callback_suite.
1658 Description: Kill join_lru obd method and its callers.
1659 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1660 in memory---a requirement implemented through ->o_join_lru() obd
1661 method in HEAD. Remove this method and its users.
1665 Description: Add asynchronous ldlm ENQUEUE completion handler.
1666 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1667 case that stock ldlm_completion_ast() cannot handle as it waits
1668 until lock is granted. Introduce new ldlm_completion_ast_async()
1669 for this. Also comment ldlm_completion_ast().
1673 Description: ldlm_error <-> errno conversion.
1674 Details : Add functions to map (rather arbitrary) between LDLM error codes
1675 and standard errno values. CLIO needs this to prevent LDLM specific
1676 constants from escaping ldlm and osc.
1680 Description: Kill unused ldlm_handle2lock_ns() function.
1681 Details : Kill unused ldlm_handle2lock_ns() function.
1685 Description: Add lu_ref support to ldlm_lock
1686 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1687 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1688 and ldlm_resource. LDLM interface has to be changed, because code that
1689 releases a reference on a lock, has to "know" what reference this is.
1690 In the most frequent case
1692 lock = ldlm_handle2lock(handle);
1694 LDLM_LOCK_PUT(lock);
1696 no changes are required. When any other reference (received _not_ from
1697 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1698 instead of LDLM_LOCK_PUT().
1700 Arguably, changes are pervasive, and interface requires some discipline
1701 for proper use. On the other hand, it was very instrumental in finding
1702 a few leaked lock references.
1706 Description: Add ldlm_lock_addref_try().
1707 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1708 attempts to addref a lock that might be being canceled
1713 Description: Add ldlm_weigh_callback().
1714 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1715 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1716 hard-coded `number of pages' logic.
1720 Description: Add lockdep annotations to llog code.
1721 Details : Use appropriately tagged _nested() locking calls in the places
1722 where llog takes more than one ->lgh_lock lock.
1726 Description: Add loi_kms_set().
1727 Details : Wrap kms updates into a helper function.
1731 Description: Constify instances of struct lsm_operations.
1732 Details : Constify instances of struct lsm_operations.
1736 Description: lu_conf support.
1737 Details : On a server, a file system object is uniquely identified
1738 by a fid, which is sufficient to locate and load all object
1739 state (inode). On a client, on the other hand, more data are
1740 necessary instantiate an object. Change lu_object_find() and
1741 friends to take additional `lu_conf' argument describing object.
1742 Typically this includes layout information.
1746 Description: lu_context fixes.
1747 Details : Introduce new lu_context functions that are needed on the client
1748 side, where some system threads (ptlrpcd) are shared by multiple
1749 modules, and so cannot be stopped during module shutdown.
1753 Description: Add start and stop methods to lu_device_type_operations.
1754 Details : Introduce two new methods in lu_device_type_operations, that are
1755 invoked when first instance of a given type is created and last one
1756 is destroyed respectively. This is need by CLIO.
1760 Description: Add lu_ref support to struct lu_device.
1761 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1762 track leaked references.
1766 Description: Introduce lu_kmem_descr.
1767 Details : lu_kmem_descr and its companion interface allow to create
1768 and destroy a number of kmem caches at once.
1772 Description: Fix lu_object finalization race.
1773 Details : Fix a race between lu_object_find() finding an object and its
1774 concurrent finalization. This race is (most likely) not possible
1775 on the server, but might happen on the client.
1779 Description: Introduce lu_ref interface.
1780 Details : lu_ref is a debugging module allowing to track references to
1781 a given object. It is quite cpu expensive, and has to be
1782 explicitly enabled with --enable-lu_ref. See usage description
1787 Description: Factor lu_site procfs stats into a separate function.
1788 Details : Separate lu_site stats printing code into a separate function
1789 that can be reused on a client.
1793 Description: Constify instances of struct {lu,dt,md}_device_operations.
1794 Details : Constify instances of struct {lu,dt,md}_device_operations.
1798 Description: Introduce struct md_site and move meta-data specific parts of
1799 struct lu_site here.
1800 Details : Move md-specific fields out of struct lu_site into special struct
1801 md_site, so that lu_site can be used on a client.
1805 Description: Kill mdd_lov_destroy().
1806 Details : Remove unused mdd code.
1810 Description: Add st_block checking to multistat.c.
1811 Details : Add st_block checking to multistat.c.
1815 Description: Add lu_ref support to struct obd_device.
1816 Details : Add lu_ref tracking to obd_device.
1820 Description: Kill obd_set_fail_loc().
1821 Details : Remove unused code.
1825 Description: Add special type for ptlrpc_request interpret functions.
1826 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1827 there. Actual usage will be in CLIO.
1831 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1832 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1833 doesn't work with lockdep.
1837 Description: Add rwv.c test program.
1838 Details : New testing program exercising readv(2) and writev(2) (Qian).
1842 Description: Add sendfile.c test program.
1843 Details : New testing program exercising sendfile(2) (Jay).
1847 Description: Ratelimit a message that can be very frequent.
1848 Details : Ratelimit a memory allocation failure message that can
1853 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1854 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1858 Description: Liblustre build fixes.
1859 Details : Liblustre build fixes.
1863 Description: libcfs: add cfs_{need,cond}_resched() interface.
1864 Details : libcfs: add cfs_{need,cond}_resched() definition and
1865 implementations for Linux, NT, and liblustre.
1867 Severity : enhancement
1869 Description: More exported tunables for mballoc
1870 Details : Add support for tunable preallocation window and new tunables for
1871 large/small requests
1875 Description: Detect corruption of block bitmap and checking for preallocations
1876 Details : Checks validity of on-disk block bitmap. Also it does better
1877 checking of number of applied preallocations. When corruption is
1878 found, it turns filesystem readonly to prevent further corruptions.
1882 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1883 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1884 window to 0. If it does hit cache miss, and read-pattern is still
1885 stride-io mode, does not reset the stride window, but also does not
1886 increase the stride window length in this case.
1890 Frequency : only for big-endian servers
1891 Description: Check if system is big-endian while mounting fs with extents feature
1892 Details : Mounting a filesystem with extents feature will fail on big-endian
1893 systems since ext3-based ldiskfs is not supported on big-endian
1894 systems. This can be over-riden with "bigendian_extents" mount option.
1896 Severity : enhancement
1898 Description: The root squash functionality
1899 Details : A security feature, which is to prevent users from being able
1900 to mount lustre on their desktop, run as root, and delete
1901 all of the files in the filesystem. The goal is accomplished by
1902 remapping user id (UID) and group id (GID) of the root user to
1903 a UID and GID specified by the system administartor via Lustre
1904 configuration management server (MGS). The functionality also
1905 allows to specify sets of clients for which the remapping does
1910 Description: Excessive recovery window
1911 Details : With AT enabled, the recovery window can be excessively long (6000+
1912 seconds). To address this problem, we no longer use
1913 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1914 timeout no longer depends on the service time, it is set to
1915 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1916 time via pb_service_time.
1920 Description: Watchdog triggered on MDS failover
1921 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1922 the OSTs know that the MDS "UUID" can be reused for the same export
1923 from a different NID, so we do not need to wait for the export to be
1926 --------------------------------------------------------------------------------
1928 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1930 * Support for kernels:
1931 2.6.5-7.283 (SLES 9),
1932 2.6.9-55.EL (RHEL 4),
1933 2.6.16.46-0.14 (SLES 10),
1934 2.6.18.8 vanilla (kernel.org)
1935 * Client support for unpatched kernels:
1936 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1937 2.6.16 - 2.6.22 vanilla (kernel.org)
1938 * Due to recently discovered recovery problems, we do not recommend
1939 using patchless RHEL 4 clients with this or any earlier release.
1940 * Recommended e2fsprogs version: 1.39.cfs8
1941 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1942 * Starting with this release, the ldiskfs backing filesystem required
1943 by Lustre is now in its own package, lustre-ldiskfs. This package
1944 should be installed. It is versioned separately from Lustre and
1945 may be released separately in future.
1949 Description: block reactivating mgc import until all deactivates complete
1950 Details : Fix race when failing back MDT/MGS to itself (testing)
1953 Frequency : at statup only
1955 Description: mds_lov_synchronize race leads to various problems
1956 Details : simultaneous MDT->OST connections at startup can cause the
1957 sync to abort, leaving the OSC in a bad state.
1959 Severity : enhancement
1961 Description: add optional extra BUILD_VERSION info
1962 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1963 to override the lustre version.
1966 Frequency : 2.6.18 servers only
1968 Description: ll_kern_mount() doesn't release the module reference
1969 Details : The ldiskfs module reference count never drops down to 0
1970 because ll_kern_mount() doesn't release the module reference.
1975 Description: server LBUG when using old ost_num_threads parameter
1976 Details : Accept the old ost_num_threads parameter but warn that it
1977 is deprecated, and fix an off-by-one error that caused an LBUG.
1982 Description: Transient SCSI error results in persistent IO issue
1983 Details : iobuf->dr_error is not reinitialized to 0 between two
1987 Frequency : sometimes when underlying device returns I/O errors
1989 Description: OSTs not going read-only during write failures
1990 Details : OSTs are not remounted read-only when the journal commit threads
1991 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1992 instead of the ext3 wrappers.
1996 Description: poor connect scaling with increasing client count
1997 Details : Don't run filter_grant_sanity_check for more than 100 exports
1998 to improve scaling for large numbers of clients.
2001 Frequency : SLES10 only
2003 Description: sanity-quota.sh quotacheck failed: rc = -22
2004 Details : Quotas cannot be enabled on SLES10.
2007 Frequency : liblustre clients only
2009 Description: getdirentries does not give error when run on compute nodes
2010 Details : getdirentries does not fail when the size specified as an argument
2011 is too small to contain at least one entry
2013 Severity : enhancement
2015 Description: Add LNET router traceability for debug purposes
2016 Details : If a checksum failure occurs with a router as part of the
2017 IO path, the NID of the last router that forwarded the bulk data
2018 is printed so it can be identified.
2023 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2024 Details : Due to a race condition, liblustre clients were occasionally
2025 evicted incorrectly.
2027 Severity : enhancement
2029 Description: lfs setstripe use optional parameters instead of postional
2032 Severity : enhancement
2034 Description: Nanosecond timestamp support for ldiskfs
2035 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2036 resolution timestamps. There is not yet support for this at
2037 the Lustre filesystem level.
2040 Frequency : during server recovery
2042 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2043 Details : request with rq_no_resend flag not awake l_wait_event if they get
2047 Frequency : nfs export on patchless client
2049 Description: connectathon hang when test nfs export over patchless client
2050 Details : Disconnected dentry cannot be found with lookup, so we do not need
2051 to unhash it or make it invalid
2054 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2055 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2056 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2061 Description: open req refcounting wrong on reconnect
2062 Details : If reconnect happened between getting open reply from server and
2063 call to mdc_set_replay_data in ll_file_open, we will schedule
2064 replay for unreferenced request that we are about to free.
2065 Subsequent close will crash in variety of ways.
2066 Check that request is still eligible for replay in
2067 mdc_set_replay_data().
2072 Description: disable writes to filesystem when reading health_check file
2073 Details : the default for reading the health_check proc file has changed
2074 to NOT do a journal transaction and write to disk, because this
2075 can cause reads of the /proc file to hang and block HA state
2076 checking on a healthy but otherwise heavily loaded system. It
2077 is possible to return to the previous behaviour during configure
2078 with --enable-health-write.
2080 Severity : enhancement
2082 Description: 64-bit inode version
2083 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2084 to the inode. This will be required for version-based recovery.
2089 Description: MDS fails to start if a duplicate client export is detected
2090 Details : in some rare cases it was possible for a client to connect to
2091 an MDS multiple times. Upon recovery the MDS would detect this
2092 and fail during startup. Handle this more gracefully.
2094 Severity : enhancement
2096 Description: Add -o localflock option to simulate old noflock
2098 Details : This will achieve local-only flock/fcntl locks
2104 Description: log_commit_thread vs filter_destroy race leads to crash
2105 Details : Take import reference before releasing llog record semaphore
2110 Description: Wrong request locking in request set processing
2111 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2112 imp_delayed_list, in this place should be used imp_lock.
2115 Frequency : when reconnection
2117 Description: Grant Leak when osc reconnect to OST
2118 Details : When osc reconnect ost, OST(filter) should check whether it
2119 should grant more space to client by comparing fed_grant and
2120 cl_avail_grant, and return the granted space to client instead
2121 of "new granted" space, because client will call osc_init_grant
2122 to update the client grant space info.
2125 Frequency : when client reconnect to OST
2127 Description: Grant Leak when osc do resend and replay bulk write
2128 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2129 bulk write request, because the grant info while be sync between
2130 OSC and OST when reconnect, and we should ignore the grant info
2131 these of resend/replay write req.
2136 Description: Grant space more than avaiable left space sometimes.
2137 Details : When then OST is about to be full, if two bulk writing from
2138 different clients came to OST. Accord the avaliable space of the
2139 OST, the first req should be permitted, and the second one
2140 should be denied by ENOSPC. But if the seconde arrived before
2141 the first one is commited. The OST might wrongly permit second
2142 writing, which will cause grant space > avaiable space.
2145 Frequency : when client is evicted
2147 Description: Grant might be wrongly erased when osc is evicted by OST
2148 Details : when the import is evicted by server, it will fork another
2149 thread ptlrpc_invalidate_import_thread to invalidate the
2150 import, where the grant will be set to 0. While the original
2151 thread will update the grant it got when connecting. So if
2152 the former happened latter, the grant will be wrongly errased
2153 because of this race.
2158 Description: Checking Stale with correct fid
2159 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2160 to check whether it is stale, because sometimes, we want the
2161 enqueue happened anyway, and op_data.fid2 will not be initialized.
2163 Severity : enhancement
2165 Description: update patchless client
2166 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2169 Frequency : only with 2.4 kernel
2171 Description: random memory corruption
2172 Details : size of struct ll_inode_info is to big for union inode.u and this
2173 can be cause of random memory corruption.
2178 Description: Memory leak in recovery
2179 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2180 It should also check obd_fail before fsfilt_start, otherwise if
2181 fsfilt_start return -EROFS,(failover mds during mds recovery).
2182 then the req will return with repmsg->transno = 0 and rc = EROFS.
2183 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2184 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2185 be freed no matter whether fsfilt_commit success or not.
2188 Frequency : only with huge count clients
2190 Description: Prevents from taking the superblock lock in llap_from_page for
2192 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2193 for llap_from_page call in ll_removepage prevents from taking the
2194 superblock lock for a soon died page.
2199 Description: Not check open intent error before release open handle
2200 Details : in some rare cases, the open intent error is not checked before
2201 release open handle, which may cause
2202 ASSERTION(open_req->rq_transno != 0), because it tries to release
2203 the failed open handle.
2208 Description: Set cat log bitmap only after create log success.
2209 Details : in some rare cases, the cat log bitmap is set too early. and it
2210 should be set only after create log success.
2214 Description: Accessing a block bevice can re-enable I/O when Lustre is
2215 tearing down a device.
2216 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2221 Description: service threads may hog cpus when there are a lot of requests
2222 Details : Insert cond_resched to give other threads a chance to use some CPU
2227 Description: the cat log was not initialized in recovery
2228 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2229 unlink log on mds will not be initialized until mds post
2230 recovery. And also in mds post recovery, the unlink log will
2231 initialization will be done asynchronausly, so there will be race
2232 between add unlink log and unlink log initialization.
2236 Description: brw_stats were being printed incorrectly
2237 Details : brw_stats were being printed as log2 but all of them were not
2238 recorded as log2. Also remove some code duplication arising from
2239 filter_tally_{read,write}.
2243 Frequency : rare, only in recovery.
2244 Description: ASSERTION(req->rq_type != LI_POISON) failed
2245 Details : imp_lock should be held while iterating over imp_sending_list for
2246 prevent destroy request after get timeout in ptlrpc_queue_wait.
2250 Description: replay-single.sh test 52 fails
2251 Details : A lock's skiplist need to be cleanup when it being unlinked
2252 from its resource list.
2256 Description: Short directio read returns full requested size rather than
2258 Details : Direct I/O operations should return actual amount of bytes
2259 transferred rather than requested size.
2261 Severity : enhancement
2263 Description: metadata RPC reduction (e.g. for rm performance)
2264 Details : decrease the amount of synchronous RPC between clients and servers
2265 by canceling conflicing lock before the operation on the client
2266 and packing thier handles into the main operation RPC to server.
2268 Severity : enhancement
2270 Description: Async OSC create to avoid the blocking unnecessarily.
2271 Details : If a OST has no remain object, system will block on the creating
2272 when need to create a new object on this OST. Now, ways use
2273 pre-created objects when available, instead of blocking on an
2274 empty osc while others are not empty. If we must block, we block
2275 for the shortest possible period of time.
2279 Description: improve handling recoverable errors
2280 Details : if request processig with error which can be recoverable on server
2281 request should be resend, otherwise page released from cache and
2284 Severity : enhancement
2286 Description: refine locking for avoid write wrong info into lov_objid file
2287 Details : fix possible races with add new target and write/update data in
2290 --------------------------------------------------------------------------------
2292 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2297 Frequency : on some architectures
2299 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2300 Details : Uninitialized flags sometimes cause configuration commands to
2304 Frequency : patchless clients only
2306 Description: missing __iget() symbol export
2307 Details : The __iget() symbol export is missing. To avoid the need for
2308 this on patchless clients the deathrow inode reaper is turned
2309 off, and we depend on the VM to clean up old inodes. This
2310 dependency was during via the fix for bug 12181.
2312 --------------------------------------------------------------------------------
2314 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2316 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2317 INTEROPERATE with older versions automatically. In many cases a
2318 special upgrade step is needed. Please read the
2319 user documentation before upgrading any part of a live system.
2320 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2321 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2322 * WARNING: Lustre configuration and startup changes are required with
2323 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2327 Severity : enhancement
2329 Description: MountConf
2330 Details : Lustre configuration is now managed via mkfs and mount
2331 commands instead of lmc and lconf. New obd types (MGS, MGC)
2332 are added for dynamic configuration management. See
2333 https://mail.clusterfs.com/wikis/lustre/MountConf for
2336 Severity : enhancement
2338 Description: dynamic OST addition
2339 Details : OSTs can now be added to a live filesystem
2341 Severity : enhancement
2343 Description: startup order invariance
2344 Details : MDTs and OSTs can be started in any order. Clients only
2345 require the MDT to complete startup.
2347 Severity : enhancement
2349 Description: parallel, asynchronous orphan cleanup
2350 Details : orphan cleanup is now performed in separate threads for each
2351 OST, allowing parallel non-blocking operation.
2353 Severity : enhancement
2355 Description: optimized stripe assignment
2356 Details : stripe assignments are now made based on ost space available,
2357 ost previous usage, and OSS previous usage, in order to try
2358 to optimize storage space and networking resources.
2360 Severity : enhancement
2362 Description: Permanently set tunables
2363 Details : All writable /proc/fs/lustre tunables can now be permanently
2364 set on a per-server basis, at mkfs time or on a live system.
2366 Severity : enhancement
2368 Description: Lustre message v2
2369 Details : Add lustre message format v2.
2371 Severity : enhancement
2373 Description: client OST exclusion list
2374 Details : Clients can be started with a list of OSTs that should be
2375 declared "inactive" for known non-responsive OSTs.
2379 Description: SPEC SFS validation failure on NFS v2 over lustre.
2380 Details : Changes the blocksize for regular files to be 2x RPC size,
2381 and not depend on stripe size.
2383 Severity : enhancement
2385 Description: Multiple MD RPCs in flight.
2386 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2387 To avoid overly-overloading MDS, introduce a limit on number of
2388 MDS RPCs in flight for a single client and add /proc controls
2389 to adjust this limit.
2391 Severity : enhancement
2393 Description: client read/write statistics
2394 Details : Add client read/write call usage stats for performance
2395 analysis of user processes.
2396 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2397 file access. extents_stats shows chunk size distribution.
2398 extents_stats_per_process show chunk size distribution per
2401 Severity : enhancement
2403 Description: mds statistics
2404 Details : Add detailed mds operations statistics in
2405 /proc/fs/lustre/mds/*/stats.
2409 Description: Failure of copying files with lustre special EAs.
2410 Details : Client side always return success for setxattr call for lustre
2411 special xattr (currently only "trusted.lov").
2416 Description: Refcount LNET uuids
2417 Details : The global LNET uuid list grew linearly with every startup;
2418 refcount repeated list entries instead of always adding to
2421 Severity : enhancement
2423 Description: Dynamic service threads
2424 Details : Within a small range, start extra service threads
2425 automatically when the request queue builds up.
2428 Frequency : mixed-endian client/server environments
2430 Description: mixed-endian crashes
2431 Details : The new msg_v2 system had some failures in mixed-endian
2434 Severity : enhancement
2436 Description: Easy OST removal
2437 Details : OSTs can be permanently deactivated with e.g. 'lctl
2438 conf_param lustre-OST0001.osc.active=0'
2440 Severity : enhancement
2442 Description: MGS proc entries
2443 Details : Added basic proc entries for the MGS showing what filesystems
2446 Severity : enhancement
2448 Description: provide MGS failover
2449 Details : Added config lock reacquisition after MGS server failover.
2451 Severity : enhancement
2453 Description: add Linux 2.4 support
2454 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2458 Description: a large application tries to do I/O to the same resource and dies
2459 in the middle of it.
2460 Details : Check the req->rq_arrival time after the call to
2461 ost_brw_lock_get(), but before we do anything about
2462 processing it & sending the BULK transfer request. This
2463 should help move old stale pending locks off the queue as
2464 quickly as obd_timeout.
2467 Frequency : when an incorrect nid is specified during startup
2469 Description: ptlrpc connect to non-existant node causes kernel crash
2470 Details : LNET can't be re-entered from an event callback, which
2471 happened when we expire a message after the export has been
2472 cleaned up. Instead, hand the zombie cleanup off to another
2475 Severity : enhancement
2477 Description: plain/inodebits lock performance improvement
2478 Details : Grouping plain/inodebits in granted list by their request modes
2479 and bits policy, thus improving the performance of search through
2483 Frequency : only if OST filesystem is corrupted
2485 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2486 Details : for a short time RPCs with bulk IO are in the replay list,
2487 but replay of bulk IOs is unimplemented. If the OST filesystem
2488 is corrupted due to disk cache incoherency and then replay is
2489 started it is possible to trip an assertion. Avoid putting
2490 committed RPCs into the replay list at all to avoid this issue.
2493 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2495 Description: System hang on startup
2496 Details : This bug allowed the liblustre (e.g. catamount) client to
2497 return to the app before handling all startup RPCs. This
2498 could leave the node unresponsive to lustre network traffic
2499 and manifested as a server ptllnd timeout.
2501 Severity : enhancement
2503 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2504 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2505 boolean to control whether to print peer debug info when a
2506 client's RPC times out.
2509 Frequency : only for kernels with patches from Lustre below 1.4.3
2511 Description: Remove old rdonly API
2512 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2515 Frequency : only for devices with external journals
2517 Description: Set external device read-only also
2518 Details : During a commanded failover stop, we set the disk device
2519 read-only while the server shuts down. We now also set any
2520 external journal device read-only at the same time.
2523 Frequency : when upgrading from 1.4 while trying to change parameters
2525 Description: The wrong (new) MDC name was used when setting parameters for
2526 upgraded MDT's. Also allows changing of OSC (and MDC)
2527 parameters if --writeconf is specified at tunefs upgrade time.
2530 Frequency : when setting specific ost indicies
2532 Description: QOS code breaks on skipped indicies
2533 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2534 created with --index need not be sequential.
2536 Severity : enhancement
2538 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2539 Details : The uninit_groups feature works in conjunction with the kernel
2540 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2541 up the pass1 processing of e2fsck. This is a read-only feature
2542 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2543 mount filesystems that have had this feature enabled.
2545 Severity : enhancement
2547 Description: Improve multi-block allocation algorithm to avoid fragmentation
2548 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2549 allocation locality and avoid filesystem fragmentation.
2551 ------------------------------------------------------------------------------
2553 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2555 * Support for kernels:
2556 2.4.21-47.0.1.EL (RHEL 3)
2557 2.6.5-7.283 (SLES 9)
2558 2.6.9-42.0.10.EL (RHEL 4)
2559 2.6.12.6 vanilla (kernel.org)
2560 2.6.16.27-0.9 (SLES 10)
2561 * Recommended e2fsprogs version: 1.39.cfs5
2563 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2567 Frequency : occasional, depends on client load and configuration
2568 Bugzilla : 12181, 12203
2569 Description: data loss for recently-modified files
2571 Details : In some cases it is possible that recently written or created
2572 files may not be written to disk in a timely manner (this should
2573 normally be within 30s unless client IO load is very high).
2574 The problem appears as zero-length files or files that are a
2575 multiple of 1MB in size after a client crash or client eviction
2576 that are missing data at the end of the file.
2578 This problem is more likely to be hit on clients where files are
2579 repeatedly created and unlinked in the same directory, clients
2580 have a large amount of RAM, have many CPUs, the filesystem has
2581 many OSTs, the clients are rebooted frequently, and/or the files
2582 are not accessed by other nodes after being written.
2584 The presence of the problem can be detected by looking at
2585 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2586 smaller than the second (nr_unused) then dirty files will not
2587 be flushed automatically to disk. "sync; sleep 10" should be
2588 run several times on the node before unmounting it to update
2589 Lustre (this is also safe to run on nodes without this problem).
2591 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2592 that can cause this same problem, so customers using that kernel
2593 also need to update the kernel in addition to Lustre. In order
2594 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2596 It is normal that files written just before a client crash (less
2597 than 30s) may not yet have been flushed to disk, even for local
2601 Frequency : frequent on thin XT3 nodes
2603 Description: UUID collision on thin XT3 Linux nodes
2604 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2605 randomly, since we relied on an insufficiently-seeded PRNG.
2610 Description: OSS hangs after "All ost request buffers busy"
2611 Details : A deadlock between quota and journal operations caused OSS
2612 hangs after printing "All ost request buffers busy."
2615 Frequency : always on liblustre builds
2617 Description: Cleanup compiler warnings on liblustre
2620 Frequency : always on liblustre builds on XT3
2622 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2626 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2627 > 32000 subdirectories
2628 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2629 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2630 subdirectory count crosses 32000. This will aid e2fsck to
2631 correctly handle more than 32000 subdirectories.
2634 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2636 Description: System hang on startup
2637 Details : This bug allowed the liblustre (e.g. catamount) client to
2638 return to the app before handling all startup RPCs. This
2639 could leave the node unresponsive to lustre network traffic
2640 and manifested as a server ptllnd timeout.
2642 Severity : enhancement
2644 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2645 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2646 boolean to control whether to print peer debug info when a
2647 client's RPC times out.
2652 Description: make O_SYNC working on 2.6 kernels
2653 Details : 2.6 kernels use different method for mark pages for write,
2654 so need add a code to lustre for O_SYNC work.
2659 Description: Failure to close file and release space on NFS
2660 Details : Put inode details into lock acquired in ll_intent_file_open.
2661 Use mdc_intent_lock in ll_intent_open to properly
2662 detect all kind of errors unhandled by mdc_enqueue.
2667 Description: proc file read during shutdown sometimes raced obd removal,
2669 Details : Add lock to prevent obd access after proc file removal.
2672 Frequency : Only for files larger than 4GB on 32-bit clients.
2674 Description: improperly doing page alignment of locks
2675 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2676 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2681 Description: RPCs being resent when they shouldn't be
2682 Details : Some RPCs that should not be resent are being resent. This
2683 can cause inconsistencies in the RPC state machine. Do not
2684 resend such requests.
2687 Frequency : rare, only with NFS export
2689 Description: Crash on NFS re-export node
2690 Details : under very unusual load conditions an assertion is hit in
2691 ll_intent_file_open()
2694 Frequency : only if OST filesystem is corrupted
2696 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2697 Details : for a short time RPCs with bulk IO are in the replay list,
2698 but replay of bulk IOs is unimplemented. If the OST filesystem
2699 is corrupted due to disk cache incoherency and then replay is
2700 started it is possible to trip an assertion. Avoid putting
2701 committed RPCs into the replay list at all to avoid this issue.
2706 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2707 Details : Large single O_DIRECT read and write calls can fail to allocate
2708 a sufficiently large buffer to process the request. In case of
2709 allocation failure the allocation is retried with a smaller
2710 buffer and broken into smaller requests.
2712 Severity : enhancement
2714 Description: Add -o localflock option to simulate old noflock behaviour.
2715 Details : This will achieve local-only flock/fcntl locks coherentness.
2720 Description: versioning check is incomplete
2721 Details : Checking the version difference of client vs. server, report
2722 error if the gap is too big.
2727 Description: add support PG_writeback bit
2728 Details : add support for PG_writeback bit for Lustre, for more carefull
2729 work with page cache in 2.6 kernel. This also fix some deadlocks
2730 and remove hack for work O_SYNC with 2.6 kernel.
2732 Severity : enhancement
2734 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2735 Details : The uninit_groups feature works in conjunction with the kernel
2736 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2737 up the pass1 processing of e2fsck. This is a read-only feature
2738 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2739 mount filesystems that have had this feature enabled.
2741 Severity : enhancement
2743 Description: Improve multi-block allocation algorithm to avoid fragmentation
2744 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2745 allocation locality and avoid filesystem fragmentation.
2747 ------------------------------------------------------------------------------
2749 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2751 * Support for kernels:
2752 2.6.9-42.0.3.EL (RHEL 4)
2753 2.6.5-7.276 (SLES 9)
2754 2.4.21-47.0.1.EL (RHEL 3)
2755 2.6.12.6 vanilla (kernel.org)
2756 2.6.16.21-0.8 (SLES10)
2757 * Recommended e2fsprogs version: 1.39.cfs2-0
2759 * The backwards-compatible /proc/sys/portals symlink has been removed
2760 in this release. Before upgrading, please ensure that you change
2761 any configuration scripts or /etc/sysctl.conf files that access
2762 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2763 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2764 in advance of the upgrade on any system running Lustre 1.4.6 or
2765 newer, since /proc/sys/lnet was added in that version.
2766 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2770 Frequency : only when quota is used
2772 Description: avoid scanning export list for quota master
2773 Details : Change the algorithms to avoid scanning export list in order
2774 to improve the efficiency.
2777 Frequency : MDS failover only, very rarely
2779 Description: "went back in time" messages on mds failover
2780 Details : The greatest transno may be lost when the current operation
2781 finishes with an error (transno==0) and the client's last_rcvd
2782 record is over-written. Save the greatest transno in the
2783 mds_last_transno for this case.
2786 Frequency : always for specific kernels and striping counts
2788 Description: client may get "Matching packet too big" without ACL support
2789 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2790 when trying to access files in certain configurations. The
2791 clients should in fact be denied when mounting because they do
2792 not understand ACLs.
2795 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2797 Description: many clients connecting with IO in progress causes connect timeouts
2798 Details : Avoid synchronous journal commits to avoid delays caused by many
2799 clients connecting/disconnecting when bulk IO is in progress.
2800 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2801 OST_IO_PORTAL to avoid delays behind potentially many pending
2805 Frequency : occasionally with multiple writers to a single file
2807 Description: shared writes to file may result in wrong size reported by stat()
2808 Details : Allow growing of kms when extent lock is cancelled
2811 Frequency : always with random mmap IO to multi-striped file
2813 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2814 Details : Only if the hole is at the end of OST object so that kms is too
2815 small. Fix is to increase kms accordingly in ll_nopage.
2818 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2820 Description: writes to a missing object would leak memory on the OST
2821 Details : If there is an inconsistency between the MDS and OST filesystems,
2822 such that the MDS references an object that doesn't exist, writes
2823 to that object will leak memory due to incorrect cleanup in the
2824 error handling path, eventually running out of memory on the OST.
2829 Description: Creating too long symlink causes lustre errors
2830 Details : Check symlink and name lengths before sending requests to MDS.
2833 Frequency : only if flock is enabled (not on by default)
2835 Description: posix locks not released on fd closure on 2.6.9+
2836 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2837 kernels, this caused such locks not to be released on fd close and
2838 then assertions on fs unmount about still used locks.
2841 Frequency : MDS failover only, very rarely
2843 Description: clients may get ASSERTION(granted_lock != NULL)
2844 Details : When request was taking a long time, and a client was resending
2845 a getattr by name lock request. The were multiple lock requests
2846 with the same client lock handle and
2847 mds_getattr_name->fixup_handle_for_resent_request found one of the
2848 lock handles but later failed with ASSERTION(granted_lock != NULL).
2853 Description: handle->h_buffer_credits > 0, assertion failure
2854 Details : h_buffer_credits is zero after truncate, causing assertion
2855 failure. This patch extends the transaction or creates a new
2859 Frequency : NFS re-export or patchless client
2860 Bugzilla : 11179, 10796
2861 Description: Crash on NFS re-export node (__d_move)
2862 Details : We do not want to hash the dentry if we don't have a lock.
2863 But if this dentry is later used in d_move, we'd hit uninitialised
2864 list head d_hash, so we just do this to init d_hash field but
2865 leave dentry unhashed.
2868 Frequency : NFS re-export or patchless client
2870 Description: NFS exports has problem with symbolic link
2871 Details : lustre client didn't properly install dentry when re-exported
2872 to NFS or running patchless client.
2875 Frequency : NFS re-export or patchless client
2877 Description: Various nfs/patchless fixes.
2878 Details : fixes reuse disconected alias for lookup process - this fixes
2879 warning "find_exported_dentry: npd != pd",
2880 fix permission error with open files at nfs.
2881 fix apply umask when do revalidate.
2884 Frequency : occasional
2886 Description: Crash on NFS re-export node
2887 Details : calling clear_page() on the wrong pointer triggered oops in
2888 generic_mapping_read().
2891 Frequency : rarely, using O_DIRECT IO
2893 Description: unaligned directio crashes client with LASSERT
2894 Details : check for unaligned buffers before trying any requests.
2897 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2899 Description: stale data returned from RAID cache
2900 Details : If only a small amount of IO is done to the RAID device before
2901 reading it again it is possible to get stale data from the RAID
2902 cache instead of reading it from disk.
2905 Frequency : always for sles10 kernel
2907 Description: sles10 support
2908 Details : ll_follow_link: compile fixes and using of nd_set_link
2909 under newer kernels.
2912 Frequency : depends on arch, kernel and compiler version, always on sles10
2915 Description: recursive or deep enough symlinks cause stack overflow
2916 Details : getting rid of large stack-allocated variable in
2920 Frequency : depends on hardware
2922 Description: lustre write performance loss in the SLES10 kernel
2923 Details : the performance loss is caused by using of write barriers in the
2924 ext3 code. The SLES10 kernel turns barrier support on by
2925 default. The fix is to undo that change for ldiskfs.
2927 ------------------------------------------------------------------------------
2929 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2931 * Support for kernels:
2932 2.6.9-42.0.3EL (RHEL 4)
2933 2.6.5-7.276 (SLES 9)
2934 2.4.21-47.0.1.EL (RHEL 3)
2935 2.6.12.6 vanilla (kernel.org)
2939 Frequency : quota enabled and large files being deleted
2941 Description: releasing more than 4GB of quota at once hangs OST
2942 Details : If a user deletes more than 4GB of files on a single OST it
2943 will cause the OST to spin in an infinite loop. Release
2944 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2949 Description: statfs data retrieved from /proc may be stale or zero
2950 Details : When reading per-device statfs data from /proc, in the
2951 {kbytes,files}_{total,free,avail} files, it may appear
2952 as zero or be out of date.
2955 Frequency : systems with MD RAID1 external journal devices
2957 Description: lconf's call to blkid is confused by RAID1 journal devices
2958 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2959 of returning all block devices that match the journal UUID.
2962 Frequency : always, for aggregate stripe size over 4GB
2964 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2965 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2966 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2969 Frequency : always if "lfs find" used on a local file/directory
2971 Description: "lfs find" segfaults if used on a local file/directory
2972 Details : The case where a directory component was not specified wasn't
2973 handled correctly. Handle this properly.
2976 Frequency : always on ppc64
2978 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2979 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2980 perform endianness conversion.
2983 Frequency : rarely (truncate to non-zero file size after write under load)
2984 Bugzilla : 10730, 10687
2985 Description: Files padded with zeros to next 4K multiple
2986 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2987 it is possible that files that are truncated to a non-zero size
2988 immediately after being written are filled with zero bytes beyond
2989 the truncated size. No file data is lost.
2991 Severity : enhancement
2993 Description: Allow recovery/failover for liblustre clients.
2994 Details : liblustre clients were unaware of failover configurations until
2997 Severity : enhancement
2999 Description: user file locks should fail when not mounting with flock option
3000 Details : Set up an error-returning stub in ll_file_operations.lock field
3001 to prevent incorrect behaviour when client is mounted without
3002 flock option. Also, set up properly f_op->flock field for
3006 Frequency : always on ia64
3008 Description: "lfs df" loops on printing out MDS statfs information
3009 Details : The obd_ioctl_data was not initialized and in some systems
3010 this caused a failure during the ioctl that did not return
3011 an error. Initialize the struct and return an error on failure.
3014 Frequency : SLES 9 only
3016 Description: Error of copying files with lustre special EAs as root
3017 Details : Client side always return success for setxattr call for lustre
3018 special xattr (currently only "trusted.lov").
3021 Frequency : rarely on clusters with both ia64+i386 clients
3023 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3024 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3025 clients are doing concurrent writes to the same file it is
3026 possible that the ia64 clients may LASSERT because the OST
3027 extent locks are not PAGE_SIZE aligned. Ensure that grown
3028 locks are always aligned on the request boundary.
3031 Frequency : specific use, occasional
3033 Description: Overwriting in use executable truncates on-disk binary image
3034 Details : If one node attempts to overwrite an executable in use by
3035 another node, we now correctly return ETXTBSY instead of
3036 truncating the file.
3038 Severity : enhancement
3040 Description: Async OSC create to avoid the blocking unnecessarily.
3041 Details : If a OST has no remain object, system will block on the creating
3042 when need to create a new object on this OST. Now, ways use
3043 pre-created objects when available, instead of blocking on an
3044 empty osc while others are not empty. If we must block, we block
3045 for the shortest possible period of time.
3050 Description: chmod on Lustre root is propagated to other clients
3051 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3052 invalid by the follow_mount time.
3055 Frequency : liblustre clients only
3057 Description: Race in 'instant cancel' lock handling could lead to such locks
3058 never to be granted in case of SMP MDS
3059 Details : Do not destroy not yet granted but cbpending locks in
3063 Frequency : replay/resend of open
3065 Description: non null lock assetion failure in mds_intent_policy
3066 Details : Trying to replay/resend lockless open requests resulted in
3067 mds_open() returning 0 with no lock. Now it sets a flag if
3068 a lock is going to be returned.
3070 Severity : enhancement
3072 Description: Checksum enhancements
3073 Details : New checksum enhancements allow for resending RPCs that failed
3076 Severity : enhancement
3078 Description: Tunables on number of dirty pages in cacche
3079 Details : Allow to set limit on number of dirty pages cached.
3084 Description: client crash on unmount - lock still has references
3085 Details : In some error handling cases it was possible to leak a lock
3086 reference on a client while accessing a file. This was not
3087 harmful to the client during operation, but would cause the
3088 client to crash when the filesystem is unmounted.
3091 Frequency : specific case, rare
3093 Description: ETXTBSY on mds though file not in use by client
3094 Details : ETXTBSY is no longer incorrectly returned when attempting to
3095 chmod or chown a directory that the user previously tried to
3096 execute or a currently-executing binary.
3099 Frequency : extremely rare except on liblustre-based clients
3101 Description: Lustre space not freed when files are deleted
3102 Details : Clean up open-unlinked files after client eviction. Previously
3103 the unlink was skipped and the files remained as orphans.
3108 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3109 Details : In some cases it was possible to send a blocking callback to
3110 a client doing a glimpse, even though that client didn't get
3111 a lock granted. When the glimpse lock is cancelled on the OST
3112 the freed lock is left on the waiting list and corrupted the list.
3115 Frequency : all core dumps
3117 Description: Broke core dumps to lustre
3118 Details : Negative dentry may be unhashed if parent does not have UPDATE
3119 lock, but some callers, e.g. do_coredump, expect dentry to be
3120 hashed after successful create, hash it in ll_create_it.
3122 ------------------------------------------------------------------------------
3124 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3126 * Support for kernels:
3127 2.6.9-42.0.2.EL (RHEL 4)
3128 2.6.5-7.276 (SLES 9)
3129 2.4.21-40.EL (RHEL 3)
3130 2.6.12.6 vanilla (kernel.org)
3134 Frequency : always on RHEL 3
3136 Description: Number of open files grows over time
3137 Details : The number of open files grows over time, whether or not
3138 Lustre is started. This was due to a filp leak introduced
3139 by one of our kernel patches.
3141 ------------------------------------------------------------------------------
3143 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3145 * Support for kernels:
3146 2.6.9-42.EL (RHEL 4)
3147 2.6.5-7.276 (SLES 9)
3148 2.4.21-40.EL (RHEL 3)
3149 2.6.12.6 vanilla (kernel.org)
3154 Bugzilla : 5719, 9635, 9792, 9684
3155 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3156 Details : If a server is under heavy load and cannot reply to new
3157 connection requests before the client resends the (re)connect,
3158 the connection handling code can behave badly if two service
3159 threads are concurrently handing separate (re)connections from
3160 the same client. Add better locking to the connection handling
3161 code, and ensure that only a single connection will be processed
3162 for a given client UUID, even if the lock is dropped.
3164 Severity : enhancement
3166 Description: add TCP zero-copy support to kernel
3167 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3168 sends if the hardware supports scatter-gather and checksumming.
3169 This allows socklnd to do client-write and server-read more
3170 efficiently and reduce CPU utilization from skbuf copying.
3173 Frequency : only if NFS exporting from client
3175 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3176 Details : If setting attributes on a file created under NFS that had
3177 never been opened it would be possible to oops the client
3178 if the file had no objects.
3181 Frequency : always for liblustre
3183 Description: liblustre client does MDS+OSTs setattr RPC for each write
3184 Details : When doing a write from a liblustre client, the client
3185 incorrectly issued an RPC to the MDS and each OST the file was
3186 striped over in order to update the timestamps. When writing
3187 with small chunks and many clients this could overwhelm the MDS
3188 with RPCs. In all cases it would slow down the write because
3189 these RPCs are unnecessary.
3191 Severity : enhancement
3193 Description: allow number of MDS service threads to be changed at module load
3194 Details : It is now possible to change the number of MDS service threads
3195 running. Adding "options mds mds_num_threads={N}" to the MDS's
3196 /etc/modprobe.conf will set the number of threads for the next
3197 time Lustre is restarted (assuming the "mds" module is also
3198 reloaded at that time). The default number of threads will
3199 stay the same, 32 for most systems.
3204 Description: OST crash if filesystem is unformatted or corrupt
3205 Details : If an OST is started on a device that has never been formatted
3206 or if the filesystem is corrupt and cannot even mount then the
3207 error handling cleanup routines would dereference a NULL pointer.
3212 Description: NULL pointer deref in llap_from_page.
3213 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3214 incorrect) mapping if the page was truncated/reclaimed while it was
3215 searched for. Check for this condition and skip such pages when
3216 doing readahead. Introduce extra check to llap_from_page() to
3217 verify page->mapping->host is non-NULL (so page is not anonymous).
3220 Frequency : Sometimes when using sys_sendfile
3222 Description: "page not covered by a lock" warnings from ll_readpage
3223 Details : sendfile called ll_readpage without right page locks present.
3224 Now we introduced ll_file_sendfile that does necessary locking
3225 around call to generic_file_sendfile() much like we do in
3229 Frequency : with certain MDS communication failures at client mount time
3231 Description: NULL pointer deref after failed client mount
3232 Details : a client connection request may delayed by the network layer
3233 and not be sent until after the PTLRPC layer has timed out the
3234 request. If the client fails the mount immediately it will try
3235 to clean up before the network times out the request. Add a
3236 reference from the request import to the obd device and delay
3237 the cleanup until the network drops the request.
3240 Frequency : occasionally during client (re)connect
3242 Description: assertion failure during client (re)connect
3243 Details : processing a client connection request may be delayed by the
3244 client or server longer than the client connect timeout. This
3245 causes the client to resend the connection request. If the
3246 original connection request is replied in this interval, the
3247 client may trip an assertion failure in ptlrpc_connect_interpret()
3248 which thought it would be the only running connect process.
3251 Frequency : only with obd_echo servers and clients that are rebooted
3253 Description: kernel BUG accessing uninitialized data structure
3254 Details : When running an obd_echo server it did not start the ping_evictor
3255 thread, and when a client was evicted an uninitialized data
3256 structure was accessed. Start the ping_evictor in the RPC
3257 service startup instead of the OBD startup.
3259 Severity : enhancement
3260 Bugzilla : 10193 (patchless)
3261 Description: Remove dependency on various unexported kernel interfaces.
3262 Details : No longer need reparent_to_init, exit_mm, exit_files,
3263 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3266 Frequency : rare (only users of deprecated and unsupported LDAP config)
3268 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3269 Details : LDAP apparently contains 'lustreName' attributes instead of
3270 'name'. A simple remapping of the name is sufficient.
3273 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3275 Description: waiting_locks_callback trips kernel BUG if client is evicted
3276 Details : Running with the dump_on_timeout debug flag turned on makes
3277 it possible that the waiting_locks_callback() can try to dump
3278 the Lustre kernel debug logs from an interrupt handler. Defer
3279 this log dumping to the expired_lock_main() thread.
3281 Severity : enhancement
3283 Description: Support NFS exporting on 2.6 kernels.
3284 Details : Implement non-rawops metadata methods for NFS server to use without
3285 changing NFS server code.
3288 Frequency : very rare (synthetic metadata workload only)
3290 Description: two racing renames might cause an MDS thread to deadlock
3291 Details : Running the "racer" program may cause one MDS thread to rename
3292 a file from being the source of a rename to being the target of
3293 a rename at exactly the same time that another thread is doing
3294 so, and the second thread has already enqueued these locks after
3295 doing a lookup of the target and is trying to relock them in
3296 order. Ensure that we don't try to re-lock the same resource.
3299 Frequency : only very large systems with liblustre clients
3301 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3302 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3303 all OSTs in parallel. This allows the eviction of stale liblustre
3304 clients to proceed much faster than if they were done in series,
3305 and also offers similar improvements for other set_info RPCs.
3310 Description: excessive CPU usage during initial read phase on client
3311 Details : During the initial read phase on a client, it would agressively
3312 retry readahead on the file, consuming too much CPU and impacting
3313 performance (since 1.4.5.8). Improve the readahead algorithm
3314 to avoid this, and also improve some other common cases (read
3315 of small files in particular, where "small" is files smaller than
3316 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3321 Description: MDS crash when receiving packet with unknown intent.
3322 Details : Do not LBUG in unknown intent case, just return -EFAULT
3324 Severity : enhancement
3325 Bugzilla : 9293, 9385
3326 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3327 Details : Do not serialize getattr (non-intent version) and statfs.
3330 Frequency : occasional, when OST network is overloaded/intermittent
3332 Description: client evicted by OST after bulk IO timeout
3333 Details : If a client sends a bulk IO request (read or write) the OST
3334 may evict the client if it is unresposive to its data GET/PUT
3335 request. This is incorrect if the network is overloaded (takes
3336 too long to transfer the RPC data) or dropped the OST GET/PUT
3337 request. There is no need to evict the client at all, since
3338 the pinger and/or lock callbacks will handle this, and the
3339 client can restart the bulk request.
3342 Frequency : Always when mmapping file with no objects
3344 Description: client crashes when mmapping file with no objects
3345 Details : Check that we actually have objects in a file before doing any
3346 operations on objects in ll_vm_open, ll_vm_close and
3352 Description: Request leak when working with deleted CWD
3353 Details : Introduce advanced request refcount tracking for requests
3354 referenced from lustre intent.
3356 Severity : Enhancement
3358 Description: Cache open file handles on client.
3359 Details : MDS now will return special lock along with openhandle, if
3360 requested and client is allowed to hold openhandle, even if unused,
3361 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3362 closing files for every read/write openration.
3364 Severity : Enhancement
3366 Description: Cache open negative dentries on client when possible.
3367 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3368 negative dentries on lock revocation.
3373 Description: Remounting a client read-only wasn't possible with a zconf mount
3374 Details : It wasn't possible to remount a client read-only with llmount.
3376 Severity : enhancement
3377 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3378 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3379 MPICH is included to add Lustre-specific ADIO interfaces.
3380 This is based closely on the UFS ADIO layer and only differs
3381 in file creation, in order to allow the OST striping to be set.
3382 This is user-contributed code and not supported by CFS.
3387 Description: extended inode attributes (immutable, append-only) work improperly
3388 when 2.4 and 2.6 kernels are used on client/server or vice versa
3389 Details : Introduce kernel-independent values for these flags.
3391 Severity : enhancement
3394 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3395 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3396 granularity. Now, Fractional MB granularity is be supported,
3397 this is very useful for low memory system.
3399 Severity : enhancement
3401 Description: Getattr by fid
3402 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3403 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3404 have locks and avoids getattr rpc requests for it on every stat.
3407 Frequency : Always, for filesystems larger than 2TB
3409 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3410 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3411 filesystems larger than 2TB to be mounted with mballoc enabled.
3414 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3416 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3417 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3418 filesystems larger than 2TB is mounted then the kernel will
3419 silently corrupt the start of the filesystem. CONFIG_LBD is
3420 enabled for all CFS-supported kernels, but the possibility of
3421 this happening with a modified kernel config exists.
3423 Severity : enhancement
3425 Description: add client O_DIRECT support for 2.6 kernels
3426 Details : It is now possible to do O_DIRECT reads and writes to files
3427 in the Lustre client mountpoint on 2.6 kernel clients.
3429 Severity : enhancement
3431 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3432 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3433 parallel, not waiting for response from every OST before sending
3434 a rpc to the next OST.
3439 Description: setattr vs write race when updating file timestamps
3440 Details : Client processes that update a file timestamp into the past
3441 right after writing to the file (e.g. tar) it is possible that
3442 the updated file modification time can be reset to the current
3443 time due to a race between processing the setattr and write RPC.
3445 Severity : enhancement
3447 Description: Bring 'lfs find' closer in line with regular Linux find.
3448 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3449 -print0 options and obtains all the needed info through the lustre
3452 Severity : enhancement
3454 Description: support up to 1024 configured devices on one node
3455 Details : change obd_dev array from statically allocated to dynamically
3456 allocated structs as they are first used to reduce memory usage
3461 Description: Flush dirty partially truncated pages during truncate
3462 Details : Immediatelly flush partially truncated pages in filter_setattr,
3463 this way we completely avoid having any pages in page cache on OST
3464 and can retire ugly workarounds during writes to flush such pages.
3469 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3470 Details : protect inode from truncation within vfs_unlink() context
3471 just take a reference before calling vfs_unlink() and release it
3472 when parent's i_sem is free.
3477 Description: last_id value checked outside lock on OST caused LASSERT failure
3478 Details : If there were multiple MDS->OST object precreate requests in
3479 flight, it was possible that the OST's last object id was checked
3480 outside a lock and incorrectly tripped an assertion. Move checks
3481 inside locks, and discard old precreate requests.
3484 Frequency : always, if extents are used on OSTs
3486 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3487 Details : OSTs using the extents format would not zero the high 16 bits of
3488 the index physical block number. This is not a problem for any
3489 OST filesystems smaller than 16TB, and no kernels support ext3
3490 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3491 new/modified files) and can be fixed for existing filesystems
3492 with e2fsprogs-1.39-cfs1.
3497 Description: import connection selection may be incorrect if timer wraps
3498 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3499 import connections to be ignored if the 32-bit jiffies counter
3500 wraps. Use a 64-bit jiffies counter.
3503 Frequency : very large clusters immediately after boot
3505 Description: LNET request buffers exhausted under heavy short-term load
3506 Details : If a large number of client requests are generated on a service
3507 that has previously never seen so many requests it is possible
3508 that the request buffer growth cannot keep up with the spike in
3509 demand. Instead of dropping incoming requests, they are held in
3510 the LND until the RPC service can accept more requests.
3513 Frequency : Sometimes during replay
3515 Description: Assertion failure in ll_local_open after replay.
3516 Details : If replay happened on an open request reply before we were able
3517 to set replay handler, reply will become not swabbed tripping the
3518 assertion in ll_local_open. Now we set the handler right after
3519 recognising of open request
3522 Frequency : very rare
3524 Description: kernel reports "badness in vsnprintf"
3525 Details : Reading from the "recovery_status" /proc file in small chunks
3526 may cause a negative length in lprocfs_obd_rd_recovery_status()
3527 call to vsnprintf() (which is otherwise harmless). Exit early
3528 if there is no more space in the output buffer.
3530 Severity : enhancement
3532 Description: clear OBD RPC statistics by writing to them
3533 Details : It is now possible to clear the OBD RPC statistics by writing
3534 to the "stats" file.
3539 Description: Client mtime is not the same on different clients after utimes
3540 Details : In some cases, the client was using the utimes() syscall on
3541 a file cached on another node. The clients now validate the
3542 ctime from the MDS + OSTs to determine which one is right.
3547 Description: Inability to activate failout mode
3548 Details : lconf script incorrectly assumed that in pythong string's numeric
3549 value is used in comparisons.
3552 Frequency : always with multiple stripes per file
3554 Description: Inefficient object allocation for mutli-stripe files
3555 Details : When selecting which OSTs to stripe files over, for files with
3556 a stripe count that divides evenly into the number of OSTs,
3557 the MDS is always picking the same starting OST for each file.
3558 Return the OST selection heuristic to the original design.
3563 Description: mount failures may take full timeout to return an error
3564 Details : Under some heavy load conditions it is possible that a
3565 failed mount can wait for the full obd_timeout interval,
3566 possibly several minutes, before reporting an error.
3567 Instead return an error as soon as the status is known.
3569 Frequency : quota enabled and large files being deleted
3571 Description: releasing more than 4GB of quota at once hangs OST
3572 Details : If a user deletes more than 4GB of files on a single OST it
3573 will cause the OST to spin in an infinite loop. Release
3574 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3579 Description: statfs data retrieved from /proc may be stale or zero
3580 Details : When reading per-device statfs data from /proc, in the
3581 {kbytes,files}_{total,free,avail} files, it may appear
3582 as zero or be out of date.
3585 Frequency : systems with MD RAID1 external journal devices
3587 Description: lconf's call to blkid is confused by RAID1 journal devices
3588 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3589 of returning all block devices that match the journal UUID.
3592 Frequency : always, for aggregate stripe size over 4GB
3594 Description: assertion fails when trying to use 4GB stripe size
3595 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3596 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3599 Frequency : always on ppc64
3601 Description: the first write on an ext3 filesystem with mballoc got stuck
3602 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3603 perform endianness conversion.
3605 ------------------------------------------------------------------------------
3607 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3609 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3610 INTEROPERATE with older versions automatically. Please read the
3611 user documentation before upgrading any part of a live system.
3612 * WARNING: Lustre networking configuration changes are required with
3613 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3616 * Support for kernels:
3617 2.6.9-22.0.2.EL (RHEL 4)
3618 2.6.5-7.244 (SLES 9)
3619 2.6.12.6 vanilla (kernel.org)
3622 Severity : enhancement
3623 Bugzilla : 7981/8208
3624 Description: Introduced Lustre Networking (LNET)
3625 Details : LNET is new networking infrastructure for Lustre, it includes
3626 a reorganized network configuration mode (see the user
3627 documentation for full details) as well as support for routing
3628 between different network fabrics. Lustre Networking Devices
3629 (LNDs) for the supported network fabrics have also been
3630 created for this new infrastructure.
3632 Severity : enhancement
3633 Description: Introduced Access control lists
3634 Details : clients can set ACLs on files and directories in order to have
3635 more fine-grained permissions than the standard Unix UGO+RWX.
3636 The MDS must be started with the "-o acl" mount option.
3638 Severity : enhancement
3639 Description: Introduced filesystem quotas
3640 Details : Administrators may now establish per-user quotas on the
3643 Severity : enhancement
3645 Description: Configuration change for the XT3
3646 The PTLLND is now used to run Lustre over Portals on the XT3
3647 The configure option(s) --with-cray-portals are no longer used.
3648 Rather --with-portals=<path-to-portals-includes> is used to
3649 enable building on the XT3. In addition to enable XT3 specific
3650 features the option --enable-cray-xt3 must be used.
3655 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3656 Details : A race between allocating a new llcd and re-getting the llcd_lock
3657 allowed another thread to grab newly-allocated llcd.
3659 Severity : enhancement
3661 Description: 2.6 OST async journal commit and locking fix to improve performance
3662 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3663 2.6 kernels are now async as they already were in 2.4 kernels so
3664 that they can commit concurrently with the network bulk transfer.
3665 For block-allocated files the filter allocation semaphore is held
3666 to avoid filesystem fragmentation during allocation. BKL lock
3667 removed for 2.6 xattr operations where it is no longer needed.
3672 Description: lconf incorrectly determined whether two IP networks could talk
3673 Details : In some more complicated routing and multiple-network
3674 configurations, lconf will avoid trying to make a network
3675 connection to a disjoint part of the IP space. It was doing the
3676 math incorrectly for one set of cases.
3681 Description: Fix for potential infinite loop processing records in an llog.
3682 Details : If an llog record is corrupted/zeroed, it is possible to loop
3683 forever in llog_process(). Validate the llog record length
3684 and skip the remainder of the block on error.
3687 Frequency : occasional (liblustre only)
3689 Description: liblustre could not open files whose last component is a symlink
3690 Details : sysio_path_walk() would incorrectly pass the open intent to
3691 intermediate path components.
3694 Frequency : rare (liblustre only with non-standard tuning)
3695 Bugzilla : 7201 (7350)
3696 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3697 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3698 after it has been released, only do this for referenced locks
3700 Severity : enhancement
3702 Description: specifying an (invalid) directory default stripe_size of -1
3703 would reset the directory default striping
3704 Details : stripe_size -1 was used internally to signal directory stripe
3705 removal, now use "all default" to signal dir stripe removal
3706 as a directory striping of "all default" is not useful
3709 Frequency : common for large clusters running liblustre clients
3711 Description: doing an ls when liblustre clients are running is slow
3712 Details : sending a glimpse AST to a liblustre client waits for every AST
3713 to time out, as liblustre clients will not respond. Since they
3714 cannot cache data we refresh the OST lock LVB from disk instead.
3716 Severity : enhancement
3718 Description: doing an ls at the same time as file IO can be slow
3719 Details : enqueue and other "small" requests can be blocked behind many
3720 large IO requests. Create a new OST IO portal for non-IO
3721 requests so they can be processed faster.
3724 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3726 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3727 created file with mode 000
3728 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3729 allow an ftruncate() as a non-root user to a file with mode 000.
3730 Lustre now respects this flag to disable mode checks when
3731 truncating a file owned by the user
3734 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3737 Description: Revoking locks from clients that went dead or catatonic might take
3739 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3740 cancellation of such locks instant on servers without waiting for
3741 any reply from clients. Clients drops these locks when cancel
3742 notification from server is received without replying.
3745 Frequency : liblustre-only, when liblustre client dies or becomes busy
3747 Description: Doing ls on Linux clients can take a long time with active
3749 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3750 granting such locks to it in the first place if possible. Locks
3751 are taken by proxy on the OST during the read or write and
3752 dropped immediately afterward. Add connect flags handling, do
3753 not grant locks to liblustre clients for glimpse ASTs.
3755 Severity : enhancement
3757 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3758 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3759 sequential reads and behaves badly with applications doing random
3760 reads. Improve it by reading ahead at least read region, and
3761 avoiding excessive large RPC for small reads.
3763 Severity : enhancement
3765 Description: Creating more than 1000 files for a single job may cause a load
3766 imbalance on the OSTs if there are also a large number of OSTs.
3767 Details : qos_prep_create() uses an OST index reseed value that is an
3768 even multiple of the number of available OSTs so that if the
3769 reseed happens in the middle of the object allocation it will
3770 still utilize the OSTs as uniformly as possible.
3775 Description: OST or MDS may oops in ping_evictor_main()
3776 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3777 but doesn't restart at beginning of obd_exports_timed list
3780 Severity : enhancement
3782 Description: improve by-nid export eviction on the MDS and OST
3783 Details : allow multiple exports with the same NID to be evicted at one
3784 time without re-searching the exports list.
3787 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3789 Description: MDS may oops in groups_free()
3790 Details : in rare race conditions a newly allocated group_info struct is
3791 freed again, and this can be NULL. The 2.4 compatibility code
3792 for groups_free() checked for a NULL pointer, but 2.6 did not.
3795 Frequency : common for liblustre clients doing little filesystem IO
3796 Bugzilla : 9352, 7313
3797 Description: server may evict liblustre clients accessing contended locks
3798 Details : if a client is granted a lock or receives a completion AST
3799 with a blocking AST already set it would not reply to the AST
3800 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3803 Frequency : lfs setstripe, only systems with more than 160 OSTs
3805 Description: unable to set striping with a starting offset beyond OST 160
3806 Details : llapi_create_file() incorrectly limited the starting stripe
3807 index to the maximum single-file stripe count.
3810 Frequency : LDAP users only
3812 Description: lconf did not handle in-kernel recovery with LDAP properly
3813 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3815 Severity : enhancement
3817 Description: bind OST threads to NUMA nodes to improve performance
3818 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3819 node and do their allocations there to localize memory access
3821 Severity : enhancement
3823 Description: llmount can determine client NID directly from Myrinet (GM)
3824 Details : the client NID code from gmnalnid was moved directly into
3825 llmount, removing the need to use this or specifying the
3826 client NID explicitly when mounting GM clients with zeroconf
3829 Frequency : if client is started with down MDS
3831 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3832 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3833 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3834 ptlrpc_import_delayed_req() because we didn't check if the
3835 request was interrupted, and we also didn't break out of the
3836 event loop if there was no timeout
3841 Description: data loss during non-page-aligned writes to a single file from
3842 both multiple nodes and multiple threads on one node at same time
3843 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3844 inconsistency led to false short-reads, that were cached and later
3845 used by ->prepare_write() to fill in partially written page,
3846 leading to data loss.
3849 Frequency : always, if lconf --abort_recovery used
3851 Description: lconf --abort_recovery fails with 'Operation not supported'
3852 Details : lconf was attempting to abort recovery on the MDT device and not
3855 Severity : enhancement
3857 Description: remove cleanup logs
3858 Details : replace lconf-generated cleanup logs with lustre internal
3859 cleanup routines. Eliminates the need for client-cleanup and
3862 Severity : enhancement
3864 Description: add support for EAs (user and system) on lustre filesystems
3865 Details : it is now possible to store extended attributes in the Lustre
3866 client filesystem, and with the user_xattr mount option it
3867 is possible to allow users to store EAs on their files also
3869 Severity : enhancement
3871 Description: Add possibility (config option) to show minimal available OST free
3873 Details : When compiled with --enable-mindf configure option, statfs(2)
3874 (and so, df) will return least minimal free space available from
3875 all OSTs as amount of free space on FS, instead of summary of
3876 free spaces of all OSTs.
3878 Severity : enhancement
3880 Description: do not expand extent locks acquired on OST-side
3881 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3882 by server: they are not cached anyway.
3885 Frequency : when mmap is used/binaries executed from Lustre
3887 Description: Unmmap pages before throwing them away from read cache.
3888 Details : llap_shrink cache now attempts to unmap pages before discarding
3889 them (if unmapping failed - do not discard). SLES9 kernel has
3890 extra checks that trigger if this unmapping is not done first.
3895 Description: lconf didn't resolve symlinks before checking to see whether a
3896 given mountpoint was already in use
3899 Frequency : when migrating failover services
3900 Bugzilla : 6395, 9514
3901 Description: When migrating a subset of services from a node (e.g. failback
3902 from a failover service node) the remaining services would
3903 time out and evict clients.
3904 Details : lconf --force (implied by --failover) sets the global obd_timeout
3905 to 5 seconds in order to quickly disconnect, but this caused
3906 other RPCs to time out too quickly. Do not change the global
3907 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3909 Severity : enhancement
3910 Frequency : if MDS is started with down OST
3911 Bugzilla : 9439,5706
3912 Description: Allow startup/shutdown of an MDS without depending on the
3913 availability of the OSTs.
3914 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3915 Add appropriate locking and lov-osc refcounts for safe
3916 cleaning. Add osc abort_inflight calls in case the
3917 synchronize never started.
3920 Frequency : occasional (Cray XT3 only)
3922 Description: root not authorized to access files in CRAY_PORTALS environment
3923 Details : The client process capabilities were not honoured on the MDS in
3924 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3925 been accessed by an authorized user then root was able to access
3926 the file on the local client also. The root user capabilities
3927 are now allowed on the MDS, as this environment has secure UID.
3930 Frequency : occasional
3932 Description: ldiskfs "too long searching" message happens too often
3933 Details : A debugging message (otherwise harmless) prints too often on
3934 the OST console. This has been reduced to only happen when
3935 there are fragmentation problems on the filesystem.
3940 Description: Division by zero in statfs when all OSCs are inactive
3941 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3942 lov_statfs divides by value returned by lov_get_stripecnt().
3946 Bugzilla : 9489, 3273
3947 Description: First write from each client to each OST was only 4kB in size,
3948 to initialize client writeback cache, which caused sub-optimal
3949 RPCs and poor layout on disk for the first writen file.
3950 Details : Clients now request an initial cache grant at (re)connect time
3951 and so that they can start streaming writes to the cache right
3952 away and always do full-sized RPCs if there is enough data.
3953 If the OST is rebooted the client also re-establishes its grant
3954 so that client cached writes will be honoured under the grant.
3959 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3960 Details : Now I/O RPCs go to different portal number and (presumably) fast
3961 lock requests (and glimses) and other RPCs get their own service
3962 threads pool that should be able to service those RPCs
3965 Severity : enhancement
3967 Description: Ability to exchange lustre version between client and servers and
3968 issue warnings at client side if client is too old. Also for
3969 liblustre clients there is ability to refuse connection of too old
3971 Details : New 'version' field is added to connect data structure that is
3972 filled with version info. That info is later checked by server and
3976 Frequency : rare, liblustre only.
3977 Bugzilla : 9296, 9581
3978 Description: Two simultaneous writes from liblustre at offset within same page
3979 might proceed at the same time overwriting eachother with stale
3981 Details : I/O lock withing llu_file_prwv was released too early, before data
3982 actually was hitting the wire. Extended lock-holding time until
3983 server acknowledges receiving data.
3986 Frequency : extremely rare. Never observed in practice.
3988 Description: avoid generating lustre_handle cookie of 0.
3989 Details : class_handle_hash() generates handle cookies by incrementing
3990 global counter, and can hit 0 occasionaly (this is unlikely, but
3991 not impossible, because initial value of cookie counter is
3992 selected randonly). Value of 0 is used as a sentinel meaning
3993 "unassigned handle" --- avoid it. Also coalesce two critical
3994 sections in this function into one.
3996 Severity : enhancement
3998 Description: allow liblustre clients to delegate truncate locking to OST
3999 Details : To avoid overhead of locking, liblustre client instructs OST to
4000 take extent lock in ost_punch() on client's behalf. New connection
4001 flag is added to handle backward compatibility.
4003 Severity : enhancement
4004 Bugzilla : 4928, 7341, 9758
4005 Description: allow number of OST service threads to be specified
4006 Details : a module parameter allows the number of OST service threads
4007 to be specified via "options ost ost_num_threads={N}" in the
4008 OSS's /etc/modules.conf or /etc/modprobe.conf.
4012 Bugzilla : 6146, 9635, 9895
4013 Description: servers crash with bad pointer in target_handle_connect()
4014 Details : In rare cases when a client is reconnecting it was possible that
4015 the connection request was the last reference for that export.
4016 We would temporarily drop the export reference and get a new
4017 one, but this may have been the last reference and the export
4018 was just destroyed. Get new reference before dropping old one.
4020 Severity : enhancement
4021 Frequency : if client is started with failover MDS
4023 Description: Allow multiple MDS hostnames in the mount command
4024 Details : Try to read the configuration from all specified MDS
4025 hostnames during a client mount in case the "primary"
4028 Severity : enhancement
4030 Description: Stop sending data to evicted clients as soon as possible.
4031 Details : Check if the client we are about to send or are sending data to
4032 was evicted already. (Check is done every second of waiting,
4033 for which l_wait_event interface was extended to allow checking
4034 of exit condition at specified intervals).
4037 Frequency : rare, normally only when NFS exporting is done from client
4039 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4041 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4044 Frequency : always, if srand() is called before liblustre initialization
4046 Description: Liblustre uses system PRNG disturbing its usage by user application
4047 Details : Introduce internal to lustre fast and high-quality PRNG for
4048 lustre usage and make liblustre and some other places in generic
4049 lustre code to use it.
4051 Severity : enhancement
4052 Bugzilla : 9477, 9557, 9870
4053 Description: Verify that the MDS configuration logs are updated when xml is
4054 Details : Check if the .xml configuration logs are newer than the config
4055 logs stored on the MDS and report an error if this is the case.
4056 Request --write-conf, or allow starting with --old_conf.
4058 Severity : enhancement
4060 Description: Handle symlinks in the path when checking if Lustre is mounted.
4061 Details : Resolve intermediate symlinks when checking if a client has
4062 mounted a filesystem to avoid duplicate client mounts.
4067 Description: lconf can hit an error exception but still return success.
4068 Details : The lconf command catches the Command error exception at the top
4069 level script context and will exit with the associated exit
4070 status, but doesn't ensure that this exit status is non-zero.
4075 Description: failure of ptlrpc thread startup can cause oops
4076 Details : Starting a ptlrpc service thread can fail if there are a large
4077 number of threads or the server memory is very fragmented.
4078 Handle this without oopsing.
4081 Frequency : always, only if liblustre and non-default acceptor port was used
4083 Description: liblustre cannot connect to servers with non-default acceptor port
4084 Details : tcpnal_set_default_params() was not called and was therefore
4085 ignoring the environment varaible TCPNAL_PORT, as well as other
4086 TCPNAL_ environment variables
4091 Description: two objects could be created on the same OST for a single file
4092 Details : If an OST is down, in some cases it was possible to create two
4093 objects on a single OST for a single file. No problems other
4094 than potential performance impact and spurious error messages.
4098 Bugzilla : 5681, 9562
4099 Description: Client may oops in ll_unhash_aliases
4100 Details : Client dcache may become inconsistent in race condition.
4101 In some cases "getcwd" can fail if the current directory is
4107 Description: Inode refcounting problems in NFS export code
4108 Details : link_raw functions used to call d_instantiate without obtaining
4109 extra inode reference first.
4113 Bugzilla : 9942, 9903
4114 Description: Referencing freed requests leading to crash, memleaks with NFS.
4115 Details : We used to require that call to ll_revalidate_it was always
4116 followed by ll_lookup_it. Also with revalidate_special() it is
4117 possible to call ll_revalidate_it() twice for the same dentry
4118 even if first occurence returned success. This fix changes semantic
4119 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4120 reference on a request referred from the intent.
4121 ll_intent_release() then releases such a request.
4124 Frequency : rare, normally benchmark loads only
4126 Description: unlinked inodes were kept in memory on the client
4127 Details : If a client is repeatedly creating and unlinking files it
4128 can accumulate a lot of stale inodes in the inode slab cache.
4129 If there is no other client load running this can cause the
4130 client node to run out of memory. Instead flush old inodes
4131 from client cache that have the same inode number as a new inode.
4134 Frequency : SLES9 2.6.5 kernel and long filenames only
4135 Bugzilla : 9969, 10379
4136 Description: utime reports stale NFS file handle
4137 Details : SLES9 uses out-of-dentry names in some cases, which confused
4138 the lustre dentry revalidation. Change it to always use the
4142 Frequency : rare, unless heavy write-truncate concurrency is continuous
4143 Bugzilla : 4180, 6984, 7171, 9963, 9331
4144 Description: OST becomes very slow and/or deadlocked during object unlink
4145 Details : filter_destroy() was holding onto the parent directory lock
4146 while truncating+unlinking objects. For very large objects this
4147 may block other threads for a long time and slow overall OST
4148 responsiveness. It may also be possible to get a lock ordering
4149 deadlock in this case, or run out of journal credits because of
4150 the combined truncate+unlink. Solution is to do object truncate
4151 first in one transaction without parent lock, and then do the
4152 final unlink in a new transaction with the parent lock. This
4153 reduces the lock hold time dramatically.
4156 Frequency : rare, 2.4 kernels only
4158 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4159 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4160 before unmounting their filesystem in order to avoid deadlock.
4161 The kernel_locked() function in 2.4 kernels only checks whether
4162 the kernel lock is held, not whether it is this process that is
4163 holding it as 2.6 kernels do.
4168 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4169 Details : The client ptlrpc code may be trying to reconnect to a down
4170 server before a previous connection attempt has timed out.
4171 Increase the reconnect interval to be longer than the connection
4172 timeout interval to avoid sending duplicate connections to
4176 Frequency : echo_client brw_test command
4178 Description: fix echo_client to work with OST preallocated code
4179 Details : OST preallocation code (5137) didn't take echo_client IO path
4180 into account: echo_client calls filter methods outside of any
4181 OST thread and, hence, there is no per-thread preallocated
4182 pages and buffers to use. Solution: hijack pga pages for IO. As
4183 a byproduct, this avoids unnecessary data copying.
4187 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4188 Description: Client can oops in mdc_commit_close() after open replay
4189 Details : It was possible for the MDS to return an open request with no
4190 transaction number in mds_finish_transno() if the client was
4191 evicted, but without actually returning an error. Clients
4192 would later try to replay that open and may trip an assertion
4193 Simplify the client close codepath, and always return an error
4194 from the MDS in case the open is not successful.
4197 Frequency : rare, 2.6 OSTs only
4199 Description: OST may deadlock under high load on fragmented files
4200 Details : If there was a heavy load and highly-fragmented OST filesystems
4201 it was possible to have all the OST threads deadlock waiting on
4202 allocation of biovecs, because the biovecs were not released
4203 until the entire RPC IO was completed. Instead, release biovecs
4204 as soon as they are complete to ensure forward IO progress.
4206 Severity : enhancement
4208 Description: Support for specifying external journal device at mount
4209 Details : If an OST or MDS device is formatted with an external journal
4210 device, this device major/minor is stored in the ext3 superblock
4211 and may not be valid for failover. Allow detecting and
4212 specifying the external journal at mount time.
4217 Description: Mounting an MDS with pending unlinked files may cause oops
4218 Details : target_finish_recovery() calls mds_postrecov() which returned
4219 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4220 considers this an error and immediately begins cleaning up the
4221 lov, just after starting the mds_lov process
4223 Severity : enhancement
4225 Description: Implement 'lfs df' to report actual free space on per-OST basis
4226 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4227 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4228 usage of objects; '-h' to report in human readable format.
4230 ------------------------------------------------------------------------------
4232 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4239 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4240 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4241 at startup during orphan unlinking. Instead, skip buddy update
4242 before setup, as it will be regenerated after recovery is complete.
4247 Description: If an OST is inactive, its locks might reference stale inodes.
4248 Details : lov_change_cbdata() must iterate over all namespaces, even if
4249 they are inactive to clear inode references from the lock.
4251 Severity : enhancement
4252 Frequency : occasional, if non-standard max_dirty_mb used
4254 Description: Client will block write RPCs if not enough grant
4255 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4256 then the client will block writes while waiting for another RPC
4257 to complete instead of consuming its dirty limit. With change
4258 we get improved performance when max_dirty_mb is small.
4260 Severity : enhancement
4261 Bugzilla : 3389, 6253
4262 Description: Add support for supplementary groups on the MDS.
4263 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4264 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4265 lookups for user supplementary groups into a cache.
4269 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4270 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4271 for liblustre O_EXCL is enforced by MDS.
4275 Description: readdir never returns NULL in liblustre.
4276 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4277 dentry in struct dirent.
4281 Frequency : liblustre only, depends on application IO pattern
4282 Description: liblustre clients evicted if not contacting servers
4283 Details : Don't put liblustre clients into the ping_evictor list, so
4284 they will not be evicted by the pinger ever.
4286 Severity : enhancement
4288 Description: Add ability to evict clients by NID from MDS.
4289 Details : By echoing "nid:$NID" string into
4290 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4291 $NID would be instantly evicted from this MDS and from all active
4292 OSTs connected to it.
4296 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4297 Details : lookup_it_finish() used to query file size from OSTs that was not
4302 Description: service threads change working directory to that of init
4303 Details : Starting lustre service threads may pin the working directory
4304 of the parent thread, making that filesystem busy. Threads
4305 now change to the working directory of init to avoid this.
4309 Frequency : during shutdown only
4310 Description: shutdown with a failed MDS or OST can cause unmount to hang
4311 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4316 Frequency : 2.6 only
4317 Description: chmod/chown may include an extra supplementary group
4318 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4319 supplementary group and if none is specified this is used.
4322 Bugzilla : 5479 (6816)
4324 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4325 Details : If lookup is in progress on a file that is unlinked we might try
4326 to revalidate the inode and fail in revalidate after lookup is
4327 complete and ll_file_open() enqueues the open again but
4328 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4331 Frequency : always, if lconf --abort_recovery used
4333 Description: lconf --abort_recovery fails with 'Operation not supported'
4334 Details : lconf was attempting to abort recovery on the MDT device and not
4337 ------------------------------------------------------------------------------
4339 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4344 Frequency : rare (only unsupported configurations with a node running as an
4346 Bugzilla : 6514, 5137
4347 Description: Mounting a Lustre file system on a node running as an OST could
4349 Details : OSTs now preallocates memory needed to write out data at
4350 startup, instead of when needed, to avoid having to
4351 allocate memory in possibly low memory situations.
4352 Specifically, if the file system is mounted on on OST,
4353 memory pressure could force it to try to write out data,
4354 which it needed to allocate memory to do. Due to the low
4355 memory, it would be unable to do so and the node would
4356 become unresponsive.
4358 Severity : enhancement
4360 Description: Addition of lconf --service command line option
4361 Details : lconf now accepts a '--service <arg>' option, which is
4362 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4364 Severity : enhancement
4366 Description: Failover mode is now the default for OSTs.
4367 Details : By default, OSTs will now run in failover mode. To return to
4368 the old behaviour, add '--failout' to the lmc line for OSTs.
4370 Severity : enhancement
4372 Description: Health checks are now provided for MDS and OSTs
4373 Details : Additional detailed health check information on MSD and OSTs
4374 is now provided through the procfs health_check value.
4377 Frequency : occasional, depends on IO load
4379 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4380 after numerous create/delete cycles
4381 Details : The ext3 inode allocation policy would not allocate new inodes
4382 very well on the OSTs because there are no new directories
4383 being created. Instead we look for groups with free space if
4384 the parent directories are nearly full.
4389 Description: Network or server problems during mount may cause partially
4390 mounted clients instead of returning an error.
4391 Details : The config llog parsing code may overwrite the error return
4392 code during mount error handling, returning success instead
4398 Description: MDS can fail to allocate large reply buffers
4399 Details : After long uptimes the MDS can fail to allocate large reply
4400 buffers (e.g. zconf client mount config records) due to memory
4401 fragmentation or consumption by the buffer cache. Preallocate
4402 some large reply buffers so that these replies can be sent even
4403 under memory pressure.
4407 Frequency : rare (liblustre)
4408 Description: fsx running with liblustre complained that using truncate() to
4409 extend the file doesn't work. This patch corrects that issue.
4410 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4411 ATTR_SIZE and lsm use in llu_setattr_raw.
4415 Frequency : rare, only 2.6 kernels
4416 Description: Unusual file access patterns on the MDS may result in inode
4417 data being lost in very rare circumstances.
4418 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4419 code in the 2.6 kernel caused the inode and/or EA data not to
4420 be read from disk, causing single-file corruption.
4424 Frequency : rare, only 2.6 filesystems using extents
4425 Description: Heavy concurrent write and delete load may cause data corruption.
4426 Details : It was possible under high-load situations to have an extent
4427 metadata block in the block device cache from a just-unlinked
4428 file overwrite a newly-allocated data block. We now unmap any
4429 metadata buffers that alias just-allocated data blocks.
4433 Frequency : filesystems with default stripe_count larger than 77
4434 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4435 Details : lconf specifies an inode size of 4096 bytes when the default
4436 stripe_count is larger than 77. This conflicts with the default
4437 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4438 this case to avoid pinning too much memory for large EAs.
4440 ------------------------------------------------------------------------------
4442 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4447 Frequency : rare (extremely heavy IO load with hundreds of clients)
4449 Description: Client is evicted, gets IO error writing to file
4450 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4451 caused clients to be evicted for AST timeouts. The fixes in
4452 bug 5192 mean we no longer need to have such short AST timeouts
4453 so ldlm_timeout has been increased.
4456 Frequency : occasional during --force or --failover shutdown under load
4457 Bugzilla : 5949, 4834
4458 Description: Server oops/LBUG if stopped with --force or --failover under load
4459 Details : a collection of import/export refcount and cleanup ordering
4460 issues fixed for safer force cleanup
4463 Frequency : only filesystems larger than 120 OSTs
4464 Bugzilla : 5990, 6223
4465 Description: lfs getstripe would oops on a very large filesystem
4466 Details : lov_getconfig used kfree on vmalloc'd memory
4469 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4470 Bugzilla : 6242, 6243
4471 Description: reading from files that had been truncated to a non-zero size
4472 but never opened returned no data
4473 Details : ll_file_read() reads zeros from no-object files to EOF
4478 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4479 Details : There's little harm in aborting MDS/OSS recovery and letting it
4480 try again, so I removed the LASSERT and return an error instead.
4482 Severity : enhancement
4484 Description: New debugging infrastructure for tracking down data corruption
4485 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4486 (b) cover more of the client-side code path, and (c) try to narrow
4487 down where problems occurred
4491 Bugzilla : 3819, 4364, 4397, 6313
4492 Description: Racing close and eviction MDS could cause assertion in mds_close
4493 Details : It was possible to get multiple mfd references during close and
4494 client eviction, leading to one thread referencing a freed mfd.
4496 Severity: : enhancement
4497 Bugzilla : 3262, 6359
4498 Description: Attempts to reconnect to servers are now more aggressive.
4499 Details : This builds on the enhanced upcall-less recovery that was added
4500 in 1.4.2. When trying to reconnect to servers, clients will
4501 now try each server in the failover group every 10 seconds. By
4502 default, clients would previously try one server every 25 seconds.
4507 Description: After recovery, certain operations trigger a failed
4508 assertion on a client.
4509 Details : Failing over an mds, using lconf -d --failover, while a
4510 client was doing a readdir() call would cause the client to
4511 LBUG after recovery completed and the readdir() was resent.
4513 Severity : enhancement
4515 Description: Default groups are now added by lconf
4516 Details : You can now run lconf --group <servicename> without having to
4517 manually add groups with lmc.
4520 Frequency : occasional
4522 Description: Nodes with an elan id of 0 trigger a failed assertion
4525 Frequency : always when accessing e.g. tty/console device nodes
4527 Description: tty and some other devices nodes cannot be used on lustre
4528 Details : file's private_data field is used by device data and lustre
4529 values in there got lost. New field was added to struct file to
4530 store fs-specific private data.
4533 Frequency : when exporting Lustre via NFS
4535 Description: NFSD failed occasionally when looking up a path component
4536 Details : NFSD is looking up ".." which was broken in ext3 directories
4537 that had grown large enough to become hashed.
4540 Frequency : Clusters with multiple interfaces not on the same subnet
4542 Description: Nodes will repeatedly try to reconnect to an interface which it
4543 cannot reach and report an error to the log.
4544 Details : Extra peer list entries will be created by lconf with some peers
4545 unreachable. lconf now validates the peer before adding it.
4548 Frequency : Only if a default stripe is set on the filesystem root.
4550 Description: Setting a default stripe on the filesystem root prevented the
4551 filesystem from being remounted.
4552 Details : The client was sending extra request flags in the root getattr
4553 request and did not allocate a reply buffer for the dir EA.
4556 Frequency : occasional, higher if lots of files are accessed by one client
4557 Bugzilla : 6159, 6097
4558 Description: Client trips assertion regarding lsm mismatch/magic
4559 Details : While revalidating inodes the VFS looks up inodes with ifind()
4560 and in rare cases can find an inode that is being freed.
4561 The ll_test_inode() code will free the lsm during ifind()
4562 when it finds an existing inode and then the VFS later attaches
4563 this free lsm to a new inode.
4567 Bugzilla : 6422, 7030
4568 Description: MDS deadlock between mkdir and client eviction
4569 Details : Creating a new file via mkdir or mknod (starting a transaction
4570 and getting the ns lock) can deadlock with client eviction
4571 (gets ns lock and trying to finish a synchronous transaction).
4574 Frequency : occasional
4575 Description: While starting a server, the fsfilt_ext3 module could not be
4577 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4578 kernels. Previously, lconf would still use the ext3 name
4579 when trying to load modules. Now, it will correctly use
4580 ext3 on 2.4 and ldiskfs on 2.6.
4582 Severity : enhancement
4583 Description: The default stripe count has been changed to 1
4584 Details : The interpretation of the default stripe count (0, to lfs
4585 or lmc) has been changed to mean striping across a single
4586 OST, rather than all available. For general usage we have
4587 found a stripe count of 1 or 2 works best.
4589 Severity : enhancement
4590 Description: Add support for compiling against Cray portals.
4591 Details : Conditional compiling for some areas that are different
4595 Frequency : occasional
4596 Bugzilla : 6409, 6834
4597 Description: Creating files with an explicit stripe count may lead to
4598 a failed assertion on the MDS
4599 Details : If some OSTs are full or unavailable, creating files may
4600 trigger a failed assertion on the MDS. Now, Lustre will
4601 try to use other servers or return an error to the
4605 Frequency : occasional
4607 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4608 caused lock timeout during readahead (since 1.4.2).
4609 Details : Processes doing readahead might match a lock that hasn't been
4610 granted yet if there are overlapping and conflicting lock
4611 requests. The readahead process waits on ungranted lock
4612 (original lock is CBPENDING), while OST waits for that process
4613 to cancel CBPENDING read lock and eventually evicts client.
4615 Severity : enhancement
4617 Description: Initial enabling of flock support for clients
4618 Details : Implements fcntl advisory locking and file status functions.
4619 This feature is provided as an optional mount flag (default
4620 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4621 locking are implemented yet, and those that are are not guaranteed
4622 to be completely correct in production environments.
4623 mount -t lustre -o [flock|noflock] ...
4626 Frequency : occasional
4628 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4629 assertion in the ext3 JBD (journaling) layer.
4630 Details : The b_committed_data struct is protected by the big kernel lock
4631 in 2.4 kernels, serializing journal_commit_transaction() and
4632 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4633 access to this struct. In 2.6 kernels there is finer grained
4634 locking to improve SMP performance of the JBD layer.
4638 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4643 Description: merge and cleanup kernel patches.
4646 -----------------------------------------------------------------------------
4648 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4650 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4651 versions of Lustre. Please update all servers and clients to
4652 version 1.4.2 or later at the same time. You must also run
4653 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4654 to update the configuration logs.
4656 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4657 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4658 - swab llog records of type '0' so we get proper header size/idx (5861)
4659 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4660 - fix rename of one directory over another leaking an inode (5953)
4661 - avoid SetPageDirty on 2.6 (5981)
4662 - don't re-add just-being-destroyed locks to the waiting list (5653)
4663 - when creating new directories, inherit the parent's custom
4664 striping settings if present parent (3048)
4665 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4666 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4667 - don't hold client locks on temporary worklist from l_lru (5666)
4668 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4669 - automatically evict dead clients (5921)
4670 - Update file size properly in create+truncate+fstat case (6196)
4671 - Do not unhash mountpoint dentries, do not allow removal of
4673 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4674 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4675 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4676 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4677 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4678 - Correctly handle reads of files with no objects (6243)
4679 - lctl recover will also mark a device active if deactivate used (5933)
4681 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4682 - allow --write-conf on an MDS with different nettype than client (5619)
4683 - don't write config llogs to MDS for mounts not from that MDS (5617)
4684 - lconf should create multiple TCP connections from a client (5201)
4685 - init scripts are now turned off by default; run chkconfig --on
4686 lustre and chkconfig --on lustrefs to use them
4687 - upcalls are no longer needed for clients to recover to failover
4689 - add --abort-recovery option to lconf to abort recovery on device
4691 - add support for an arbitrary number of OSTs (3026)
4692 - Quota support protocol changes.
4693 - forward compatibility changes to wire structs (6007)
4694 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4695 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4696 - improved "lustre status" script
4697 - initialize blocksize for non-regular files (6062)
4698 - added --disable-server and --disable-client configure options (5782)
4699 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4700 - Vanilla 2.4.29 support
4701 - increase maximum number of obd devices to 520 (6242)
4702 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4703 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4704 SLES 9 2.6.5 kernels
4705 - limit stripes per file to 160 (the maximum EA size) (6093)
4707 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4710 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4711 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4712 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4713 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4714 - small ext3 extents cleanups and fixes (5733)
4715 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4716 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4717 - avoid needless client->OST connect, fix handle mismatch (5317)
4718 - fix DLM error path that led to out-of-sync client, long delays (5779)
4719 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4720 - fix several locking issues related to i_size (5492,5624,5654,5672)
4721 - don't move pending lock onto export if it is already evicted (5683)
4722 - fix kernel oops when creating .foo in unlinked directory (5548)
4723 - fix deadlock in obdfilter statistics vs. object create (5811)
4724 - use time_{before,after} to avoid timer jiffies wrap (5882)
4725 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4726 - Do not leak request if server was not able to process it (5154)
4727 - If mds_open unable to find parent dir, make that negative lookup(5154)
4728 - don't create new directories with extent-mapping (5909, 5936)
4730 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4731 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4732 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4733 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4734 - support for RHEL4 kernel on i686 (5773)
4735 - provide error messages when incompatible logs are encountered (5898)
4737 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4738 * version 1.4.0.10 (1.4.1 release candidate 1)
4740 - don't keep a lock reference when lock is not granted (4238)
4741 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4742 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4743 - drop import inflight refcount on signal_completed_replay error (5255)
4744 - unlock page after async write error during send (3677)
4745 - handle missing objects in filter_preprw_read properly (5265)
4746 - no transno return for symlink open, don't save no-trasno open (3440)
4747 - don't try to complete elan receive that already failed (4012)
4748 - free RPC server reply state on error (5406)
4749 - clean up thread from ptlrpc_start_thread() on error (5160)
4750 - readahead could read extra page into cache that wasn't ejected (5388)
4751 - prevent races in class_attach/setup/cleanup/detach (5260)
4752 - don't dereference de->d_inode after l_dput of de (5458)
4753 - use "int" for stripe value returned from lock_to_stripe (5544)
4754 - mballoc allocation and error-checking fixes in 2.6 (5504)
4755 - block device patches to fix I/O request sizes in 2.6 (5482)
4756 - look up hostnames for IB nals (5602)
4757 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4758 - don't start multiple acceptors for the same port (5277)
4759 - fix incorrect LASSERT in mds_getattr_name (5635)
4760 - export a proc file for general "ping" checking (5628)
4761 - fix "lfs check" to not block when the MDS is down (5628)
4763 - service request history (4965)
4764 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4765 - create an "evict_client" /proc entry on OSTs, like the MDS has
4766 - fix mount usage message, return errors per mount(8) (5168)
4767 - change grep [] to grep "[]" in tests so they work in more UMLs
4768 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4769 - remove ancient LOV_MAGIC_V0 EA support (5047)
4770 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4771 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4772 - allow readahead allocations to fail when low on memory (5383)
4773 - mmap locking landed again, after considerable improvement (2828)
4774 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4776 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4779 - send OST transaction number in read/write reply to free req (4966)
4780 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4781 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4782 - the watchdog thread now runs as interruptible (5246)
4783 - flock/lockf fixes (but it's still disabled, pending 5135)
4784 - don't use EXT3 constants in llite code (5094)
4785 - memory shortage at startup could cause assertion (5176)
4787 - reorganization of lov code
4788 - single portals codebase
4790 - add extents/mballoc support (5025)
4791 - direct I/O reads in the obdfilter (4048)
4792 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4794 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4797 - send OST transaction number in read/write reply to free req (4966)
4798 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4799 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4800 - the watchdog thread now runs as interruptible (5246)
4801 - handle missing objects in filter_preprw_read properly (5265)
4802 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4803 - drop import inflight refcount on signal_completed_replay error (5255)
4804 - unlock page after async write error during send (3677)
4805 - return original error code on reconstructed replies (3761)
4806 - no transno return for symlink open, don't save no-trasno open (3440)
4808 - add pid to ldlm debugging output (4922)
4809 - bump the watchdog timeouts -- we can't handle 30sec yet
4810 - extra debugging for orphan dentry/inode bug (5259)
4812 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4815 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4816 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4817 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4818 - fix NULL dereference in /proc/sys/portals/routes (4827)
4819 - allow failed mdc_close() operations to be interrupted (4561)
4820 - stop precreate on OST before MDS would time out on it (4778)
4821 - don't send partial-page writes before EOF from client (4410)
4822 - discard client grant for sub-page writes on large-page clients (4520)
4823 - don't free dentries not owned by NFS code, check generation (4806)
4824 - fix lsm leak if mds_create_objects() fails (4801)
4825 - limit debug_daemon file size, always print CERROR messages (4789)
4826 - use transno after validating reply (3892)
4827 - process timed out requests if import state changes (3754)
4828 - update mtime on OST during writes, return in glimpse (4829)
4829 - add mkfsoptions to LDAP (4679)
4830 - use ->max_readahead method instead of zapping global ra (5039)
4831 - don't interrupt __l_wait_event() during strace
4833 - add software watchdogs to catch hung threads quickly (4941)
4834 - make lustrefs init script start after nfs is mounted
4835 - fix CWARN/ERROR duplication (4930)
4836 - return async write errors to application if possible (2248)
4837 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4838 - print NAL number in %x format (4645)
4839 - update barely-supported suse-2.4.21-171 series (4842)
4840 - support for sles 9 %post scripts
4841 - support for building 2.6 kernel-source packages
4842 - support for sles km_* packages
4844 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4847 - ignore -ENOENT errors in osc_destroy (3639)
4848 - notify osc create thread that OSC is being cleaned up (4600)
4849 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4850 - reconstruct ost_handle() like mds_handle() (4657)
4851 - create a new thread to do import eviction to avoid deadlock (3969)
4852 - let lconf resolve symlinked-to devices (4629)
4853 - don't unlink "objects" from directory with default EA (4554)
4854 - hold socknal file ref over connect in case target is down (4394)
4855 - allow more than 32000 subdirectories in a single directory (3244)
4856 - fix blocks count for O_DIRECT writes (3751)
4857 - OST returns ENOSPC from object create when no space left (4539)
4858 - don't send truncate RPC if file size isn't changing (4410)
4859 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4860 - bind to privileged port in socknal and tcpnal (3689)
4862 - rate limit CERROR/CWARN console message to avoid overload (4519)
4863 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4864 - basic mmap support (3918)
4865 - kernel patch series update from b1_4 (4711)
4867 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4870 - avoid crash during MDS cleanup with OST shut down (2775)
4871 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4872 - don't use bad inodes on the MDS (3744)
4873 - dynamic object preallocation to improve recovery speed (4236)
4874 - don't hold spinlock over lock dumping or change debug flags (4401)
4875 - don't zero obd_dev when it is force cleaned (3651)
4876 - print grants to console if they go negative (4431)
4877 - "lctl deactivate" will stop automatic recovery attempts (3406)
4878 - look for existing locks in ldlm_handle_enqueue() (3764)
4879 - don't resolve lock handle twice in recovery avoiding race (4401)
4880 - revalidate should check working dir is a directory (4134)
4882 - don't always mark "slow" obdfilter messages as errors (4418)
4884 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4887 - don't close LustreDB during write_conf until it is done (3860)
4888 - fix typo in lconf for_each_profile (3821)
4889 - allow dumping logs from multiple threads at one time (3820)
4890 - don't allow multiple threads in OSC recovery (3812)
4891 - fix debug_size parameters (3864)
4892 - fix mds_postrecov to initialize import for llog ctxt (3121)
4893 - replace config semaphore with spinlock (3306)
4894 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4895 - don't allow enqueue to complete on a destroyed export (3822)
4896 - down write_lock before checking llog header bitmap (3825)
4897 - recover from lock replay timeout (3764)
4898 - up llog sem before sending rpc (3652)
4899 - reduce ns lock hold times when setting kms (3267)
4900 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4901 - fix NULL deref and obd_dev leak on setup error (3312)
4902 - replace some LBUG about llog ops with error handling (3841)
4903 - don't match INVALID dentries from d_lookup and spin (3784)
4904 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4905 - fix invalid assertion in ptlrpc_set_wait (3880)
4907 - add libwrap support for the TCP acceptor (3996)
4908 - add /proc/sys/portals/routes for non-root route listing (3994)
4909 - allow setting MDS UUID in .xml (2580)
4910 - print the stack of a process that LBUGs (4228)
4912 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4915 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4916 - make sure to unset replay flag from failed open requests (3440)
4917 - if default stripe count is 0, use OST count for inode size (3636)
4918 - update parent mtime/ctime on client for create/unlink (2611)
4919 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4920 - free recovery state on server during a forced cleanup (3571)
4921 - unregister_reply for resent reqs (3063)
4922 - loop back devices mounting and status check on 2.6 (3563)
4923 - fix resource-creation race that can provoke i_size == 0 (3513)
4924 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4925 - more debugging for page-accounting assertion (3746)
4926 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4927 - avoid deadlock after precreation failure (3758)
4928 - fix race and lock order deadlock in orphan handling (3450, 3750)
4929 - add validity checks when grabbing inodes from l_ast_data (3599)
4931 - add /proc/.../recovery_status to obdfilter (3428)
4932 - lightweight CDEBUG infrastructure, debug daemon (3668)
4933 - change default OSC RPC parameters to be better on small clusters
4934 - turn off OST read cache for files smaller than 32MB
4935 - install man pages and include them in rpms (3100)
4936 - add new init script for (un)mounting lustre filesystems (2593)
4937 - run chkconfig in %post for init scripts (3701)
4938 - drop scimac NAL (unmaintained)
4940 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4943 - clean kiobufs before and after use (3485)
4944 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4945 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4946 - change init script to fail more clearly if not run as root (1528)
4947 - allow clients to reconnect during replay (1742)
4948 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4949 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4950 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4951 - fix cond_resched (3554)
4952 - only evict unfinished clients after recovery (3515)
4953 - allow bulk resend, prevent data loss (3570)
4954 - dynamic ptlrpc request buffer allocation (2102)
4955 - don't allow unlinking open directory if it isn't empty (2904)
4956 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4957 - remove extraneous obd dereference causing LASSERT failure (3334)
4958 - don't use get_cycles() when creating temp. files on the mds (3156)
4959 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4960 - handle EEXIST for set-stripe, set proper directory name (3336)
4962 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4963 - fix ksocknal_fmb_callback() error messages (2918)
4965 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4968 - don't copy lvb into (possibly NULL) reply on error (2983)
4969 - don't deref dentry after dput, don't free lvb on error (2922)
4970 - use the kms to determine writeback rpc length (2947)
4971 - increment oti_logcookies when osc is inactive (2948)
4972 - update client's i_blocks count via lvb messages (2543)
4973 - handle intent open/close of special files properly (1557)
4974 - mount MDS with errors=remount-ro, like obdfilter (2009)
4975 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4976 - don't use cancelling-locks' kms values (2947)
4977 - use highest lock extent for kms, not last one (2925)
4978 - don't dereference ERR_PTR() dentry in error handling path (3107)
4979 - fix thread race in portals_debug_dumplog() (3122)
4980 - create lprocfs device entries at setup instead of at attach (1519)
4981 - common AST error handler, don't evict client on completion race (3145)
4982 - zero nameidata in detach_mnt in 2.6 (3118)
4983 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4984 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4985 - initialize RPC timeout timer earlier for 2.6 (3219)
4986 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4987 - print nal/nid for unknown nid (3258)
4988 - additional checks for oscc recovery before doing precreate (3284)
4989 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4990 - don't crash in mdc_close for bad permissions on open (3285)
4991 - zero i_rdev for non-device files (3147)
4992 - clear page->private before handing to FS, better assertion (3119)
4993 - tune the read pipeline (3236)
4994 - fix incorrect decref of invalidated dentry (2350)
4995 - provide read-ahead stats and refine rpc in flight stats (3328)
4996 - don't hold journal transaction open across create RPC (3313)
4997 - update atime on MDS at close time (3265)
4998 - close LDAP connection when recovering to avoid server load (3315)
4999 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5000 - don't leak open file on MDS after open resend (3325)
5001 - serialize filter_precreate and filter_destroy_precreated (3329)
5002 - loop device shouldn't call sync_dev() for nul device (3092)
5003 - clear page cache after eviction (2766)
5004 - resynchronize MDS->OST in background (2824)
5005 - refuse to mount the same filesystem twice on same mountpoint (3394)
5006 - allow llmount to create routes for mounting behind routers (3320)
5007 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5008 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5009 - fix rare problem with rename on htree directories (3417)
5011 - allow default OST striping configuration per directory (1414)
5012 - fix compilation for qswnal for 2.6 kernels (3125)
5013 - increase maximum number of MDS request buffers for large systems
5014 - change liblustreapi to be useful for external progs like lfsck (3098)
5015 - increase local configuration timeout for slow disks (3353)
5016 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5018 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5021 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5022 - don't overwrite extent policy data in reply if lock was blocked (2901)
5023 - drop filter export grants atomically with removal from device (2663)
5024 - del obd_self_export from work_list in class_disconnect_exports (2908)
5025 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5026 - swab reply message in mdc_close, other PPC fixes (2464)
5027 - fix destroying of named logs (2325)
5028 - overwrite old logs when running lconf --write_conf (2264)
5029 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5030 - fix race in target_handle_connect (2898)
5031 - mds_reint_create() should take same inode create lock (2926)
5032 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5033 - don't close files for self_export to avoid uninitialized obd (2936)
5034 - allow MDS with the same name as client node (2939)
5035 - hold dentry reference for closed log files for unlink (2325)
5036 - reserve space for all logs during transactions (2059)
5037 - don't evict page beyond end of stripe extent (2925)
5038 - don't oops on a deleted current working directory (2399)
5039 - handle hard links to targets without a parent properly (2517)
5040 - don't dereference NULL lock when racing during eviction (2867)
5041 - don't grow lock extents when lots of conflicting locks (2919)
5043 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5046 - account for cache space usage on clients to avoid data loss (974)
5047 - lfsck support in lustre kernel code (2349)
5048 - reduce journal credits needed for BRW writes (2370)
5049 - orphan handling to avoid losing space on client/server crashes
5050 - ptlrpcd can be blocked, stopping ALL progress (2477)
5051 - use lock value blocks to assist in proper KMS, faster stat (1021)
5052 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5053 - recovery for initial connections (2355)
5054 - fixes for mds_cleanup_orphans (1934)
5055 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5056 - block all file creations until orphan recovery completes (1901)
5057 - client remove rq_connection from request struct (2423)
5058 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5059 - recovery timer race (2670)
5060 - mdc_close recovey bug (2532)
5061 - ptlrpc cleanup bug (2710)
5062 - mds timeout on local locks (2588)
5063 - namespace lock held during RPCs (2431)
5064 - handle interrupted sync write properly (2503)
5065 - don't try to handle a message that hasn't been replied to (2699)
5066 - client assert failure during cleanup after abort recovery (2701)
5067 - leak mdc device after failed mount (2712)
5068 - ptlrpc_check_set allows timedout requests to complete (2714)
5069 - wait for inflight reqs when ptlrpcd finishes (2710)
5070 - make sure unregistered services are removed from the srv_list
5071 - reset bulk XID's when resending them (caught by 1138 test)
5072 - unregister_bulk after timeout
5073 - fix lconf error (2694)
5074 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5075 - readahead locks pages, leaves pending causing memory pressure (2673)
5076 - increase OST request buffers to 4096 on large machines (2729)
5077 - fix up permission of existing directories in simple_mkdir (2661)
5078 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5079 - don't assign transno to errored transactions (2742)
5080 - don't delete objects on OST if given a bogus objid from MDS (2751)
5081 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5082 - if rq_no_resend, then timeout request after recovery (2432)
5083 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5084 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5085 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5086 - MDS recovery shouldn't delete objects that it has given out (2730)
5087 - if enqueue arrives after completion, don't clobber LVB (2819)
5088 - don't unlock pages twice when trigger_group_io returns error (2814)
5089 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5090 - don't write pages to disk if there was an error (1450)
5091 - don't ping imports that have recovery disabled (2676)
5092 - take buffered bytes into account when balancing socknal conn (2817)
5093 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5094 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5095 - remove little-endian swabbing of llog records (1987)
5096 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5097 - retry reposting request buffers if they fail (1191)
5098 - grow extent at grant time to avoid granting a revoked lock (2809)
5099 - lock revoke doesn't evict page if covered by a second lock (2765)
5100 - disable VM readahead to avoid reading outside lock extents (2805)
5102 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5103 - updated kernel patches for hp-2.4.20 kernel (2681)
5105 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5108 - fix truncated write corruption (2366)
5109 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5111 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5112 - ensure len(uuid) < 37 in lmc (1171)
5113 - fix ia64 OOPS in llog_test (2255)
5114 - zero end of page at obdfilter for partial page writes (2648)
5115 - don't leave stale dentries around after renames (bug 2428)
5116 - fix timeouts when evicting a client with a single lock held (2642)
5117 - set deadline for the initial HELLO message to drain (2634)
5118 - print out dotted-quad IP addresses in the socknal (2302)
5120 - additional debugging for MDS client eviction problem (2443)
5121 - fix mkfsoptions support for osts (2603, 2604)
5123 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5126 - add series for the vanilla 2.6.0 kernel
5127 - add series for the vanilla 2.4.24 kernel
5128 - add series for a cray x86/64 UL kernel drop
5129 - fix xattr patches for the vanilla 2.4.19 series
5131 - generate true UUIDs in lmc (1171)
5132 - have portals stack dumping break in UML (2466)
5133 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5134 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5135 - wake callback waiting threads on client eviction (2460)
5136 - Add --ptldebug and --subsystem to lmc (1719)
5137 - update assertion to allow safe interrupt allocation
5138 - set rq_no_resend for cancel requests (2432)
5139 - recalculate ptlrpcd timeout after resend (2494)
5140 - call vfs_rmdir when removing pending directories (2368)
5141 - fix renaming a file to itself (2429)
5142 - lmc creates a default one-stripe lov (2454)
5143 - expand procfs space to handle large clusters (2326)
5144 - increase UML stack to avoid overflow
5145 - update lconf's list of debug and subsystem masks
5146 - fix lfs find --obd (2510)
5147 - /proc tunable for disabling filter read caching (2591)
5148 - stop rpm packages from altering slapd.conf (2301)
5149 - disable nagle in the socknal under 0conf (2578)
5150 - choose mds inode size based on stripe count (2572)
5151 - fix kernel-source rpm problems (2516)
5153 - add --disable-doc to avoid pdf generation (2421)
5154 - update documentation, tests, type-os, comments
5155 - avoid format warnings on ia64
5156 - remove the TOE NAL
5157 - tiny code cleanups by removing unused fields
5159 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5162 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5163 - remove the most blatant lies from BUILDING file (2371)
5164 - change default debug level to reasonable production setting
5165 - reduce client side cache size to reduce cache flush time
5166 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5167 - make TCP zerocopy and pinger support enabled by default (2476)
5168 - sync writes completed after process exits caused crashes (2319)
5169 - maintain correct mount count on the MDS (2356)
5170 - backout 1557, because 2316 wasn't really fixed
5171 - better file I/O statistics gathering in /proc
5172 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5173 - another kernel patch to fix zero-copy TCP function export
5174 - don't take duplicate lock when processing re-sent getattr (2420)
5175 - lctl uses obd_self_export instead of creating new conn (2353)
5176 - MDS/OST recovery case which requires object creation asserted (2425)
5177 - move lfs from /usr/sbin to /usr/bin in packages
5178 - fix race between mds_client_add and mds_client_free (2417)
5179 - use kmalloc instead of slabs in portals (2430)
5180 - don't create duplicate records when a failover MDS is present (2442)
5181 - remove unnecessary mount age check (2332)
5182 - don't remove directory inodes from locks prematurely (2451)
5183 - don't break if MDS service name is the same as hostname (2103)
5184 - fix races in client write RPC generation when cache full (2482)
5186 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5189 - remove now-unused request->rq_obd (278)
5190 - if an allocation fails, print out how much memory we've used (1933)
5191 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5192 - add the "configurable stack size" patch to most series files (1256)
5193 - ability to write large log records, for 100+ OST configs (2306)
5194 - fix NULL deref when filter_prep fails (2314)
5195 - fix operator precedence error in filter_sync
5196 - dynamic allocation of socknal TX descriptors (2315)
5197 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5198 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5199 - ensure that CWARN messages in Portals always get to the syslog
5200 - __init/__exit are not for prototype decls (ldlm_init/exit)
5201 - x86-64 compile warning fixes
5202 - fix gateway LMC keyword conflict (2318)
5203 - fix MDS lock inversions in getattr/reint paths (1844)
5204 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5205 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5206 - DLM race condition prevented some lock evictions (2328)
5207 - ENOMEM detection and retry on socknal sends (2230)
5208 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5209 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5210 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5211 - better error messages when a client is rejected during recovery (1505)
5212 - avoid cancelling locks which were never granted, after failure (2330)
5213 - fix i_sem/journal inversion in mds_client_add (2333)
5214 - fix truncate/getattr lock cycle deadlock (2334)
5215 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5216 - fix two rare exit paths which could leak an l_lock() ref (2321)
5217 - fencepost error in MDS/OST orphan recovery (2226)
5218 - make log record alignment 8 bytes (1988)
5219 - lstripe now fails when requested offset > ost_count (2237)
5220 - ensure that all kernel series have a complete list.h (1607)
5221 - fix crashes in special-file operations (2316)
5222 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5224 - allow configurable automake binary, for testing new versions
5225 - small update to the lfs documentation
5227 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5229 * fix negative export reference count in fsfilt_sync (2312)
5231 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5232 * release candidate 0.9.1
5234 - orphans are moved into the PENDING directory for possible recovery
5235 - replayed opens now open by fid for orphan/rename safety (1042)
5236 - last close of an orphan inode generates a transno (683)
5237 - chdir() and mount() now pin the directory entry (1020)
5238 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5239 - discard very old requests without processing them (1502)
5240 - remove some common, well-understood CERRORs (1505)
5241 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5242 - clear "grant" flags in OST replies until OST grant code lands (1644)
5243 - fix read performance by not clobbering i_blksize on client (1598)
5244 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5245 - make LRU size a /proc tunable, clears locks when reduced (707)
5246 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5247 - server locks take references on exports now (1558)
5248 - build fixes for 2.4.20-rh trees (1663)
5249 - return an error from lov_create if all OSCs are inactive (1751)
5250 - fix import levels when a reconnect happens without a timeout (1597)
5251 - exit early from mds_open if we get a lookup error (1749)
5252 - partial page read at EOF wouldn't wait for disk before sending (1642)
5253 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5254 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5255 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5256 - write/truncate lock inversion (1639)
5257 - Don't auto-load obdclass, portals modules during cleanup (1495)
5258 - fix timestamps from jumping to "now" (1763)
5259 - extra journal assertions (1648)
5260 - add an extra multiunlink test (1771)
5261 - fix read_record/write_record API (1776)
5262 - fix leak of offset_extent, possible incorrect i_size later (1772)
5263 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5264 - Debugging for the kqswnal_get_idle_tx problems (1820)
5265 - Allow recovery to be attempted multiple times (1536)
5266 - Write out MDS last_rcvd file after it is first created (1600)
5267 - Fix tx_descriptor leak in failed transmit situations (1827)
5268 - ext3 journaling fixes for assertion failure after IO error (1871)
5269 - class_export_put() on freed export after completion AST error (1896)
5270 - Fix revalidate looping in VFS (1322)
5271 - Don't access a freed export during MDS_REINT timeout (1521)
5272 - Add open-unlink recovery support on the MDS (1673,1764)
5273 - Return an error if no MDS data was read from last_rcvd (1946)
5274 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5275 - Don't setup a disk device that doesn't match exported UUID (317)
5276 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5277 - avoid committing NULL handle in force close
5278 - local.sh is now a one-stripe LOV configuration
5279 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5280 - fix ext3 htree duplicate directory entry corruption (1516)
5281 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5282 - update to new LOV EA format (2097)
5283 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5285 2003-06-15 Phil Schwan <phil@clusterfs.com>
5288 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5289 - per-import/export recovery handling (958, 931, 959)
5290 - multiple last-rcvd slots, for serving multiple FSes (949)
5291 - connections are again shared between multiple imp/exports (963, 964)
5292 - "umount -f" would hang if any requests needed to be sent (393, 978)
5293 - avoid pinning large req buffer by copying for queued messages (989)
5294 - add "uuid" to "lctl device" command to help upcalls (991)
5295 - "open" RPCs with transnos would confuse recovery counters (1037)
5296 - do proper endian conversion of all wire messages (288, 340, 891)
5297 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5298 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5299 - fix ptlrpc_connection leak in target_handle_connect (1174)
5300 - fix import refcounting bug in OST and MDS cleanup (1134)
5301 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5302 - fix very unlikely obd_types race condition (501)
5303 - remove osc_open hack for echo_client (1187)
5304 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5305 - a failure in read_inode2 leads to deadlock (1139)
5306 - cancel ack-locks as soon as transaction is committed (1072)
5307 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5308 - make sure to commitrw after any preprw to avoid deadlock (1162)
5309 - failing to execute a file in a lustre FS would lock inode (1203)
5310 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5311 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5312 - obd_uuid2tgt would walk off the end of the list (1255)
5313 - on IA64 the osc would give portals incorrect bulk size (1258)
5314 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5315 - fix lock inversion caused by new llite matching code (1282)
5316 - limit the number of dirty pages on a client to 10MB (1286)
5317 - timed out locks were not being corrected cancelled (1289)
5318 - fix O_DIRECT above 4GB on IA-32 (1292)
5319 * major user-visible changes
5320 - fail out/fail over policy now controlled by the upcall (993)
5322 - add OBD_PING to check server availability and failure (954)
5323 - lustre messages are now sent in sending host order (288, 340, 891)
5324 - add eadatalen to MDS getattr reply (340)
5325 - OST read replies may contain second buffer, with per-page status (593)
5327 2003-03-11 Phil Schwan <phil@clusterfs.com>
5330 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5331 - failed open()s could cause deadlock; fixed (867, 869)
5332 - stop cancelling OST locks when files are closed (481)
5333 - overlapping XID spaces caused network corruption (851, 853)
5334 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5335 - setattr_raw added, to do single-RPC, server-side setattrs
5336 - lmc/lconf syntax change for OST UUIDs
5337 - fix crashy race condition between ptlrpc_free_req and osc_close
5338 - don't use request in mdc_enqueue if we hit a timeout (889)
5339 - don't set the inode i_size for regular files from the MDS (896)
5340 - handle out of order completion AST (842)
5341 - don't LBUG if a lock request times out after receiving AST (913)
5342 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5343 - if a bad lock AST arrives, send an error instead of dropping entirely
5344 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5345 - fix leak in bulk IO when only partially completed (899, 900, 926)
5346 - fix O_DIRECT for ia64 (55)
5347 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5348 - C-z after timeout could hang a process forever; fixed (977)
5350 - client-side I/O cache (678, 924, 929, 941, 970)
5352 - READPAGE and SETATTRs which don't take server-side locks get
5355 2003-02-11 Phil Schwan <phil@clusterfs.com>
5358 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5359 - Fix fsfilt_extN_readpage() to read a full page of directory
5360 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5361 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5362 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5363 - Most intent operations were removed, in favour of a new RPC mode
5364 that does a single RPC to the server and bypasses most of the VFS
5365 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5366 - Aggressively cancel local locks on DLM servers
5367 - mds_reint_unlink sends EA to the client if it's the last nlink.
5368 client uses that EA to unlink OST objects.
5369 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5370 - recursive symlinks were fixed (439)
5371 - fixed NULL deref in DEBUG_REQ
5372 - filter_update_lastobjid no longer calls sync, which annoyed extN
5373 - fixed multi-client small-writes to a single file problem (445)
5374 - fixed mtime updates during file writes (607)
5375 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5376 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5377 - fixed Linux OST size reporting problem (444, 656)
5378 - OST now updates object mtime with writes or setattr (607, 619)
5379 - client verifies file size before zeroing page past EOF (445)
5380 - OST now writes last allocated objid to disk with allocation (108)
5381 - LOV on echo now works (409)
5383 - mds_reint_unlink sends a new buffer, with the EA included. this
5384 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5385 set if a regular file was being unlinked, and it was the last link
5386 - use PtlGet from the target for bulk writes (315)
5387 - OST now updates object mtime with writes or setattr (607, 619)
5388 - LDLM now has a grant-time callback to revalidate locked items, if
5390 - Many MDS operations were reorganized to combat race conditions
5392 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5393 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5395 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5398 - Fully reactivate OST imports after reconnection (512, others)
5399 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5400 - More graceful error handling for truncating on dead OST (515)
5401 - Don't error out unless we're actually accessing dead stripes (474)
5402 - Fix garbage sizes when stripes are missing (410)
5403 - LRU counters were broken, causing constant lock purge (433, 432)
5404 - garbage on read from stripes with failed OSTs (441)
5405 - mark OSCs as active before reconnecting during recovery (438)
5406 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5407 - lfind did not preserve OST order in output (443)
5408 - symlinks cause hung clients, incorrect data (439)
5409 - stop dereferencing request after dropping refcount (457)
5410 - don't LASSERT(spin_is_locked) on non-SMP (455)
5411 - fixes for many rename() bugs
5412 - fstat didn't correctly synchronize attributes (399)
5413 - server must handle lock cancellation during blocking AST prep (487)
5414 - bulk descriptors were free()d too soon (511)
5415 - fix paths in lconf, which would load incorrect modules (451, 507)
5416 - fix confusing lconf 'host not found' error message (386)
5417 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5418 - fix race condition in mdc_blocking_ast() for inode access (526)
5419 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5420 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5421 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5422 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5423 - fix lstripe in several ways (kernel side) (527)
5424 - fix request leak in ldlm_cli_enqueue (262)
5425 - incorrect OSC was marked inactive after OST failure
5426 - call mds_fs_cleanup before unmounting filesystem (524)
5427 - fix races between taking ns_lock and ldlm_lock_change_resource
5428 - fix races updating LOV export open file list
5429 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5430 - fix recovery NULL deref in ldlm_cli_cancel_unused
5431 - fix some DLM races by using new hash table for lock handles (419)
5432 - permit the client to specify desired inodes, at replay
5433 - duplicate requests when we queue them for replay reintegration
5434 - fix last_rcvd offset calculation
5435 - sync after each recovered transaction, so we always make progress
5436 - never, not always, ERESTART requests without transnos
5437 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5439 - skip replay if the MDS says that the client is already connected
5440 - don't check for a recovery-enabled export to match lctl's UUID
5441 - don't INC_USE_COUNT for phantom exports
5442 - don't crash when cleaning up phantom exports (567)
5443 - don't double-finish or set replay data for errored mdc_open requests
5444 - abort requests when they time out, so we don't get old replies
5445 - send/receive replies for AST messages again
5446 - if the client says that it doesn't have the lock, cancel it on the
5448 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5449 mark it as destroyed, it will all work out when decref is called
5450 - fix module use counts (22, 581)
5452 - ASTs now expect a reply (server cancels lock on error reply)
5454 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5457 - fix many simultaneous client startup (392)
5458 - fix dentry->d_it clobbering
5459 - credentials weren't being shipped for readdir/getattr operations
5460 - remove invalid assertions triggered during some concurrent MD
5462 - proper Lustre versions added (336, 389)
5463 - fix memory leak for create error case (398)
5464 - fix LOV locking bug that would get cli/srv out of sync
5465 - fix echo client over LOV (409)
5466 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5467 - fix double-O_EXCL intent crash (424)
5468 - avoid sending multiple lock CANCELs (352)
5470 - MDS can do multi-client recovery (modulo bugs in new code)
5472 - many updates, edits, cleanups
5474 2002-11-18 Phil Schwan <phil@clusterfs.com>
5477 - fix null d_it dereference (346)
5478 - fix full OST/dbench hang (333)
5479 - fix permission problem with file removal (286)
5480 - fix removal of OSCs from LOV when they fail
5481 - fix NULL deref during bulk timeout (214)
5482 - fix problems related to multiple filesystems on one MDS (241)
5483 - fixed serious subtle metadata locking bugs
5484 - free locks on clients when inodes are removed due to memory
5486 - fix inode pointer in lock data (285)
5487 - partial support for multiple MDS on a single host (241)
5488 - data locks weren't cancelled at clear_inode time (290, 311)
5489 - intent locks could lead to unbounded lock growth (205)
5490 - added a maximum lock count, an LRU list, and a flusher
5491 - fix multiple rename (365)
5492 - properly abstracted the echo client
5493 - OSC locked 1 byte too many; fixed
5494 - rewrote brw callback code:
5495 - fixed recovery bugs related to LOVs (306)
5496 - fixed too-many-pages-in-one-write crash (191)
5497 - fixed (again) crash in sync_io_timeout (214)
5498 - probably fixed callback-related race (385)
5500 - Add capability to MDS protocol
5501 - LDLM cancellations and callbacks on different portals
5503 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5506 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5507 - defer open object destruction to close time (601981, 138)
5508 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5509 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5510 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5511 - don't call obd_setattr in ll_file_release for destroyed objects
5512 * protocol change to lustre_msg: move |version| and add |flags|
5513 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5514 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5515 * added replay of create, unlink, link and rename operations during
5516 MDS failover; recovery should be much more robust now
5517 * remove failed OSCs from LOVs (only lov_create uses this so far)
5518 * the lustre-HOWTO was brought (more) up to date (582544)
5520 2002-10-23 Phil Schwan <phil@clusterfs.com>
5523 - in-use dentries weren't being reused properly (617851)
5524 - prevent multiple LDLM setup (599178)
5525 - fix LOV size calculations for truncate (617853)
5526 - fix client handling of MDS intent errors (POSIX)
5527 - fix permission bug in lovstripe.c test (624321)
5528 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5529 - truncate past end of file could corrupt data
5530 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5531 - a race in recovery could return ETIMEDOUT to apps (623947)
5532 - building outside the source directory was fixed
5533 * the lustre-HOWTO was brought (more) up to date (582544)
5534 * major progress was made on recovery functionality
5536 2002-10-10 Phil Schwan <phil@clusterfs.com>
5539 - recovery deadlock fix
5540 - rm -rf causes LBUG fix (617817)
5541 - file open by multiple tasks fix (618962)
5542 - directory permissions bugs (602707 and 620007)
5543 - journal_stop fixed with locking (611313)
5544 - O_APPEND failures resolved (618273, perhaps 614459)
5545 - lconf PATH fix (619770)
5546 - IA64 build fix (621450)
5547 - RPC buffer sizes scale with amount of memory
5549 2002-10-01 Phil Schwan <phil@clusterfs.com>
5552 - locks would be cancelled without throwing away data pages,
5553 resulting in inconsistent data (605627)
5554 - inode attributes were not always being refreshed (605627, 612449)
5555 - lconf now continues to cleanup after lctl reports an error
5556 - MDS now enforces user permissions (602707)
5557 - lprocfs cleanup fixed, but not yet enabled (614157)
5558 - fixed infinite server hang, should a client not respond to an AST
5559 - avoid going into recovery if user calls readlink() with a buffer
5560 that's too small (613941)
5561 - AST RPCs no longer require replies (614867) -- this may be changed
5562 - don't crash server if client sends an IOV that's too big (611336)
5563 - fixed lock conversion deadlock (611892)
5564 - fixed the following of symlinks (614622)
5565 * recovery: the server can remove locks from a client that dies, other
5566 clients can make progress
5567 * more extN patch fixes
5568 * compile-time configurable ptlrpc buffer allocations
5570 - collaborative read cache document
5571 - Lustre Lite Performance CDR document-in-progress
5573 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5576 - fix typo in patch-2.4.18
5578 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5581 - clear ptlrpc request each time in handle_incoming_request()
5582 - unlink of files now destroys the object on the OST
5584 2002-09-19 Peter Braam <braam@clusterfs.com>
5586 * add hard link support
5587 * change obdfile creation method
5588 * kernel patch changed
5590 2002-09-19 Peter Braam <braam@clusterfs.com>
5593 - stack overflow bug in extN fixed
5595 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5597 * documentation updates
5598 - add man pages for config tools
5599 - update tests/README to describe testing with new config tools
5600 - finish metadata API descriptions
5601 * bug fixes and cleanups
5602 - statfs workaround for 16TB limit
5603 - LOV stripe allocation improved, can stripe on subset of OSTs
5604 - LOV file size/IO offset was wrong for files > 4GB in size
5605 - object EA data was being dropped, caused files to be unreadable
5606 - memory overflow with non-LOV OST caused memory corruption
5607 - fixed regression tests to work with new config tools, obdfilter
5608 - fixed bug when directory size became larger than 1 block
5609 - fixed bug (for single client case) when PWD was deleted
5610 - invalidate local directory pages when doing intent-based ops
5611 - avoid LDLM oops when lock callback contained bad data
5613 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5615 * documentation updates
5616 * bug fixes and cleanups
5617 - configuration tools
5620 - 64-bit compile warnings
5621 - 64-bit internal statfs data
5623 * test_brw on persistent OST devices
5625 * lprocfs (disabled)
5627 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5629 * documentation updates
5630 * bug fixes and cleanups
5631 * configuration tools
5633 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5636 * many small fixes to 0_5_4
5637 * io/network handling
5638 * thinkos in MDS operations
5640 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5643 * crucial basic fixes to 0.5.3
5644 * IOR, Iozone work over Elan
5647 2002-08-07 Phil Schwan <phil@clusterfs.com>
5648 * version 0_5_3, our first alpha
5649 * we use the new Portals iovs
5650 * documentation updates
5651 * bug fixes and cleanups
5652 * small changes in the DLM wire protocol
5654 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5655 * version 0_5_1 with some initial stability,
5656 * locking on MD and file I/O.
5657 * documentation updates
5658 * several bug fixes since 0.5.0
5659 * small changes in wire protocol
5661 2002-07-18 Phil Schwan <phil@clusterfs.com>
5663 * delivered as Lustre Light Alpha
5664 * fixed a crash after handling invalid MDS requests
5665 * fixed directory pages for architectures with non-4k pages sizes
5667 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5668 * release version v0_4_4
5669 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5671 * Fixes a number of interruption problems with OST operations.
5672 * Update documentation for portals header changes
5673 * Move all wire protocol structs/defines to lustre_idl.h
5674 * Fixes symlink length bug.
5675 * Add tcpdump to repository.
5677 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5678 * release version v0_4_3
5679 * Fixes statfs for inodes on extN.
5680 * Fixes bug in runtests which would delete /etc/hosts.
5681 * Use 64-bit object IDs wherever possible (not into VFS though)
5682 Remove ost_get_info, which is unused by lustre, and out of date.
5684 2002-07-03 Peter Braam <braam@clusterfs.com>
5685 * release version v0_4_2 Fixes a lookup error (type not passed)
5686 * move forward to head of Portals
5687 * move forward to latest Lustre kernel
5689 2002-06-25 Peter Braam <braam@clusterfs.com>
5690 * release version v0_4_1. Hopefully stable on single node use.