1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.31 (SLES 10),
5 2.6.18-92.1.17.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 : on remount
19 Description: external journal device not working after the remount
20 Details : clear dev_rdonly flag for external journal devices in
24 Frequency : race on file read and write
26 Description: Lustre doesn't delete files
27 Details : Clients drop lock reference and release openhandle when they find
33 Description: shutdown vs evict race
34 Details : client_disconnect_export vs connect request race.
35 if client will evicted at this time - we start invalidate
36 thread without referece to import and import can be freed
40 Frequency : rare, need acl's on inode.
42 Description: client can't handle ost additional correctly
43 Details : if ost was added after client connected to mds client can have
44 hit lnet_try_match_md ... to big messages to wide striped files.
45 in this case need teach client to handle config events about add
46 lov target and update client max ea size at that event.
48 Severity : enhancement
50 Description: Changelogs
51 Details : Changelogs are a lightweight mechanism to track filesystem
52 metadata and namespace changes. The changelog is recorded
53 permanently on the MDTs, and is periodically "consumed" / purged
54 when records are no longer needed.
56 Severity : enhancement
58 Description: compact fld format with extents
59 Details : Store range of seq rather than every seq in FLD. Seq
60 controller update FLD rather than clients. In Case of CMD, mdt0
61 has FLD, all other metadata server act as non persistent proxy
62 for FLD queries and cache fld entries in fld cache.
67 Description: don't skip ost target if they assigned to file
68 Details : Drop slow OSCs if we can, but not for requested start idx.
69 This means "if OSC is slow and it is not the requested
70 start OST, then it can be skipped, otherwise skip it only
71 if it is inactive/recovering/out-of-space.
75 Description: more cleanup in mds_lov
76 Details : not send LOV EA under replay, we can't know about they size at this
77 time. Don't allow client connect to mds before any ost connected,
78 for avoid problems with LOV EA size and returning EIO to client.
80 Severity : enhancement
82 Description: Interoperability at server side (Disk interoperability)
84 Severity : enhancement
86 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
88 Severity : enhancement
90 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
92 Severity : enhancement
94 Description: New client IO stack (CLIO).
96 Severity : enhancement
98 Description: Commit on sharing. Eliminate inter-client dependencies between
99 uncommitted transactions by doing transaction commits.
100 Thereby clients may recovery independently.
103 Frequency : Create a symlink file with a very long name
105 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
106 Details : If there is no extra space in the request for early cancels,
107 ldlm_req_handles_avail() returns 0 instead of a negative value.
109 Severity : enhancement
111 Description: Add /proc entry for import status
112 Details : The mdc, osc, and mgc import directories now have
113 an import directory that contains useful import data for debugging
116 Severity : enhancement
118 Description: Re-disable certain /proc logging
119 Details : Enable and disable client's offset_stats, extents_stats and
120 extents_stats_per_process stats logging on the fly.
123 Frequency : Only on FC kernels 2.6.22+
125 Description: oops in statahead
126 Details : Do not drop reference count for the dentry from VFS when lookup,
127 VFS will do that by itself.
129 Severity : enhancement
131 Description: Generic /proc file permissions
132 Details : Set /Proc file permissions in a more generic way to enable non-
133 root users operate on some /proc files.
137 Description: Hitting mdc_commit_close() ASSERTION
138 Details : Properly handle request reference release in
139 ll_release_openhandle().
143 Description: quota recovery deadlock during mds failover
144 Details : This patch includes att18982, att18236, att18237 in bz14840.
146 1. fix osts hang when mds does failover with quotaon
147 2. prevent watchdog storm when osts threads wait for the
152 Frequency : only patchless client
153 Description: add workaround for race between add/remove dentry from hash
155 Severity : enhancement
157 Description: Allow OST glimpses to return PW locks
161 Description: LBUG when llog conf file is full
162 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
166 Description: Prevent import from entering FULL state when server in recovery
170 Description: service mount cannot take device name with ":"
171 Details : Only when device name contains ":/" will mount treat it as
177 Description: replace ptlrpcd with the statahead thread to interpret the async
178 statahead RPC callback
182 Frequency : on recovery
183 Description: I/O failures after umount during fail back
184 Details : if client reconnected to restarted server we need join to recovery
185 instead of find server handler is changed and process self
186 eviction with cancel all locks.
188 Severity : enhancement
190 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
192 Severity : enhancement
194 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
196 Severity : enhancement
198 Description: Upcall on Lustre log has been dumped
199 Details : Allow for a user mode script to be called once a Lustre log has
200 been dumped. It passes the filename of the dumped log to the
201 script, the location of the script can be specified via
202 /proc/sys/lnet/debug_log_upcall.
207 Description: avoid idr_remove called for id which is not allocated.
208 Details : Move assigment s_dev for clustered nfs to end of initialization,
209 to avoid problem with error handling.
214 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
215 Details : When connection is reused this not moved from CONN_UNUSED_HASH
216 into CONN_USED_HASH and this prodice warning when put connection
217 again in unused hash.
220 Severity : enhancement
222 Description: File striping can now be set to use an arbitrary pool of OSTs.
224 Severity : enhancement
226 Description: Export bytes_read/bytes_write count on OSC/OST.
230 Description: Early reply size mismatch, MGC loses connection
231 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
232 the connect flags are properly negotiated.
237 Description: quotas are not honored with O_DIRECT
238 Details : all writes with the flag O_DIRECT will use grants which leads to
239 this problem. Now using OBD_BRW_SYNC to guard this.
243 Description: add quota statistics
244 Details : 1. sort out quota proc entries and proc code.
245 2. add quota statistics
247 Severity : enhancement
249 Description: enable quota support for HEAD.
253 Description: Properly propagate oinfo flags from lov to osc for statfs
254 Details : restore missing copy oi_flags to lov requests.
256 Severity : enhancement
258 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
259 : l_getgroups(8), lst(8), routerstat(8)
260 Details : included man pages for llobdstat(8), llstat(8),
261 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
263 Severity : enhancement
265 Description: configure's --enable-quota should check the
266 : kernel .config for CONFIG_QUOTA
267 Details : configure is terminated if --enable-quota is passed but
268 : no quota support is in kernel
272 Description: Remove portals compatibility
273 Details : Remove portals compatibility, not interoperable with releases
278 Description: Resolve device initialization race
279 Details : Prevent proc handler from accessing devices added to the
280 obd_devs array but yet be intialized.
282 Severity : enhancement
284 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
286 Severity : enhancement
288 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
293 Description: Using wrong pointer in osc_brw_prep_request
294 Details : Access to array[-1] can produce panic if kernel compiled with
295 CONFIG_PAGE_ALLOC enabled
299 Description: Client runs out of low memory
300 Details : Consider only lowmem when counting initial number of llap pages
304 Description: *optional* service tags registration
305 Details : if the "service tags" package is installed on a Lustre node
306 When the filesystem is mounted, a local-node service tag will
307 be created. See http://inventory.sun.com/ for more information
308 about the Service Tags asset management system.
312 Description: Kernel BUG tries to release flock
313 Details : Lustre does not destroy flock lock before last reference goes
314 away. So always drop flock locks when client is evicted and
315 perform unlock regardless of successfulness of speaking to MDS.
319 Description: add refcount protection for osc callbacks, avoid panic on shutdown
323 Description: sanity test 65a fails if stripecount of -1 is set
324 Details : handle -1 striping on filesystem in ll_dirstripe_verify
329 Description: ASSERTION(CheckWriteback(page,cmd)) failed
330 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
333 Severity : enhancement
335 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
339 Description: do not process already freed flock
340 Details : flock can possibly be freed by another thread before it reaches
341 to ldlm_flock_completion_ast.
345 Description: LBUG during stress test
346 Details : Need properly lock accesses the flock deadlock detection list.
350 Description: oops in page fault handler
351 Details : kernel page fault handler can return two special 'pages' in error
352 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
356 Description: timeout with invalidate import.
357 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
358 should be handled by ptlrpcd. This produce long age waiting and
359 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
361 Severity : enhancement
363 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
367 Description: enable MGS and MDT services start separately
368 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
369 a co-located MGS without starting the MGS, which is a complement
370 to 'nosvc' mount option.
374 Frequency : after recovery
375 Description: precreate to many object's after del orphan.
376 Details : del orphan st in oscc last_id == next_id and this triger growing
377 count of precreated objects. Set flag LOW to skip increase count
378 of precreated objects.
382 Frequency : rare, on clear nid stats
383 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
384 Details : when clean nid stats sometimes try destroy live entry,
385 and this produce panic in free.
389 Description: Stack overflow during MDS log replay
390 ease stack pressure by using a thread dealing llog_process.
394 Description: wait until IO finished before start new when do lock cancel.
395 Details : VM protocol want old IO finished before start new, in this case
396 need wait until PG_writeback is cleared until check dirty flag and
397 call writepages in lock cancel callback.
399 Severity : enhancement
401 Description: using special macro for print time and cleanup in includes.
405 Description: mds_mfd_close() ASSERTION(rc == 0)
406 Details : In mds_mfd_close(), we need protect inode's writecount change
407 within its orphan write semaphore to prevent possible races.
411 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
415 Frequency : rare, on shutdown ost
416 Description: don't hit live lock with umount ost.
417 Details : shrink_dcache_parent can be in long loop with destroy dentries,
418 use shrink_dcache_sb instead.
422 Description: don't panic with use echo client
423 Details : echo client pass NULL as client nid pointer and this produce null
428 Description: fix build on ppc32
429 Details : compile code with -m64 flag produce wrong object file for ppc32.
433 Description: add message levels for liblustreapi
437 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
438 Details : Move the 'good_osts' check before the 'total_bavail' check. This
439 will result in an -EAGAIN and in the exit call path we call
440 alloc_rr() which will with increasing aggressiveness attempt to
441 aquire precreated objects on the minimum number of required OSCs.
445 Description: Use old size assignment to avoid deadlock
446 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
447 the scheduling while holding a spinlock. We do not need locking
448 for size in ll_update_inode() because size is only updated from
449 the MDS for directories or files without objects, so there is no
450 other place to do the update, and concurrent access to such inodes
451 are protected by the inode lock.
455 Description: resolve "_IOWR redefined" build error on SLES10
459 Description: dump the memory debugging after all modules are unloaded to
460 suppress false negative in conf_sanity test 39
462 Severity : enhancement
464 Description: build kernel-ib packages for OFED 1.3 in our release cycle
469 Description: fix SLES kernel versioning
470 Details : the kernel version for our SLES 10 kernel did not include a "-"
471 before the "smp" at the end. while this was not a problem in
472 general, it did mean that software trying to use the kernel
473 version to try to detect a vendor specific kernel would fail.
474 this was most evident by the OFED build scripts.
478 Description: Don't update lov_desc members until making sure they are valid
479 Details : When updating lov_desc members via proc fs, need fix their
480 validities before doing the real update.
484 Description: don't put request into delay list while invalidate in flight.
485 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
486 import in flight. this produce timeout for invalidate and sometimes
487 can cause stale data.
491 Frequency : on ppc only
492 Description: not convert ost objects for directory because it's not exist.
493 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
497 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
499 Severity : enhancement
501 Description: Update to support 2.6.22.14 vanilla kernel.
505 Frequency : rare, on recovery
506 Description: read procfs can produce deadlock in some situation
507 Details : Holding lprocfs lock which send rpc can produce block for destroy
508 obd objects and this also block reconnect with -EALREADY. This
509 isn't fix all lprocfs bugs - but make it rare.
511 Severity : enhancement
513 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
516 Frequency : frequent on X2 node
518 Description: mdc_set_open_replay_data LBUG
519 Details : Set replay data for requests that are eligible for replay.
523 Description: lustre_mgs: operation 101 on unconnected MGS
524 Details : When MGC is disconnected from MGS long enough, MGS will evict the
525 MGC, and late on MGC cannot successfully connect to MGS and a lot
526 of the error messages complaining that MGS is not connected.
530 Frequency : on network error
531 Description: panic with double free request if network error
532 Details : mdc_finish_enqueue is finish request if any network error ocuring,
533 but it's true only for synchronus enqueue, for async enqueue
534 (via ptlrpcd) this incorrect and ptlrpcd want finish request
537 Severity : enhancement
539 Description: client-side metadata stat-ahead during readdir(directory readahead)
540 Details : perform client-side metadata stat-ahead when the client detects
541 readdir and sequential stat of dir entries therein
544 Frequency : on start mds
546 Description: Implement get_info(last_id) in obdfilter.
549 Frequency : occasional
551 Description: Correctly check stale fid, not start epoch if ost not support SOM
552 Details : open with flag O_CREATE need set old fid in op_fid3 because
553 op_fid2 was overwritten with new generated fid, but mds can answer
554 with one of these two fids and both is not stale. Setattr
555 incorrectly started an epoch and assume will be called
556 done_writing, but without SOM done_writing ever being called.
559 Frequency : rare, depends on device drivers and load
561 Description: MDS or OSS nodes crash due to stack overflow
562 Details : Code changes in 1.8.0 increased the stack usage of some functions.
563 In some cases, in conjunction with device drivers that use a lot
564 of stack the MDS (or possibly OSS) service threads could overflow
565 the stack. One change which was identified to consume additional
566 stack has been reworked to avoid the extra stack usage.
569 Frequency : occasional
571 Description: Do not fail import if osc_interpret_create gets -EAGAIN
572 Details : If osc_interpret_create got -EAGAIN it immediately exits and
573 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
574 oscc_has_objects and see OSC has no objests and call
575 oscc_internal_create to resend create request.
577 Severity : enhancement
579 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
581 Severity : enhancement
583 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
586 Frequency : very rare
588 Description: Fix replay if there is an un-replied request and open
589 Details : In some cases, older replay request will revert the
590 mcd->mcd_last_xid on MDS which is used to record the client's
593 Severity : enhancement
595 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
597 Severity : enhancement
599 Description: Add rhel5 support to HEAD.
601 Serverity : enhancement
603 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
608 Description: Don't allow skipping OSTs if index has been specified.
609 Details : Don't allow skipping OSTs if index has been specified, make
610 locking in internal create lots better.
614 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
615 Details : access to bitfield in structure is always rounded to long
616 and this produce problem with not atomic change any bit.
620 Description: Lustre make rpms failed.
621 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
622 builds Lustre rpms from tarball.
625 Frequency : rare on shutdown ost
627 Description: If llog cancel was not send before clean_exports phase, this can
628 produce deadlock in llog code.
629 Details : If llog thread has last reference to obd and call class_import_put
630 this produce deadlock because llog_cleanup_commit_master wait when
631 last llog_commit_thread exited, but this never success because was
632 called from llog_commit_thread.
636 Description: allow userland application know is lost one of stripes.
637 Details : fill lvb_blocks with error code on ost and return it to
638 application if error flag found.
642 Description: NULL lov_tgts causing MDS oops
643 Details : more safe checks for NULL lov_tgts for avoid oops.
645 Severity : enhancement
647 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
651 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
652 Details : If an OST is down the MDS will hang indefinitely in
653 obd_statfs_rqset() waiting for the statfs data. While for
654 MDS QOS usage of statfs, it should not stuck in waiting.
656 Severity : enhancement
658 Description: remote_acl support
659 Details : Support ACL-based permission check for remote user.
660 Support setfacl/getfacl for remote user with the utils
661 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
662 the system "{s,g}etfacl" utils.
664 Severity : enhancement
666 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
668 Severity : enhancement
670 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
674 Description: make mgs_setparam() handle fsname containing dash
675 Details : fsname containing a dash does not work with lctl conf_param
677 Severity : enhancement
679 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
681 Severity : enhancement
683 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
687 Frequency : rare, at shutdown
688 Description: access already free / zero obd_namespace.
689 Details : if client_disconnect_export was called without force flag set,
690 and exist connect request in flight, this can produce access to
691 NULL pointer (or already free pointer) when connect_interpret
692 store ocd flags in obd_namespace.
696 Frequency : only at startup
697 Description: not alloc memory with spinlock held.
698 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
699 if any spinlock held.
701 Severity : enhancement
703 Description: make lustre randomly fail allocating memory
704 Details : Make lustre randomly failed allocating memory for testing purpose.
706 Severity : enhancement
708 Description: lost problems with lov objid file
709 Details : Fixes some scability and access to not inited memory problems
710 in work with lov objdid file.
715 Description: lfs find does not continue on file error
716 Details : Continue other files processing when a file/dir is absent.
720 Description: Inconsistent usage of lustre_pack_reply()
721 Details : Standardize the usage of lustre_pack_reply() such that it
722 always generate a CERROR on failure.
725 Frequency : occasional
727 Description: MDS hang or stay in waiting lock
728 Details : If client receive lock with CBPENDING flag ldlm need send lock
729 cancel as separate rpc, to avoid situation when cancel request
730 can't processed due all i/o threads stay in wait lock.
734 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
738 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
742 Description: Update to SLES9 kernel-2.6.5-7.287.3.
746 Description: Update to RHEL4 latest kernel.
748 Severity : enhancement
750 Description: Build SLES10 patchless client fails
751 Details : The configure was broken by run ./configure with
752 --with-linux-obj=.... argument for patchless client. When the
753 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
754 can't search header adequately. Use absolute path such as
755 -I($LINUX)/include instead.
759 Description: interrupt oig_wait produce painc on resend.
760 Details : brw_redo_request can be used for resend requests from ptlrpcd and
761 private set, and this produce situation when rq_ptlrpcd_data not
762 copyed to new allocated request and triggered LBUG on assert
763 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
764 ptlrpcd set if request is changed and can be safety changed to use
769 Description: LASSERT_{REQ,REP}SWAB macros are buggy
770 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
771 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
772 these macros and replace them with their difinitions instead.
776 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
777 Details : Update which_patch & target file for SLES10 latest kernel.
780 Description: Remove client patches from SLES 10 kernel.
781 Details : This causes SLES 10 clients to behave as patchless clients
782 even on a Lustre-patched (server) kernel.
784 Severity : enhancement
786 Description: self-adjustable client's lru lists
787 Details : use adaptive algorithm for managing client cached locks lru
788 lists according to current server load, other client's work
789 pattern, memory activities, etc. Both, server and client
790 side namespaces provide number of proc tunables for controlling
793 Severity : enhancement
795 Description: light-weight GSS support
796 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
797 and reduce performance overhead.
799 Severity : enhancement
801 Description: Linux keyring support
802 Details : Support using service of Linux keyring for Lustre GSS internal
803 context refresh/cache mechanism.
807 Description: Fix errors in lfs documentation
808 Details : Fixes man pages
812 Description: don't use GFP_* in generic Lustre code.
813 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
817 Description: obdclass is limited by single OBD_ALLOC(idarray)
818 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
822 Description: warnings with build patchless client with vanila 2.6.19 and up
823 Details : change old ctl_table style and replace ctl_table/ctl_table_header
824 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
828 Frequency : Only for SLES
829 Description: sanity_quota fail test_1
830 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
831 file, which makes TSTID[2] inunique.
835 Frequency : Only for SLES9
836 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
837 Details : Update target/ChangeLog/which_patch .
841 Frequency : Only for test_10 in sanity-quota.sh
842 Description: a bug in quota test code
843 Details : checking a test flag in a wrong place
847 Frequency : Only for test_13 in sanity-quota.sh
848 Description: a bug in quota test script
849 Details : shouldn't check overfull allocated quota
853 Description: jbd statistics
854 Details : Port older jbd statistics patch for sles10
856 Severity : enhancement
858 Description: Add printing inode info into message about error in writepage.
863 Description: reply_lock_interpret crash due to race with it and lock cancel.
864 Details : Do not replay locks that are being cancelled. Do not reference
865 locks by their address during replay, just by their handle.
869 Frequency : When flocks are used.
870 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
871 Details : Flock locks might destroy just granted lock if it could be merged
872 with another existing flock, this is done in completion handler,
873 so teach ldlm_cli_enquque_fini that this is a valid case for
879 Description: Oops in read and write path when failing to allocate lock.
880 Details : Check if lock allocation failed and return error back.
884 Description: lstripe command fails for valid OST index
885 Details : The stripe offset is compared to lov->desc.ld_tgt_count
886 instead of lov->desc.ld_active_tgt_count.
890 Description: sanity.sh failed test 103
891 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
896 Description: df doesn't work properly if diskfs blocksize != 4K
897 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
901 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
902 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
904 Severity : enhancement
906 Description: Adaptive timeouts
907 Details : RPC timeouts adapt to changing server load and network
908 conditions to reduce resend attempts and improve recovery time.
912 Description: llapi_file_create() does not allow some changes
913 Details : add llapi_file_open() that allows specifying the mode and
914 open flags, and also returns an open file handle.
918 Description: merge and cleanup kernel patches.
919 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
923 Description: Add journal checksum support.(Kernel part)
924 Details : The journal checksum feature adds two new flags i.e
925 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
926 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
927 indicates that the commit block contains the checksum for
928 the blocks described by the descriptor blocks. Now commit
929 record can be sent to disk without waiting for descriptor
930 blocks to be written to disk. This behavior is controlled
931 using JBD2_FEATURE_ASYNC_COMMIT flag.
935 Description: OSS needs mutliple precreate threads
936 Details : Add ability to start more than one create thread per OSS.
940 Description: Sanity.sh test_65k failed: setstripe should have succeeded
941 Details : execute lfs setstripe on client
945 Description: mds_obd_create error creating tmp object
946 Details : When the user sets quota on root, llog will be affected and can't
947 create files and write files.
950 Frequency : Always on ia64 patchless client, and possibly others.
952 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
953 Details : This allows the patchless client to be loaded on architectures
958 Description: RedHat Update kernel for RHEL5
959 Details : Add the kernel config file for RHEL5.
963 Description: RedHat Update kernel for RHEL5
964 Details : Modify the kernel config file more closer RHEL5.
968 Description: Build failure against Centos5 (RHEL5)
969 Details : Define PAGE_SIZE when it isn't present.
973 Description: test_7 on sanity-quota.sh can't be used on separate machines
974 Details : add facet to handle it
978 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
979 Details : with NFS, the anon dentry's parent was set to itself in
980 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
981 corresponding dentry other than use rec->ur_name.
983 Severity : enhancement
985 Description: lfs setstripe enhancement
986 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
990 Description: enable data checksumming by default
991 Details : enable checksum by default, allow --disable-checksum
992 configure option and "-o nochecksum" mount option. Checksums
993 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
997 Description: lustre support for RHEL5
998 Details : Add support for RHEL5.
1002 Description: Client eviction due to failover config
1003 Details : after a connection loss, the lustre client should attempt to
1004 reconnect to the last active server first before trying the
1005 other potential connections.
1009 Description: when mds and osts use different quota unit(32bit and 64bit),
1010 quota will be released repeatly.
1011 Details : void sending multiple quota reqs to mds, which will keep the
1012 status between the reqs.
1016 Description: osts not allocated evenly to files
1017 Details : change the condition to increase offset_idx
1020 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1022 Description: Data corruption for OSTs that are formatted larger than 2TB
1024 Details : When generating the bio request for lustre file writes the
1025 sector number would overflow a temporary variable before being
1026 used for the IO. The data reads correctly from Lustre (which
1027 will overflow in a similar manner) but other file data or
1028 filesystem metadata may be corrupted in some cases.
1032 Description: Tune the kernel for good SCSI performance.
1033 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1034 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1039 Description: rewrite ext2-derived code in obdclass/uuid.c
1040 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1041 from scratch preserving functionality.
1045 Description: "lfs find -obd UUID" prints directories
1046 Details : "lfs find -obd UUID" will return all directory names instead
1047 of just file names. It is incorrect because the directories
1048 do not reside on the OSTs.
1052 Description: use i_size_read and i_size_write in 2.6 port
1053 Details : replace inode->i_size access with i_size_read/write()
1056 Frequency : when removing large files
1058 Description: scheduling issue during removal of large Lustre files
1059 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1060 It causes scheduling issues when removing large files (17TB in the
1064 Frequency : only with liblustre clients on XT3
1066 Description: evictions taking too long
1067 Details : allow llrd to evict clients directly on OSTs
1070 Frequency : only on ppc
1072 Description: /proc/fs/lustre/devices broken on ppc
1073 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1074 We should make sure the type of 'index' is loff_t and then cast
1075 explicitly as needed below. Do not assign an explicitly cast
1079 Frequency : only for rhel5
1081 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1082 Details : Modify the target file & which_kernel.
1084 Severity : enhancement
1086 Description: omit set fsid for export NFS
1087 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1088 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1092 Frequency : Always, for kernels after 2.6.16
1093 Description: Fix warning idr_remove called for id=.. which is not allocated.
1094 Details : Last kernels save old s_dev before kill super and not allow
1095 to restore from callback - restore it before call kill_anon_super.
1099 Description: buffer overruns could theoretically occur
1100 Details : llapi_semantic_traverse() modifies the "path" argument by
1101 appending values to the end of the origin string, and a
1102 overrun may occur. Adding buffer overrun check in liblustreapi.
1106 Description: Fix error on 'ls .' at the top of the Lustre mount.
1107 Details : Don't revalidate dentry if it is a root dentry.
1111 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1112 Details : Modify vm-tunables-rhel4.patch.
1116 Description: Kernel config for 2.6.18-vanilla.
1117 Details : Modify targets/2.6-vanilla.target.in.
1118 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1119 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1120 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1121 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1125 Description: improve handling recoverable errors
1126 Details : if request processig with error which can be recoverable on server
1127 request should be resend, otherwise page released from cache and
1132 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1133 Details : Modify target file & which_patch.
1134 A flaw was found in the IA32 system call emulation provided
1135 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1136 value could be stored in the %RAX register, which could trigger an
1137 out-of-bounds system call table access. An untrusted local user
1138 could exploit this flaw to run code in the kernel
1139 (ie a root privilege escalation). (CVE-2007-4573).
1143 Description: O_DIRECT bypasses client statistics.
1144 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1145 read_ahead_stats were updated. copied the stats section from
1146 osc_send_oap_rpc() into async_internal().
1150 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1154 Description: change order of libsysio includes
1155 Details : '#include sysio.h' should always come before '#include xtio.h'
1159 Description: handle "serious error: objid * already exists" more gracefully
1160 Details : If LAST_ID value on disk is smaller than the objects existing in
1161 the O/0/d* directories, it indicates disk corruption and causes an
1162 LBUG(). If the object is 0-length, then we should use the existing
1163 object. This will help to avoid a full fsck in most cases.
1165 Severity : enhancement
1167 Description: adapt the lustre_config script to support the upgrade case
1168 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1169 server targets to 1.6 in parallel.
1173 Description: To avoid grant space > avaible space when the disk is almost
1174 full. Without this patch you might see the error "grant XXXX >
1175 available" or some LBUG about grant, when the disk is almost
1177 Details : In filter_check_grant, for non_grant cache write, we should
1178 check the left space by if (*left > ungranted + bytes), instead
1179 of (*left > ungranted), because only we are sure the left space
1180 is enough for another "bytes", then the ungrant space should be
1181 increase. In client, we should update cl_avail_grant only there
1182 is OBD_MD_FLGRANT in the reply.
1186 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1187 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1188 described in CVE-2007-4573. This problem could allow untrusted
1189 local users to gain root access.
1192 Frequency : when using O_DIRECT and quotas
1194 Description: Incorrect file ownership on O_DIRECT output files
1195 Details : block usage reported by 'lfs quota' does not take into account
1196 files that have been written with O_DIRECT.
1201 Description: touch file failed when fs is not full
1202 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1203 otherwise we can get ENOSP while fs is not full.
1207 Description: parallel lock callbacks
1208 Details : Instead of sending blocking and completion callbacks as separated
1209 requests, adding them to a set and sending in parallel.
1212 Frequency : only for Cray XT3
1213 Bugzilla : 12829/13455
1214 Description: Changing primary group doesn't change the group lustre assigns to
1216 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1217 overridden with the primary group provided by the group upcall,
1218 whereas the supplied fsgid can be trusted if it is in the list of
1219 supplementary groups returned by the group upcall.
1221 Severity : enhancement
1223 Description: Allow masking D_WARNING, D_ERROR messages from console
1224 Details : Console messages can now be disabled via lnet.printk.
1228 Description: User code with malformed file open parameter crashes client node
1229 Details : Before packing join_file req, all the related reference should be
1230 checked carefully in case some malformed flags cause fake
1231 join_file req on client.
1235 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1236 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1237 mode is equal to the granted mode. However, it should also check
1238 that LDLM_FL_CP_REQD is not set.
1242 Description: Heavy nfs access might result in deadlocks
1243 Details : After ELC code landed, it is now improper to enqueue any mds
1244 locks under och_sem, because enqueue might want to decide to
1245 cancel open locks for same inode we are holding och_sem for.
1249 Description: Client eviction while running blogbench
1250 Details : A lot of unlink operations with concurrent I/O can lead to a
1251 deadlock causing evictions. To address the problem, the number of
1252 oustanding OST_DESTROY requests is now throttled to
1253 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1254 callbacks are processed in priority.
1258 Description: enable ACLs on MDS by default
1259 Details : ACLs must be enabled on MDS by default.
1262 Frequency : PPC/PPC64 only
1264 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1265 Details : fix duplicated definitions between asm-ppc64/types.h and
1266 lustre_types.h on PPC.
1269 Frequency : PPC/PPC64 only
1271 Description: asm-ppc/segment.h does not exist
1272 Details : fix compile issue on PPC.
1276 Description: better handle error messages in extents code
1279 Frequency : RHEL4 only
1281 Description: mkfs is very slow on IA64/RHEL4
1282 Details : A performance regression has been discovered in the MPT Fusion
1283 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1284 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1285 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1287 Severity : enhancement
1289 Description: SNMP support enhancement
1290 Details : Adding total number of sampled request for an MDS node in snmp
1293 Severity : enhancement
1295 Description: Optimize ldlm waiting list processing for PR extent locks
1296 Details : When processing waiting list for read extent lock and meeting read
1297 lock that is same or wider to it that is not contended, skip
1298 processing rest of the list and immediatelly return current
1299 status of conflictness, since we are guaranteed there are no
1300 conflicting locks in the rest of the list.
1304 Description: Time out and refuse to reconnect
1305 Details : When the failover node is the primary node, it is possible
1306 to have two identical connections in imp_conn_list. We must
1307 compare not conn's pointers but NIDs, otherwise we can defeat
1308 connection throttling.
1312 Description: port llog fixes from b1_6 into HEAD
1313 Details : Port llog reference couting and some llog cleanups from b1_6
1314 (bug 10800) into HEAD, for protect from panic and access to already
1315 free llog structures.
1319 Description: Detect stride IO mode in read-ahead
1320 Details : When a client does stride read, read-ahead should detect that and
1321 read-ahead pages according to the detected stride pattern.
1325 Description: data checksumming impacts single node performance
1326 Details : add support for several checksum algorithm. Currently, only CRC32
1327 and Adler-32 are supported. The checksum type can be changed on
1328 the fly via /proc/fs/lustre/osc/*/checksum_type.
1332 Description: use adler32 for page checksums
1333 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1338 Description: build for x2 fails
1339 Details : fix compile issue on Cray systems.
1343 Description: Properly match for duplicate locks
1344 Details : Due to different lock order from skiplists code, we need to
1345 traverse entire list for now
1348 Frequency : only on PPC/SLES10
1350 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1351 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1352 be defined. Add a hack in mkfs_lustre.c to work around this compile
1357 Description: LASSERT on MDS when client holding flock lock dies
1358 Details : ldlm pool logic depends on number of granted locks equal to
1359 number of released locks which is not true for flock locks, so
1360 just exclude such locks from consideration.
1364 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1365 Details : Use fsfilt_sync() for both the whole filesystem sync and
1366 individual file sync to eliminate dangerous inode locking
1367 with I_LOCK that can lead to a deadlock.
1371 Description: performance in 1.6.3
1372 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1373 sure that 1MB requests are not fragmented by the block layer.
1375 Severity : enhancement
1377 Description: organize the server-side client stats on per-nid basis
1378 Details : Change the structure of stats under obdfilter and mds to
1386 The "uuid"s file would list the uuids of _active_ exports.
1387 And the clear entry is to clear all stats and stale nids.
1389 Severity : enhancement
1391 Description: eliminate client locks in face of contention
1392 Details : file contention detection and lockless i/o implementation
1393 for contended files.
1397 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1400 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1402 Description: MDS may lose file striping (and hence file data) in some cases
1403 Details : If there are additional extended attributes stored on the MDS,
1404 in particular ACLs, SELinux, or user attributes (if user_xattr
1405 is specified for the client mount options) then there is a risk
1406 of attribute loss. Additionally, the Lustre file striping
1407 needs to be larger than default (e.g. striped over all OSTs),
1408 and an additional attribute must be stored initially in the
1409 inode and then increase in size enough to be moved to the
1410 external attribute block (e.g. ACL growing in size) for file
1415 Description: skiplist implementation simplification
1416 Details : skiplists are used to group compatible locks on granted list
1417 that was implemented as tracking first and last lock of each
1418 lock group the patch changes that to using doubly linked lists
1422 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1423 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1424 because LASSERT on both the data supplied by a client, and the
1425 data on disk is dangerous and incorrect.
1427 Severity : enhancement
1429 Description: Slow truncate/writes to huge files at high offsets.
1430 Details : Directly associate cached pages to lock that protect those pages,
1431 this allows us to quickly find what pages to write and remove
1432 once lock callback is received.
1436 Description: more ldlm soft lockups
1437 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1438 starve other threads from the resource lock for a long time in
1439 case of long waiting queue, so change the debug level from
1440 D_OTHER to the less frequently used D_INFO.
1442 Severity : enhancement
1444 Description: add -gid, -group, -uid, -user options to lfs find
1448 Description: Hung threads in invalidate_inode_pages2_range
1449 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1450 once one is completed. As a result, some RPCs are stuck in the
1451 queue and are never sent.
1455 Description: filter threads hungs on waiting journal commit
1456 Details : Cleanup filter group llog code, then only filter group llog will
1457 be only created in the MDS/OST syncing process.
1461 Description: Procfs and llog threads access destoryed import sometimes.
1462 Details : Sync the import destoryed process with procfs and llog threads by
1463 the import refcount and semaphore.
1465 Severity : enhancement
1467 Description: openlock cache of b1_6 port to HEAD
1472 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1473 Details : If insertion of an extent fails, then discard the inode
1474 preallocation and free data blocks else it can lead to duplicate
1479 Description: don't always update ctime in ext3_xattr_set_handle()
1480 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1481 In some cases the ctime should not be updated, for example for
1482 2.0->1.8 compatibility it is necessary to delete an xattr and it
1483 should not update the ctime.
1487 Bugzilla : 15713/16362
1488 Description: Assertion in iopen_connect_dentry in 1.6.3
1489 Details : looking up an inode via iopen with the wrong generation number can
1490 populate the dcache with a disconneced dentry while the inode
1491 number is in the process of being reallocated. This causes an
1492 assertion failure in iopen since the inode's dentry list contains
1493 both a connected and disconnected dentry.
1497 Description: assertion failure in ldlm_handle2lock()
1498 Details : fix a race between class_handle_unhash() and class_handle2object()
1499 introduced in lustre 1.6.5 by bug 13622.
1504 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1505 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1510 Description: oops in sd_iostats_seq_show()
1511 Details : unloading/reloading the scsi low level driver triggers a kernel
1512 bug when trying to access the sd iostat file.
1517 Description: Kernel panics during QLogic driver reload
1518 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1519 patch, causing memory corruption.
1524 Description: journal_dev option does not work in b1_6
1525 Details : pass mount option during pre-mount.
1527 Severity : enhancement
1529 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1530 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1531 extent information of a file. It can be used to map logical blocks
1532 in a file to physical blocks in the block device.
1536 Description: LDLM soft lockups - improvement
1537 Details : It is be possible to send the lock handle along with each read
1538 or write request because the client is already doing a lock match
1539 itself so there isn't any reason the OST should have to re-do that
1543 Frequency : only on Cray X2
1545 Description: X2 build failures
1546 Details : fix build failures on Cray X2.
1550 Description: xid & resent requests
1551 Details : Initialize RPC XID from clock at startup (randomly if clock is
1554 Severity : enhancement
1556 Description: Add lustre_start utility to start or stop multiple Lustre servers
1561 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1562 Details : In case of memory pressure, list_del() can be called twice on
1563 req->rq_history_list, causing a kernel oops.
1567 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1568 Details : fix stack overflow in the distributed lock manager by defering
1569 export eviction after a failed AST to the elt thread instead of
1570 handling it in the dlm interpret routine.
1574 Description: Convert some comments to new format.
1575 Details : Update documenting comments to match doxygen conventions.
1579 Description: Grammar fixes.
1580 Details : A couple of trivial spelling fixes.
1584 Description: OSD_COUNTERS-mandatory
1585 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1586 to simplify the code.
1590 Description: simplify cmm_device freeing logic.
1591 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1595 Description: Add lockdep support to dt_object_operations locking interface.
1596 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1597 parameter indicating lock ordering. Update mdd code to use new
1602 Description: Introduce failloc constants for lockless IO tests.
1603 Details : Add two new failloc constants to test lockless IO. Only one of
1604 them in implemented---another is checked in yet to be landed
1609 Description: Add lockdep support for inode mutex.
1610 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1611 in the situations where Lustre has to lock more than one inode
1616 Description: Add optional invariants checking support.
1617 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1618 configure switch. This macro is to be used for consistency and
1619 sanity checks that are too expensive to be left in `production'
1624 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1625 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1630 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1631 Details : Use spin_lock_nested() in (the only) situation where more than
1632 one ldlm_lock is locked simultaneously. Also, fix possible
1633 dead-lock in ldlm_lock_change_resource() by enforcing particular
1638 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1639 Details : Instead of specifying each ldlm_lock call-back through separate
1640 parameter, wrap them into struct ldlm_callback_suite.
1644 Description: Kill join_lru obd method and its callers.
1645 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1646 in memory---a requirement implemented through ->o_join_lru() obd
1647 method in HEAD. Remove this method and its users.
1651 Description: Add asynchronous ldlm ENQUEUE completion handler.
1652 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1653 case that stock ldlm_completion_ast() cannot handle as it waits
1654 until lock is granted. Introduce new ldlm_completion_ast_async()
1655 for this. Also comment ldlm_completion_ast().
1659 Description: ldlm_error <-> errno conversion.
1660 Details : Add functions to map (rather arbitrary) between LDLM error codes
1661 and standard errno values. CLIO needs this to prevent LDLM specific
1662 constants from escaping ldlm and osc.
1666 Description: Kill unused ldlm_handle2lock_ns() function.
1667 Details : Kill unused ldlm_handle2lock_ns() function.
1671 Description: Add lu_ref support to ldlm_lock
1672 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1673 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1674 and ldlm_resource. LDLM interface has to be changed, because code that
1675 releases a reference on a lock, has to "know" what reference this is.
1676 In the most frequent case
1678 lock = ldlm_handle2lock(handle);
1680 LDLM_LOCK_PUT(lock);
1682 no changes are required. When any other reference (received _not_ from
1683 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1684 instead of LDLM_LOCK_PUT().
1686 Arguably, changes are pervasive, and interface requires some discipline
1687 for proper use. On the other hand, it was very instrumental in finding
1688 a few leaked lock references.
1692 Description: Add ldlm_lock_addref_try().
1693 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1694 attempts to addref a lock that might be being canceled
1699 Description: Add ldlm_weigh_callback().
1700 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1701 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1702 hard-coded `number of pages' logic.
1706 Description: Add lockdep annotations to llog code.
1707 Details : Use appropriately tagged _nested() locking calls in the places
1708 where llog takes more than one ->lgh_lock lock.
1712 Description: Add loi_kms_set().
1713 Details : Wrap kms updates into a helper function.
1717 Description: Constify instances of struct lsm_operations.
1718 Details : Constify instances of struct lsm_operations.
1722 Description: lu_conf support.
1723 Details : On a server, a file system object is uniquely identified
1724 by a fid, which is sufficient to locate and load all object
1725 state (inode). On a client, on the other hand, more data are
1726 necessary instantiate an object. Change lu_object_find() and
1727 friends to take additional `lu_conf' argument describing object.
1728 Typically this includes layout information.
1732 Description: lu_context fixes.
1733 Details : Introduce new lu_context functions that are needed on the client
1734 side, where some system threads (ptlrpcd) are shared by multiple
1735 modules, and so cannot be stopped during module shutdown.
1739 Description: Add start and stop methods to lu_device_type_operations.
1740 Details : Introduce two new methods in lu_device_type_operations, that are
1741 invoked when first instance of a given type is created and last one
1742 is destroyed respectively. This is need by CLIO.
1746 Description: Add lu_ref support to struct lu_device.
1747 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1748 track leaked references.
1752 Description: Introduce lu_kmem_descr.
1753 Details : lu_kmem_descr and its companion interface allow to create
1754 and destroy a number of kmem caches at once.
1758 Description: Fix lu_object finalization race.
1759 Details : Fix a race between lu_object_find() finding an object and its
1760 concurrent finalization. This race is (most likely) not possible
1761 on the server, but might happen on the client.
1765 Description: Introduce lu_ref interface.
1766 Details : lu_ref is a debugging module allowing to track references to
1767 a given object. It is quite cpu expensive, and has to be
1768 explicitly enabled with --enable-lu_ref. See usage description
1773 Description: Factor lu_site procfs stats into a separate function.
1774 Details : Separate lu_site stats printing code into a separate function
1775 that can be reused on a client.
1779 Description: Constify instances of struct {lu,dt,md}_device_operations.
1780 Details : Constify instances of struct {lu,dt,md}_device_operations.
1784 Description: Introduce struct md_site and move meta-data specific parts of
1785 struct lu_site here.
1786 Details : Move md-specific fields out of struct lu_site into special struct
1787 md_site, so that lu_site can be used on a client.
1791 Description: Kill mdd_lov_destroy().
1792 Details : Remove unused mdd code.
1796 Description: Add st_block checking to multistat.c.
1797 Details : Add st_block checking to multistat.c.
1801 Description: Add lu_ref support to struct obd_device.
1802 Details : Add lu_ref tracking to obd_device.
1806 Description: Kill obd_set_fail_loc().
1807 Details : Remove unused code.
1811 Description: Add special type for ptlrpc_request interpret functions.
1812 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1813 there. Actual usage will be in CLIO.
1817 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1818 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1819 doesn't work with lockdep.
1823 Description: Add rwv.c test program.
1824 Details : New testing program exercising readv(2) and writev(2) (Qian).
1828 Description: Add sendfile.c test program.
1829 Details : New testing program exercising sendfile(2) (Jay).
1833 Description: Ratelimit a message that can be very frequent.
1834 Details : Ratelimit a memory allocation failure message that can
1839 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1840 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1844 Description: Liblustre build fixes.
1845 Details : Liblustre build fixes.
1849 Description: libcfs: add cfs_{need,cond}_resched() interface.
1850 Details : libcfs: add cfs_{need,cond}_resched() definition and
1851 implementations for Linux, NT, and liblustre.
1853 Severity : enhancement
1855 Description: More exported tunables for mballoc
1856 Details : Add support for tunable preallocation window and new tunables for
1857 large/small requests
1861 Description: Detect corruption of block bitmap and checking for preallocations
1862 Details : Checks validity of on-disk block bitmap. Also it does better
1863 checking of number of applied preallocations. When corruption is
1864 found, it turns filesystem readonly to prevent further corruptions.
1868 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1869 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1870 window to 0. If it does hit cache miss, and read-pattern is still
1871 stride-io mode, does not reset the stride window, but also does not
1872 increase the stride window length in this case.
1876 Frequency : only for big-endian servers
1877 Description: Check if system is big-endian while mounting fs with extents feature
1878 Details : Mounting a filesystem with extents feature will fail on big-endian
1879 systems since ext3-based ldiskfs is not supported on big-endian
1880 systems. This can be over-riden with "bigendian_extents" mount option.
1882 Severity : enhancement
1884 Description: The root squash functionality
1885 Details : A security feature, which is to prevent users from being able
1886 to mount lustre on their desktop, run as root, and delete
1887 all of the files in the filesystem. The goal is accomplished by
1888 remapping user id (UID) and group id (GID) of the root user to
1889 a UID and GID specified by the system administartor via Lustre
1890 configuration management server (MGS). The functionality also
1891 allows to specify sets of clients for which the remapping does
1896 Description: Excessive recovery window
1897 Details : With AT enabled, the recovery window can be excessively long (6000+
1898 seconds). To address this problem, we no longer use
1899 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1900 timeout no longer depends on the service time, it is set to
1901 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1902 time via pb_service_time.
1906 Description: Watchdog triggered on MDS failover
1907 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
1908 the OSTs know that the MDS "UUID" can be reused for the same export
1909 from a different NID, so we do not need to wait for the export to be
1912 --------------------------------------------------------------------------------
1914 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1916 * Support for kernels:
1917 2.6.5-7.283 (SLES 9),
1918 2.6.9-55.EL (RHEL 4),
1919 2.6.16.46-0.14 (SLES 10),
1920 2.6.18.8 vanilla (kernel.org)
1921 * Client support for unpatched kernels:
1922 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1923 2.6.16 - 2.6.22 vanilla (kernel.org)
1924 * Due to recently discovered recovery problems, we do not recommend
1925 using patchless RHEL 4 clients with this or any earlier release.
1926 * Recommended e2fsprogs version: 1.39.cfs8
1927 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1928 * Starting with this release, the ldiskfs backing filesystem required
1929 by Lustre is now in its own package, lustre-ldiskfs. This package
1930 should be installed. It is versioned separately from Lustre and
1931 may be released separately in future.
1935 Description: block reactivating mgc import until all deactivates complete
1936 Details : Fix race when failing back MDT/MGS to itself (testing)
1939 Frequency : at statup only
1941 Description: mds_lov_synchronize race leads to various problems
1942 Details : simultaneous MDT->OST connections at startup can cause the
1943 sync to abort, leaving the OSC in a bad state.
1945 Severity : enhancement
1947 Description: add optional extra BUILD_VERSION info
1948 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1949 to override the lustre version.
1952 Frequency : 2.6.18 servers only
1954 Description: ll_kern_mount() doesn't release the module reference
1955 Details : The ldiskfs module reference count never drops down to 0
1956 because ll_kern_mount() doesn't release the module reference.
1961 Description: server LBUG when using old ost_num_threads parameter
1962 Details : Accept the old ost_num_threads parameter but warn that it
1963 is deprecated, and fix an off-by-one error that caused an LBUG.
1968 Description: Transient SCSI error results in persistent IO issue
1969 Details : iobuf->dr_error is not reinitialized to 0 between two
1973 Frequency : sometimes when underlying device returns I/O errors
1975 Description: OSTs not going read-only during write failures
1976 Details : OSTs are not remounted read-only when the journal commit threads
1977 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1978 instead of the ext3 wrappers.
1982 Description: poor connect scaling with increasing client count
1983 Details : Don't run filter_grant_sanity_check for more than 100 exports
1984 to improve scaling for large numbers of clients.
1987 Frequency : SLES10 only
1989 Description: sanity-quota.sh quotacheck failed: rc = -22
1990 Details : Quotas cannot be enabled on SLES10.
1993 Frequency : liblustre clients only
1995 Description: getdirentries does not give error when run on compute nodes
1996 Details : getdirentries does not fail when the size specified as an argument
1997 is too small to contain at least one entry
1999 Severity : enhancement
2001 Description: Add LNET router traceability for debug purposes
2002 Details : If a checksum failure occurs with a router as part of the
2003 IO path, the NID of the last router that forwarded the bulk data
2004 is printed so it can be identified.
2009 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2010 Details : Due to a race condition, liblustre clients were occasionally
2011 evicted incorrectly.
2013 Severity : enhancement
2015 Description: lfs setstripe use optional parameters instead of postional
2018 Severity : enhancement
2020 Description: Nanosecond timestamp support for ldiskfs
2021 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2022 resolution timestamps. There is not yet support for this at
2023 the Lustre filesystem level.
2026 Frequency : during server recovery
2028 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2029 Details : request with rq_no_resend flag not awake l_wait_event if they get
2033 Frequency : nfs export on patchless client
2035 Description: connectathon hang when test nfs export over patchless client
2036 Details : Disconnected dentry cannot be found with lookup, so we do not need
2037 to unhash it or make it invalid
2040 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2041 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2042 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2047 Description: open req refcounting wrong on reconnect
2048 Details : If reconnect happened between getting open reply from server and
2049 call to mdc_set_replay_data in ll_file_open, we will schedule
2050 replay for unreferenced request that we are about to free.
2051 Subsequent close will crash in variety of ways.
2052 Check that request is still eligible for replay in
2053 mdc_set_replay_data().
2058 Description: disable writes to filesystem when reading health_check file
2059 Details : the default for reading the health_check proc file has changed
2060 to NOT do a journal transaction and write to disk, because this
2061 can cause reads of the /proc file to hang and block HA state
2062 checking on a healthy but otherwise heavily loaded system. It
2063 is possible to return to the previous behaviour during configure
2064 with --enable-health-write.
2066 Severity : enhancement
2068 Description: 64-bit inode version
2069 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2070 to the inode. This will be required for version-based recovery.
2075 Description: MDS fails to start if a duplicate client export is detected
2076 Details : in some rare cases it was possible for a client to connect to
2077 an MDS multiple times. Upon recovery the MDS would detect this
2078 and fail during startup. Handle this more gracefully.
2080 Severity : enhancement
2082 Description: Add -o localflock option to simulate old noflock
2084 Details : This will achieve local-only flock/fcntl locks
2090 Description: log_commit_thread vs filter_destroy race leads to crash
2091 Details : Take import reference before releasing llog record semaphore
2096 Description: Wrong request locking in request set processing
2097 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2098 imp_delayed_list, in this place should be used imp_lock.
2101 Frequency : when reconnection
2103 Description: Grant Leak when osc reconnect to OST
2104 Details : When osc reconnect ost, OST(filter) should check whether it
2105 should grant more space to client by comparing fed_grant and
2106 cl_avail_grant, and return the granted space to client instead
2107 of "new granted" space, because client will call osc_init_grant
2108 to update the client grant space info.
2111 Frequency : when client reconnect to OST
2113 Description: Grant Leak when osc do resend and replay bulk write
2114 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2115 bulk write request, because the grant info while be sync between
2116 OSC and OST when reconnect, and we should ignore the grant info
2117 these of resend/replay write req.
2122 Description: Grant space more than avaiable left space sometimes.
2123 Details : When then OST is about to be full, if two bulk writing from
2124 different clients came to OST. Accord the avaliable space of the
2125 OST, the first req should be permitted, and the second one
2126 should be denied by ENOSPC. But if the seconde arrived before
2127 the first one is commited. The OST might wrongly permit second
2128 writing, which will cause grant space > avaiable space.
2131 Frequency : when client is evicted
2133 Description: Grant might be wrongly erased when osc is evicted by OST
2134 Details : when the import is evicted by server, it will fork another
2135 thread ptlrpc_invalidate_import_thread to invalidate the
2136 import, where the grant will be set to 0. While the original
2137 thread will update the grant it got when connecting. So if
2138 the former happened latter, the grant will be wrongly errased
2139 because of this race.
2144 Description: Checking Stale with correct fid
2145 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2146 to check whether it is stale, because sometimes, we want the
2147 enqueue happened anyway, and op_data.fid2 will not be initialized.
2149 Severity : enhancement
2151 Description: update patchless client
2152 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2155 Frequency : only with 2.4 kernel
2157 Description: random memory corruption
2158 Details : size of struct ll_inode_info is to big for union inode.u and this
2159 can be cause of random memory corruption.
2164 Description: Memory leak in recovery
2165 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2166 It should also check obd_fail before fsfilt_start, otherwise if
2167 fsfilt_start return -EROFS,(failover mds during mds recovery).
2168 then the req will return with repmsg->transno = 0 and rc = EROFS.
2169 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2170 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2171 be freed no matter whether fsfilt_commit success or not.
2174 Frequency : only with huge count clients
2176 Description: Prevents from taking the superblock lock in llap_from_page for
2178 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2179 for llap_from_page call in ll_removepage prevents from taking the
2180 superblock lock for a soon died page.
2185 Description: Not check open intent error before release open handle
2186 Details : in some rare cases, the open intent error is not checked before
2187 release open handle, which may cause
2188 ASSERTION(open_req->rq_transno != 0), because it tries to release
2189 the failed open handle.
2194 Description: Set cat log bitmap only after create log success.
2195 Details : in some rare cases, the cat log bitmap is set too early. and it
2196 should be set only after create log success.
2200 Description: Accessing a block bevice can re-enable I/O when Lustre is
2201 tearing down a device.
2202 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2207 Description: service threads may hog cpus when there are a lot of requests
2208 Details : Insert cond_resched to give other threads a chance to use some CPU
2213 Description: the cat log was not initialized in recovery
2214 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2215 unlink log on mds will not be initialized until mds post
2216 recovery. And also in mds post recovery, the unlink log will
2217 initialization will be done asynchronausly, so there will be race
2218 between add unlink log and unlink log initialization.
2222 Description: brw_stats were being printed incorrectly
2223 Details : brw_stats were being printed as log2 but all of them were not
2224 recorded as log2. Also remove some code duplication arising from
2225 filter_tally_{read,write}.
2229 Frequency : rare, only in recovery.
2230 Description: ASSERTION(req->rq_type != LI_POISON) failed
2231 Details : imp_lock should be held while iterating over imp_sending_list for
2232 prevent destroy request after get timeout in ptlrpc_queue_wait.
2236 Description: replay-single.sh test 52 fails
2237 Details : A lock's skiplist need to be cleanup when it being unlinked
2238 from its resource list.
2242 Description: Short directio read returns full requested size rather than
2244 Details : Direct I/O operations should return actual amount of bytes
2245 transferred rather than requested size.
2247 Severity : enhancement
2249 Description: metadata RPC reduction (e.g. for rm performance)
2250 Details : decrease the amount of synchronous RPC between clients and servers
2251 by canceling conflicing lock before the operation on the client
2252 and packing thier handles into the main operation RPC to server.
2254 Severity : enhancement
2256 Description: Async OSC create to avoid the blocking unnecessarily.
2257 Details : If a OST has no remain object, system will block on the creating
2258 when need to create a new object on this OST. Now, ways use
2259 pre-created objects when available, instead of blocking on an
2260 empty osc while others are not empty. If we must block, we block
2261 for the shortest possible period of time.
2265 Description: improve handling recoverable errors
2266 Details : if request processig with error which can be recoverable on server
2267 request should be resend, otherwise page released from cache and
2270 Severity : enhancement
2272 Description: refine locking for avoid write wrong info into lov_objid file
2273 Details : fix possible races with add new target and write/update data in
2276 --------------------------------------------------------------------------------
2278 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2283 Frequency : on some architectures
2285 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2286 Details : Uninitialized flags sometimes cause configuration commands to
2290 Frequency : patchless clients only
2292 Description: missing __iget() symbol export
2293 Details : The __iget() symbol export is missing. To avoid the need for
2294 this on patchless clients the deathrow inode reaper is turned
2295 off, and we depend on the VM to clean up old inodes. This
2296 dependency was during via the fix for bug 12181.
2298 --------------------------------------------------------------------------------
2300 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2302 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2303 INTEROPERATE with older versions automatically. In many cases a
2304 special upgrade step is needed. Please read the
2305 user documentation before upgrading any part of a live system.
2306 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2307 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2308 * WARNING: Lustre configuration and startup changes are required with
2309 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2313 Severity : enhancement
2315 Description: MountConf
2316 Details : Lustre configuration is now managed via mkfs and mount
2317 commands instead of lmc and lconf. New obd types (MGS, MGC)
2318 are added for dynamic configuration management. See
2319 https://mail.clusterfs.com/wikis/lustre/MountConf for
2322 Severity : enhancement
2324 Description: dynamic OST addition
2325 Details : OSTs can now be added to a live filesystem
2327 Severity : enhancement
2329 Description: startup order invariance
2330 Details : MDTs and OSTs can be started in any order. Clients only
2331 require the MDT to complete startup.
2333 Severity : enhancement
2335 Description: parallel, asynchronous orphan cleanup
2336 Details : orphan cleanup is now performed in separate threads for each
2337 OST, allowing parallel non-blocking operation.
2339 Severity : enhancement
2341 Description: optimized stripe assignment
2342 Details : stripe assignments are now made based on ost space available,
2343 ost previous usage, and OSS previous usage, in order to try
2344 to optimize storage space and networking resources.
2346 Severity : enhancement
2348 Description: Permanently set tunables
2349 Details : All writable /proc/fs/lustre tunables can now be permanently
2350 set on a per-server basis, at mkfs time or on a live system.
2352 Severity : enhancement
2354 Description: Lustre message v2
2355 Details : Add lustre message format v2.
2357 Severity : enhancement
2359 Description: client OST exclusion list
2360 Details : Clients can be started with a list of OSTs that should be
2361 declared "inactive" for known non-responsive OSTs.
2365 Description: SPEC SFS validation failure on NFS v2 over lustre.
2366 Details : Changes the blocksize for regular files to be 2x RPC size,
2367 and not depend on stripe size.
2369 Severity : enhancement
2371 Description: Multiple MD RPCs in flight.
2372 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2373 To avoid overly-overloading MDS, introduce a limit on number of
2374 MDS RPCs in flight for a single client and add /proc controls
2375 to adjust this limit.
2377 Severity : enhancement
2379 Description: client read/write statistics
2380 Details : Add client read/write call usage stats for performance
2381 analysis of user processes.
2382 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2383 file access. extents_stats shows chunk size distribution.
2384 extents_stats_per_process show chunk size distribution per
2387 Severity : enhancement
2389 Description: mds statistics
2390 Details : Add detailed mds operations statistics in
2391 /proc/fs/lustre/mds/*/stats.
2395 Description: Failure of copying files with lustre special EAs.
2396 Details : Client side always return success for setxattr call for lustre
2397 special xattr (currently only "trusted.lov").
2402 Description: Refcount LNET uuids
2403 Details : The global LNET uuid list grew linearly with every startup;
2404 refcount repeated list entries instead of always adding to
2407 Severity : enhancement
2409 Description: Dynamic service threads
2410 Details : Within a small range, start extra service threads
2411 automatically when the request queue builds up.
2414 Frequency : mixed-endian client/server environments
2416 Description: mixed-endian crashes
2417 Details : The new msg_v2 system had some failures in mixed-endian
2420 Severity : enhancement
2422 Description: Easy OST removal
2423 Details : OSTs can be permanently deactivated with e.g. 'lctl
2424 conf_param lustre-OST0001.osc.active=0'
2426 Severity : enhancement
2428 Description: MGS proc entries
2429 Details : Added basic proc entries for the MGS showing what filesystems
2432 Severity : enhancement
2434 Description: provide MGS failover
2435 Details : Added config lock reacquisition after MGS server failover.
2437 Severity : enhancement
2439 Description: add Linux 2.4 support
2440 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2444 Description: a large application tries to do I/O to the same resource and dies
2445 in the middle of it.
2446 Details : Check the req->rq_arrival time after the call to
2447 ost_brw_lock_get(), but before we do anything about
2448 processing it & sending the BULK transfer request. This
2449 should help move old stale pending locks off the queue as
2450 quickly as obd_timeout.
2453 Frequency : when an incorrect nid is specified during startup
2455 Description: ptlrpc connect to non-existant node causes kernel crash
2456 Details : LNET can't be re-entered from an event callback, which
2457 happened when we expire a message after the export has been
2458 cleaned up. Instead, hand the zombie cleanup off to another
2461 Severity : enhancement
2463 Description: plain/inodebits lock performance improvement
2464 Details : Grouping plain/inodebits in granted list by their request modes
2465 and bits policy, thus improving the performance of search through
2469 Frequency : only if OST filesystem is corrupted
2471 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2472 Details : for a short time RPCs with bulk IO are in the replay list,
2473 but replay of bulk IOs is unimplemented. If the OST filesystem
2474 is corrupted due to disk cache incoherency and then replay is
2475 started it is possible to trip an assertion. Avoid putting
2476 committed RPCs into the replay list at all to avoid this issue.
2479 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2481 Description: System hang on startup
2482 Details : This bug allowed the liblustre (e.g. catamount) client to
2483 return to the app before handling all startup RPCs. This
2484 could leave the node unresponsive to lustre network traffic
2485 and manifested as a server ptllnd timeout.
2487 Severity : enhancement
2489 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2490 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2491 boolean to control whether to print peer debug info when a
2492 client's RPC times out.
2495 Frequency : only for kernels with patches from Lustre below 1.4.3
2497 Description: Remove old rdonly API
2498 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2501 Frequency : only for devices with external journals
2503 Description: Set external device read-only also
2504 Details : During a commanded failover stop, we set the disk device
2505 read-only while the server shuts down. We now also set any
2506 external journal device read-only at the same time.
2509 Frequency : when upgrading from 1.4 while trying to change parameters
2511 Description: The wrong (new) MDC name was used when setting parameters for
2512 upgraded MDT's. Also allows changing of OSC (and MDC)
2513 parameters if --writeconf is specified at tunefs upgrade time.
2516 Frequency : when setting specific ost indicies
2518 Description: QOS code breaks on skipped indicies
2519 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2520 created with --index need not be sequential.
2522 Severity : enhancement
2524 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2525 Details : The uninit_groups feature works in conjunction with the kernel
2526 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2527 up the pass1 processing of e2fsck. This is a read-only feature
2528 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2529 mount filesystems that have had this feature enabled.
2531 Severity : enhancement
2533 Description: Improve multi-block allocation algorithm to avoid fragmentation
2534 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2535 allocation locality and avoid filesystem fragmentation.
2537 ------------------------------------------------------------------------------
2539 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2541 * Support for kernels:
2542 2.4.21-47.0.1.EL (RHEL 3)
2543 2.6.5-7.283 (SLES 9)
2544 2.6.9-42.0.10.EL (RHEL 4)
2545 2.6.12.6 vanilla (kernel.org)
2546 2.6.16.27-0.9 (SLES 10)
2547 * Recommended e2fsprogs version: 1.39.cfs5
2549 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2553 Frequency : occasional, depends on client load and configuration
2554 Bugzilla : 12181, 12203
2555 Description: data loss for recently-modified files
2557 Details : In some cases it is possible that recently written or created
2558 files may not be written to disk in a timely manner (this should
2559 normally be within 30s unless client IO load is very high).
2560 The problem appears as zero-length files or files that are a
2561 multiple of 1MB in size after a client crash or client eviction
2562 that are missing data at the end of the file.
2564 This problem is more likely to be hit on clients where files are
2565 repeatedly created and unlinked in the same directory, clients
2566 have a large amount of RAM, have many CPUs, the filesystem has
2567 many OSTs, the clients are rebooted frequently, and/or the files
2568 are not accessed by other nodes after being written.
2570 The presence of the problem can be detected by looking at
2571 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2572 smaller than the second (nr_unused) then dirty files will not
2573 be flushed automatically to disk. "sync; sleep 10" should be
2574 run several times on the node before unmounting it to update
2575 Lustre (this is also safe to run on nodes without this problem).
2577 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2578 that can cause this same problem, so customers using that kernel
2579 also need to update the kernel in addition to Lustre. In order
2580 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2582 It is normal that files written just before a client crash (less
2583 than 30s) may not yet have been flushed to disk, even for local
2587 Frequency : frequent on thin XT3 nodes
2589 Description: UUID collision on thin XT3 Linux nodes
2590 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2591 randomly, since we relied on an insufficiently-seeded PRNG.
2596 Description: OSS hangs after "All ost request buffers busy"
2597 Details : A deadlock between quota and journal operations caused OSS
2598 hangs after printing "All ost request buffers busy."
2601 Frequency : always on liblustre builds
2603 Description: Cleanup compiler warnings on liblustre
2606 Frequency : always on liblustre builds on XT3
2608 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2612 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2613 > 32000 subdirectories
2614 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2615 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2616 subdirectory count crosses 32000. This will aid e2fsck to
2617 correctly handle more than 32000 subdirectories.
2620 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2622 Description: System hang on startup
2623 Details : This bug allowed the liblustre (e.g. catamount) client to
2624 return to the app before handling all startup RPCs. This
2625 could leave the node unresponsive to lustre network traffic
2626 and manifested as a server ptllnd timeout.
2628 Severity : enhancement
2630 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2631 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2632 boolean to control whether to print peer debug info when a
2633 client's RPC times out.
2638 Description: make O_SYNC working on 2.6 kernels
2639 Details : 2.6 kernels use different method for mark pages for write,
2640 so need add a code to lustre for O_SYNC work.
2645 Description: Failure to close file and release space on NFS
2646 Details : Put inode details into lock acquired in ll_intent_file_open.
2647 Use mdc_intent_lock in ll_intent_open to properly
2648 detect all kind of errors unhandled by mdc_enqueue.
2653 Description: proc file read during shutdown sometimes raced obd removal,
2655 Details : Add lock to prevent obd access after proc file removal.
2658 Frequency : Only for files larger than 4GB on 32-bit clients.
2660 Description: improperly doing page alignment of locks
2661 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2662 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2667 Description: RPCs being resent when they shouldn't be
2668 Details : Some RPCs that should not be resent are being resent. This
2669 can cause inconsistencies in the RPC state machine. Do not
2670 resend such requests.
2673 Frequency : rare, only with NFS export
2675 Description: Crash on NFS re-export node
2676 Details : under very unusual load conditions an assertion is hit in
2677 ll_intent_file_open()
2680 Frequency : only if OST filesystem is corrupted
2682 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2683 Details : for a short time RPCs with bulk IO are in the replay list,
2684 but replay of bulk IOs is unimplemented. If the OST filesystem
2685 is corrupted due to disk cache incoherency and then replay is
2686 started it is possible to trip an assertion. Avoid putting
2687 committed RPCs into the replay list at all to avoid this issue.
2692 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2693 Details : Large single O_DIRECT read and write calls can fail to allocate
2694 a sufficiently large buffer to process the request. In case of
2695 allocation failure the allocation is retried with a smaller
2696 buffer and broken into smaller requests.
2698 Severity : enhancement
2700 Description: Add -o localflock option to simulate old noflock behaviour.
2701 Details : This will achieve local-only flock/fcntl locks coherentness.
2706 Description: versioning check is incomplete
2707 Details : Checking the version difference of client vs. server, report
2708 error if the gap is too big.
2713 Description: add support PG_writeback bit
2714 Details : add support for PG_writeback bit for Lustre, for more carefull
2715 work with page cache in 2.6 kernel. This also fix some deadlocks
2716 and remove hack for work O_SYNC with 2.6 kernel.
2718 Severity : enhancement
2720 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2721 Details : The uninit_groups feature works in conjunction with the kernel
2722 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2723 up the pass1 processing of e2fsck. This is a read-only feature
2724 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2725 mount filesystems that have had this feature enabled.
2727 Severity : enhancement
2729 Description: Improve multi-block allocation algorithm to avoid fragmentation
2730 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2731 allocation locality and avoid filesystem fragmentation.
2733 ------------------------------------------------------------------------------
2735 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2737 * Support for kernels:
2738 2.6.9-42.0.3.EL (RHEL 4)
2739 2.6.5-7.276 (SLES 9)
2740 2.4.21-47.0.1.EL (RHEL 3)
2741 2.6.12.6 vanilla (kernel.org)
2742 2.6.16.21-0.8 (SLES10)
2743 * Recommended e2fsprogs version: 1.39.cfs2-0
2745 * The backwards-compatible /proc/sys/portals symlink has been removed
2746 in this release. Before upgrading, please ensure that you change
2747 any configuration scripts or /etc/sysctl.conf files that access
2748 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2749 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2750 in advance of the upgrade on any system running Lustre 1.4.6 or
2751 newer, since /proc/sys/lnet was added in that version.
2752 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2756 Frequency : only when quota is used
2758 Description: avoid scanning export list for quota master
2759 Details : Change the algorithms to avoid scanning export list in order
2760 to improve the efficiency.
2763 Frequency : MDS failover only, very rarely
2765 Description: "went back in time" messages on mds failover
2766 Details : The greatest transno may be lost when the current operation
2767 finishes with an error (transno==0) and the client's last_rcvd
2768 record is over-written. Save the greatest transno in the
2769 mds_last_transno for this case.
2772 Frequency : always for specific kernels and striping counts
2774 Description: client may get "Matching packet too big" without ACL support
2775 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2776 when trying to access files in certain configurations. The
2777 clients should in fact be denied when mounting because they do
2778 not understand ACLs.
2781 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2783 Description: many clients connecting with IO in progress causes connect timeouts
2784 Details : Avoid synchronous journal commits to avoid delays caused by many
2785 clients connecting/disconnecting when bulk IO is in progress.
2786 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2787 OST_IO_PORTAL to avoid delays behind potentially many pending
2791 Frequency : occasionally with multiple writers to a single file
2793 Description: shared writes to file may result in wrong size reported by stat()
2794 Details : Allow growing of kms when extent lock is cancelled
2797 Frequency : always with random mmap IO to multi-striped file
2799 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2800 Details : Only if the hole is at the end of OST object so that kms is too
2801 small. Fix is to increase kms accordingly in ll_nopage.
2804 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2806 Description: writes to a missing object would leak memory on the OST
2807 Details : If there is an inconsistency between the MDS and OST filesystems,
2808 such that the MDS references an object that doesn't exist, writes
2809 to that object will leak memory due to incorrect cleanup in the
2810 error handling path, eventually running out of memory on the OST.
2815 Description: Creating too long symlink causes lustre errors
2816 Details : Check symlink and name lengths before sending requests to MDS.
2819 Frequency : only if flock is enabled (not on by default)
2821 Description: posix locks not released on fd closure on 2.6.9+
2822 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2823 kernels, this caused such locks not to be released on fd close and
2824 then assertions on fs unmount about still used locks.
2827 Frequency : MDS failover only, very rarely
2829 Description: clients may get ASSERTION(granted_lock != NULL)
2830 Details : When request was taking a long time, and a client was resending
2831 a getattr by name lock request. The were multiple lock requests
2832 with the same client lock handle and
2833 mds_getattr_name->fixup_handle_for_resent_request found one of the
2834 lock handles but later failed with ASSERTION(granted_lock != NULL).
2839 Description: handle->h_buffer_credits > 0, assertion failure
2840 Details : h_buffer_credits is zero after truncate, causing assertion
2841 failure. This patch extends the transaction or creates a new
2845 Frequency : NFS re-export or patchless client
2846 Bugzilla : 11179, 10796
2847 Description: Crash on NFS re-export node (__d_move)
2848 Details : We do not want to hash the dentry if we don't have a lock.
2849 But if this dentry is later used in d_move, we'd hit uninitialised
2850 list head d_hash, so we just do this to init d_hash field but
2851 leave dentry unhashed.
2854 Frequency : NFS re-export or patchless client
2856 Description: NFS exports has problem with symbolic link
2857 Details : lustre client didn't properly install dentry when re-exported
2858 to NFS or running patchless client.
2861 Frequency : NFS re-export or patchless client
2863 Description: Various nfs/patchless fixes.
2864 Details : fixes reuse disconected alias for lookup process - this fixes
2865 warning "find_exported_dentry: npd != pd",
2866 fix permission error with open files at nfs.
2867 fix apply umask when do revalidate.
2870 Frequency : occasional
2872 Description: Crash on NFS re-export node
2873 Details : calling clear_page() on the wrong pointer triggered oops in
2874 generic_mapping_read().
2877 Frequency : rarely, using O_DIRECT IO
2879 Description: unaligned directio crashes client with LASSERT
2880 Details : check for unaligned buffers before trying any requests.
2883 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2885 Description: stale data returned from RAID cache
2886 Details : If only a small amount of IO is done to the RAID device before
2887 reading it again it is possible to get stale data from the RAID
2888 cache instead of reading it from disk.
2891 Frequency : always for sles10 kernel
2893 Description: sles10 support
2894 Details : ll_follow_link: compile fixes and using of nd_set_link
2895 under newer kernels.
2898 Frequency : depends on arch, kernel and compiler version, always on sles10
2901 Description: recursive or deep enough symlinks cause stack overflow
2902 Details : getting rid of large stack-allocated variable in
2906 Frequency : depends on hardware
2908 Description: lustre write performance loss in the SLES10 kernel
2909 Details : the performance loss is caused by using of write barriers in the
2910 ext3 code. The SLES10 kernel turns barrier support on by
2911 default. The fix is to undo that change for ldiskfs.
2913 ------------------------------------------------------------------------------
2915 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2917 * Support for kernels:
2918 2.6.9-42.0.3EL (RHEL 4)
2919 2.6.5-7.276 (SLES 9)
2920 2.4.21-47.0.1.EL (RHEL 3)
2921 2.6.12.6 vanilla (kernel.org)
2925 Frequency : quota enabled and large files being deleted
2927 Description: releasing more than 4GB of quota at once hangs OST
2928 Details : If a user deletes more than 4GB of files on a single OST it
2929 will cause the OST to spin in an infinite loop. Release
2930 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2935 Description: statfs data retrieved from /proc may be stale or zero
2936 Details : When reading per-device statfs data from /proc, in the
2937 {kbytes,files}_{total,free,avail} files, it may appear
2938 as zero or be out of date.
2941 Frequency : systems with MD RAID1 external journal devices
2943 Description: lconf's call to blkid is confused by RAID1 journal devices
2944 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2945 of returning all block devices that match the journal UUID.
2948 Frequency : always, for aggregate stripe size over 4GB
2950 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2951 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2952 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2955 Frequency : always if "lfs find" used on a local file/directory
2957 Description: "lfs find" segfaults if used on a local file/directory
2958 Details : The case where a directory component was not specified wasn't
2959 handled correctly. Handle this properly.
2962 Frequency : always on ppc64
2964 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2965 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2966 perform endianness conversion.
2969 Frequency : rarely (truncate to non-zero file size after write under load)
2970 Bugzilla : 10730, 10687
2971 Description: Files padded with zeros to next 4K multiple
2972 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2973 it is possible that files that are truncated to a non-zero size
2974 immediately after being written are filled with zero bytes beyond
2975 the truncated size. No file data is lost.
2977 Severity : enhancement
2979 Description: Allow recovery/failover for liblustre clients.
2980 Details : liblustre clients were unaware of failover configurations until
2983 Severity : enhancement
2985 Description: user file locks should fail when not mounting with flock option
2986 Details : Set up an error-returning stub in ll_file_operations.lock field
2987 to prevent incorrect behaviour when client is mounted without
2988 flock option. Also, set up properly f_op->flock field for
2992 Frequency : always on ia64
2994 Description: "lfs df" loops on printing out MDS statfs information
2995 Details : The obd_ioctl_data was not initialized and in some systems
2996 this caused a failure during the ioctl that did not return
2997 an error. Initialize the struct and return an error on failure.
3000 Frequency : SLES 9 only
3002 Description: Error of copying files with lustre special EAs as root
3003 Details : Client side always return success for setxattr call for lustre
3004 special xattr (currently only "trusted.lov").
3007 Frequency : rarely on clusters with both ia64+i386 clients
3009 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3010 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3011 clients are doing concurrent writes to the same file it is
3012 possible that the ia64 clients may LASSERT because the OST
3013 extent locks are not PAGE_SIZE aligned. Ensure that grown
3014 locks are always aligned on the request boundary.
3017 Frequency : specific use, occasional
3019 Description: Overwriting in use executable truncates on-disk binary image
3020 Details : If one node attempts to overwrite an executable in use by
3021 another node, we now correctly return ETXTBSY instead of
3022 truncating the file.
3024 Severity : enhancement
3026 Description: Async OSC create to avoid the blocking unnecessarily.
3027 Details : If a OST has no remain object, system will block on the creating
3028 when need to create a new object on this OST. Now, ways use
3029 pre-created objects when available, instead of blocking on an
3030 empty osc while others are not empty. If we must block, we block
3031 for the shortest possible period of time.
3036 Description: chmod on Lustre root is propagated to other clients
3037 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3038 invalid by the follow_mount time.
3041 Frequency : liblustre clients only
3043 Description: Race in 'instant cancel' lock handling could lead to such locks
3044 never to be granted in case of SMP MDS
3045 Details : Do not destroy not yet granted but cbpending locks in
3049 Frequency : replay/resend of open
3051 Description: non null lock assetion failure in mds_intent_policy
3052 Details : Trying to replay/resend lockless open requests resulted in
3053 mds_open() returning 0 with no lock. Now it sets a flag if
3054 a lock is going to be returned.
3056 Severity : enhancement
3058 Description: Checksum enhancements
3059 Details : New checksum enhancements allow for resending RPCs that failed
3062 Severity : enhancement
3064 Description: Tunables on number of dirty pages in cacche
3065 Details : Allow to set limit on number of dirty pages cached.
3070 Description: client crash on unmount - lock still has references
3071 Details : In some error handling cases it was possible to leak a lock
3072 reference on a client while accessing a file. This was not
3073 harmful to the client during operation, but would cause the
3074 client to crash when the filesystem is unmounted.
3077 Frequency : specific case, rare
3079 Description: ETXTBSY on mds though file not in use by client
3080 Details : ETXTBSY is no longer incorrectly returned when attempting to
3081 chmod or chown a directory that the user previously tried to
3082 execute or a currently-executing binary.
3085 Frequency : extremely rare except on liblustre-based clients
3087 Description: Lustre space not freed when files are deleted
3088 Details : Clean up open-unlinked files after client eviction. Previously
3089 the unlink was skipped and the files remained as orphans.
3094 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3095 Details : In some cases it was possible to send a blocking callback to
3096 a client doing a glimpse, even though that client didn't get
3097 a lock granted. When the glimpse lock is cancelled on the OST
3098 the freed lock is left on the waiting list and corrupted the list.
3101 Frequency : all core dumps
3103 Description: Broke core dumps to lustre
3104 Details : Negative dentry may be unhashed if parent does not have UPDATE
3105 lock, but some callers, e.g. do_coredump, expect dentry to be
3106 hashed after successful create, hash it in ll_create_it.
3108 ------------------------------------------------------------------------------
3110 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3112 * Support for kernels:
3113 2.6.9-42.0.2.EL (RHEL 4)
3114 2.6.5-7.276 (SLES 9)
3115 2.4.21-40.EL (RHEL 3)
3116 2.6.12.6 vanilla (kernel.org)
3120 Frequency : always on RHEL 3
3122 Description: Number of open files grows over time
3123 Details : The number of open files grows over time, whether or not
3124 Lustre is started. This was due to a filp leak introduced
3125 by one of our kernel patches.
3127 ------------------------------------------------------------------------------
3129 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3131 * Support for kernels:
3132 2.6.9-42.EL (RHEL 4)
3133 2.6.5-7.276 (SLES 9)
3134 2.4.21-40.EL (RHEL 3)
3135 2.6.12.6 vanilla (kernel.org)
3140 Bugzilla : 5719, 9635, 9792, 9684
3141 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3142 Details : If a server is under heavy load and cannot reply to new
3143 connection requests before the client resends the (re)connect,
3144 the connection handling code can behave badly if two service
3145 threads are concurrently handing separate (re)connections from
3146 the same client. Add better locking to the connection handling
3147 code, and ensure that only a single connection will be processed
3148 for a given client UUID, even if the lock is dropped.
3150 Severity : enhancement
3152 Description: add TCP zero-copy support to kernel
3153 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3154 sends if the hardware supports scatter-gather and checksumming.
3155 This allows socklnd to do client-write and server-read more
3156 efficiently and reduce CPU utilization from skbuf copying.
3159 Frequency : only if NFS exporting from client
3161 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3162 Details : If setting attributes on a file created under NFS that had
3163 never been opened it would be possible to oops the client
3164 if the file had no objects.
3167 Frequency : always for liblustre
3169 Description: liblustre client does MDS+OSTs setattr RPC for each write
3170 Details : When doing a write from a liblustre client, the client
3171 incorrectly issued an RPC to the MDS and each OST the file was
3172 striped over in order to update the timestamps. When writing
3173 with small chunks and many clients this could overwhelm the MDS
3174 with RPCs. In all cases it would slow down the write because
3175 these RPCs are unnecessary.
3177 Severity : enhancement
3179 Description: allow number of MDS service threads to be changed at module load
3180 Details : It is now possible to change the number of MDS service threads
3181 running. Adding "options mds mds_num_threads={N}" to the MDS's
3182 /etc/modprobe.conf will set the number of threads for the next
3183 time Lustre is restarted (assuming the "mds" module is also
3184 reloaded at that time). The default number of threads will
3185 stay the same, 32 for most systems.
3190 Description: OST crash if filesystem is unformatted or corrupt
3191 Details : If an OST is started on a device that has never been formatted
3192 or if the filesystem is corrupt and cannot even mount then the
3193 error handling cleanup routines would dereference a NULL pointer.
3198 Description: NULL pointer deref in llap_from_page.
3199 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3200 incorrect) mapping if the page was truncated/reclaimed while it was
3201 searched for. Check for this condition and skip such pages when
3202 doing readahead. Introduce extra check to llap_from_page() to
3203 verify page->mapping->host is non-NULL (so page is not anonymous).
3206 Frequency : Sometimes when using sys_sendfile
3208 Description: "page not covered by a lock" warnings from ll_readpage
3209 Details : sendfile called ll_readpage without right page locks present.
3210 Now we introduced ll_file_sendfile that does necessary locking
3211 around call to generic_file_sendfile() much like we do in
3215 Frequency : with certain MDS communication failures at client mount time
3217 Description: NULL pointer deref after failed client mount
3218 Details : a client connection request may delayed by the network layer
3219 and not be sent until after the PTLRPC layer has timed out the
3220 request. If the client fails the mount immediately it will try
3221 to clean up before the network times out the request. Add a
3222 reference from the request import to the obd device and delay
3223 the cleanup until the network drops the request.
3226 Frequency : occasionally during client (re)connect
3228 Description: assertion failure during client (re)connect
3229 Details : processing a client connection request may be delayed by the
3230 client or server longer than the client connect timeout. This
3231 causes the client to resend the connection request. If the
3232 original connection request is replied in this interval, the
3233 client may trip an assertion failure in ptlrpc_connect_interpret()
3234 which thought it would be the only running connect process.
3237 Frequency : only with obd_echo servers and clients that are rebooted
3239 Description: kernel BUG accessing uninitialized data structure
3240 Details : When running an obd_echo server it did not start the ping_evictor
3241 thread, and when a client was evicted an uninitialized data
3242 structure was accessed. Start the ping_evictor in the RPC
3243 service startup instead of the OBD startup.
3245 Severity : enhancement
3246 Bugzilla : 10193 (patchless)
3247 Description: Remove dependency on various unexported kernel interfaces.
3248 Details : No longer need reparent_to_init, exit_mm, exit_files,
3249 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3252 Frequency : rare (only users of deprecated and unsupported LDAP config)
3254 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3255 Details : LDAP apparently contains 'lustreName' attributes instead of
3256 'name'. A simple remapping of the name is sufficient.
3259 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3261 Description: waiting_locks_callback trips kernel BUG if client is evicted
3262 Details : Running with the dump_on_timeout debug flag turned on makes
3263 it possible that the waiting_locks_callback() can try to dump
3264 the Lustre kernel debug logs from an interrupt handler. Defer
3265 this log dumping to the expired_lock_main() thread.
3267 Severity : enhancement
3269 Description: Support NFS exporting on 2.6 kernels.
3270 Details : Implement non-rawops metadata methods for NFS server to use without
3271 changing NFS server code.
3274 Frequency : very rare (synthetic metadata workload only)
3276 Description: two racing renames might cause an MDS thread to deadlock
3277 Details : Running the "racer" program may cause one MDS thread to rename
3278 a file from being the source of a rename to being the target of
3279 a rename at exactly the same time that another thread is doing
3280 so, and the second thread has already enqueued these locks after
3281 doing a lookup of the target and is trying to relock them in
3282 order. Ensure that we don't try to re-lock the same resource.
3285 Frequency : only very large systems with liblustre clients
3287 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3288 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3289 all OSTs in parallel. This allows the eviction of stale liblustre
3290 clients to proceed much faster than if they were done in series,
3291 and also offers similar improvements for other set_info RPCs.
3296 Description: excessive CPU usage during initial read phase on client
3297 Details : During the initial read phase on a client, it would agressively
3298 retry readahead on the file, consuming too much CPU and impacting
3299 performance (since 1.4.5.8). Improve the readahead algorithm
3300 to avoid this, and also improve some other common cases (read
3301 of small files in particular, where "small" is files smaller than
3302 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3307 Description: MDS crash when receiving packet with unknown intent.
3308 Details : Do not LBUG in unknown intent case, just return -EFAULT
3310 Severity : enhancement
3311 Bugzilla : 9293, 9385
3312 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3313 Details : Do not serialize getattr (non-intent version) and statfs.
3316 Frequency : occasional, when OST network is overloaded/intermittent
3318 Description: client evicted by OST after bulk IO timeout
3319 Details : If a client sends a bulk IO request (read or write) the OST
3320 may evict the client if it is unresposive to its data GET/PUT
3321 request. This is incorrect if the network is overloaded (takes
3322 too long to transfer the RPC data) or dropped the OST GET/PUT
3323 request. There is no need to evict the client at all, since
3324 the pinger and/or lock callbacks will handle this, and the
3325 client can restart the bulk request.
3328 Frequency : Always when mmapping file with no objects
3330 Description: client crashes when mmapping file with no objects
3331 Details : Check that we actually have objects in a file before doing any
3332 operations on objects in ll_vm_open, ll_vm_close and
3338 Description: Request leak when working with deleted CWD
3339 Details : Introduce advanced request refcount tracking for requests
3340 referenced from lustre intent.
3342 Severity : Enhancement
3344 Description: Cache open file handles on client.
3345 Details : MDS now will return special lock along with openhandle, if
3346 requested and client is allowed to hold openhandle, even if unused,
3347 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3348 closing files for every read/write openration.
3350 Severity : Enhancement
3352 Description: Cache open negative dentries on client when possible.
3353 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3354 negative dentries on lock revocation.
3359 Description: Remounting a client read-only wasn't possible with a zconf mount
3360 Details : It wasn't possible to remount a client read-only with llmount.
3362 Severity : enhancement
3363 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3364 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3365 MPICH is included to add Lustre-specific ADIO interfaces.
3366 This is based closely on the UFS ADIO layer and only differs
3367 in file creation, in order to allow the OST striping to be set.
3368 This is user-contributed code and not supported by CFS.
3373 Description: extended inode attributes (immutable, append-only) work improperly
3374 when 2.4 and 2.6 kernels are used on client/server or vice versa
3375 Details : Introduce kernel-independent values for these flags.
3377 Severity : enhancement
3380 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3381 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3382 granularity. Now, Fractional MB granularity is be supported,
3383 this is very useful for low memory system.
3385 Severity : enhancement
3387 Description: Getattr by fid
3388 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3389 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3390 have locks and avoids getattr rpc requests for it on every stat.
3393 Frequency : Always, for filesystems larger than 2TB
3395 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3396 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3397 filesystems larger than 2TB to be mounted with mballoc enabled.
3400 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3402 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3403 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3404 filesystems larger than 2TB is mounted then the kernel will
3405 silently corrupt the start of the filesystem. CONFIG_LBD is
3406 enabled for all CFS-supported kernels, but the possibility of
3407 this happening with a modified kernel config exists.
3409 Severity : enhancement
3411 Description: add client O_DIRECT support for 2.6 kernels
3412 Details : It is now possible to do O_DIRECT reads and writes to files
3413 in the Lustre client mountpoint on 2.6 kernel clients.
3415 Severity : enhancement
3417 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3418 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3419 parallel, not waiting for response from every OST before sending
3420 a rpc to the next OST.
3425 Description: setattr vs write race when updating file timestamps
3426 Details : Client processes that update a file timestamp into the past
3427 right after writing to the file (e.g. tar) it is possible that
3428 the updated file modification time can be reset to the current
3429 time due to a race between processing the setattr and write RPC.
3431 Severity : enhancement
3433 Description: Bring 'lfs find' closer in line with regular Linux find.
3434 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3435 -print0 options and obtains all the needed info through the lustre
3438 Severity : enhancement
3440 Description: support up to 1024 configured devices on one node
3441 Details : change obd_dev array from statically allocated to dynamically
3442 allocated structs as they are first used to reduce memory usage
3447 Description: Flush dirty partially truncated pages during truncate
3448 Details : Immediatelly flush partially truncated pages in filter_setattr,
3449 this way we completely avoid having any pages in page cache on OST
3450 and can retire ugly workarounds during writes to flush such pages.
3455 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3456 Details : protect inode from truncation within vfs_unlink() context
3457 just take a reference before calling vfs_unlink() and release it
3458 when parent's i_sem is free.
3463 Description: last_id value checked outside lock on OST caused LASSERT failure
3464 Details : If there were multiple MDS->OST object precreate requests in
3465 flight, it was possible that the OST's last object id was checked
3466 outside a lock and incorrectly tripped an assertion. Move checks
3467 inside locks, and discard old precreate requests.
3470 Frequency : always, if extents are used on OSTs
3472 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3473 Details : OSTs using the extents format would not zero the high 16 bits of
3474 the index physical block number. This is not a problem for any
3475 OST filesystems smaller than 16TB, and no kernels support ext3
3476 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3477 new/modified files) and can be fixed for existing filesystems
3478 with e2fsprogs-1.39-cfs1.
3483 Description: import connection selection may be incorrect if timer wraps
3484 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3485 import connections to be ignored if the 32-bit jiffies counter
3486 wraps. Use a 64-bit jiffies counter.
3489 Frequency : very large clusters immediately after boot
3491 Description: LNET request buffers exhausted under heavy short-term load
3492 Details : If a large number of client requests are generated on a service
3493 that has previously never seen so many requests it is possible
3494 that the request buffer growth cannot keep up with the spike in
3495 demand. Instead of dropping incoming requests, they are held in
3496 the LND until the RPC service can accept more requests.
3499 Frequency : Sometimes during replay
3501 Description: Assertion failure in ll_local_open after replay.
3502 Details : If replay happened on an open request reply before we were able
3503 to set replay handler, reply will become not swabbed tripping the
3504 assertion in ll_local_open. Now we set the handler right after
3505 recognising of open request
3508 Frequency : very rare
3510 Description: kernel reports "badness in vsnprintf"
3511 Details : Reading from the "recovery_status" /proc file in small chunks
3512 may cause a negative length in lprocfs_obd_rd_recovery_status()
3513 call to vsnprintf() (which is otherwise harmless). Exit early
3514 if there is no more space in the output buffer.
3516 Severity : enhancement
3518 Description: clear OBD RPC statistics by writing to them
3519 Details : It is now possible to clear the OBD RPC statistics by writing
3520 to the "stats" file.
3525 Description: Client mtime is not the same on different clients after utimes
3526 Details : In some cases, the client was using the utimes() syscall on
3527 a file cached on another node. The clients now validate the
3528 ctime from the MDS + OSTs to determine which one is right.
3533 Description: Inability to activate failout mode
3534 Details : lconf script incorrectly assumed that in pythong string's numeric
3535 value is used in comparisons.
3538 Frequency : always with multiple stripes per file
3540 Description: Inefficient object allocation for mutli-stripe files
3541 Details : When selecting which OSTs to stripe files over, for files with
3542 a stripe count that divides evenly into the number of OSTs,
3543 the MDS is always picking the same starting OST for each file.
3544 Return the OST selection heuristic to the original design.
3549 Description: mount failures may take full timeout to return an error
3550 Details : Under some heavy load conditions it is possible that a
3551 failed mount can wait for the full obd_timeout interval,
3552 possibly several minutes, before reporting an error.
3553 Instead return an error as soon as the status is known.
3555 Frequency : quota enabled and large files being deleted
3557 Description: releasing more than 4GB of quota at once hangs OST
3558 Details : If a user deletes more than 4GB of files on a single OST it
3559 will cause the OST to spin in an infinite loop. Release
3560 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3565 Description: statfs data retrieved from /proc may be stale or zero
3566 Details : When reading per-device statfs data from /proc, in the
3567 {kbytes,files}_{total,free,avail} files, it may appear
3568 as zero or be out of date.
3571 Frequency : systems with MD RAID1 external journal devices
3573 Description: lconf's call to blkid is confused by RAID1 journal devices
3574 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3575 of returning all block devices that match the journal UUID.
3578 Frequency : always, for aggregate stripe size over 4GB
3580 Description: assertion fails when trying to use 4GB stripe size
3581 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3582 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3585 Frequency : always on ppc64
3587 Description: the first write on an ext3 filesystem with mballoc got stuck
3588 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3589 perform endianness conversion.
3591 ------------------------------------------------------------------------------
3593 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3595 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3596 INTEROPERATE with older versions automatically. Please read the
3597 user documentation before upgrading any part of a live system.
3598 * WARNING: Lustre networking configuration changes are required with
3599 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3602 * Support for kernels:
3603 2.6.9-22.0.2.EL (RHEL 4)
3604 2.6.5-7.244 (SLES 9)
3605 2.6.12.6 vanilla (kernel.org)
3608 Severity : enhancement
3609 Bugzilla : 7981/8208
3610 Description: Introduced Lustre Networking (LNET)
3611 Details : LNET is new networking infrastructure for Lustre, it includes
3612 a reorganized network configuration mode (see the user
3613 documentation for full details) as well as support for routing
3614 between different network fabrics. Lustre Networking Devices
3615 (LNDs) for the supported network fabrics have also been
3616 created for this new infrastructure.
3618 Severity : enhancement
3619 Description: Introduced Access control lists
3620 Details : clients can set ACLs on files and directories in order to have
3621 more fine-grained permissions than the standard Unix UGO+RWX.
3622 The MDS must be started with the "-o acl" mount option.
3624 Severity : enhancement
3625 Description: Introduced filesystem quotas
3626 Details : Administrators may now establish per-user quotas on the
3629 Severity : enhancement
3631 Description: Configuration change for the XT3
3632 The PTLLND is now used to run Lustre over Portals on the XT3
3633 The configure option(s) --with-cray-portals are no longer used.
3634 Rather --with-portals=<path-to-portals-includes> is used to
3635 enable building on the XT3. In addition to enable XT3 specific
3636 features the option --enable-cray-xt3 must be used.
3641 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3642 Details : A race between allocating a new llcd and re-getting the llcd_lock
3643 allowed another thread to grab newly-allocated llcd.
3645 Severity : enhancement
3647 Description: 2.6 OST async journal commit and locking fix to improve performance
3648 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3649 2.6 kernels are now async as they already were in 2.4 kernels so
3650 that they can commit concurrently with the network bulk transfer.
3651 For block-allocated files the filter allocation semaphore is held
3652 to avoid filesystem fragmentation during allocation. BKL lock
3653 removed for 2.6 xattr operations where it is no longer needed.
3658 Description: lconf incorrectly determined whether two IP networks could talk
3659 Details : In some more complicated routing and multiple-network
3660 configurations, lconf will avoid trying to make a network
3661 connection to a disjoint part of the IP space. It was doing the
3662 math incorrectly for one set of cases.
3667 Description: Fix for potential infinite loop processing records in an llog.
3668 Details : If an llog record is corrupted/zeroed, it is possible to loop
3669 forever in llog_process(). Validate the llog record length
3670 and skip the remainder of the block on error.
3673 Frequency : occasional (liblustre only)
3675 Description: liblustre could not open files whose last component is a symlink
3676 Details : sysio_path_walk() would incorrectly pass the open intent to
3677 intermediate path components.
3680 Frequency : rare (liblustre only with non-standard tuning)
3681 Bugzilla : 7201 (7350)
3682 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3683 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3684 after it has been released, only do this for referenced locks
3686 Severity : enhancement
3688 Description: specifying an (invalid) directory default stripe_size of -1
3689 would reset the directory default striping
3690 Details : stripe_size -1 was used internally to signal directory stripe
3691 removal, now use "all default" to signal dir stripe removal
3692 as a directory striping of "all default" is not useful
3695 Frequency : common for large clusters running liblustre clients
3697 Description: doing an ls when liblustre clients are running is slow
3698 Details : sending a glimpse AST to a liblustre client waits for every AST
3699 to time out, as liblustre clients will not respond. Since they
3700 cannot cache data we refresh the OST lock LVB from disk instead.
3702 Severity : enhancement
3704 Description: doing an ls at the same time as file IO can be slow
3705 Details : enqueue and other "small" requests can be blocked behind many
3706 large IO requests. Create a new OST IO portal for non-IO
3707 requests so they can be processed faster.
3710 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3712 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3713 created file with mode 000
3714 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3715 allow an ftruncate() as a non-root user to a file with mode 000.
3716 Lustre now respects this flag to disable mode checks when
3717 truncating a file owned by the user
3720 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3723 Description: Revoking locks from clients that went dead or catatonic might take
3725 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3726 cancellation of such locks instant on servers without waiting for
3727 any reply from clients. Clients drops these locks when cancel
3728 notification from server is received without replying.
3731 Frequency : liblustre-only, when liblustre client dies or becomes busy
3733 Description: Doing ls on Linux clients can take a long time with active
3735 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3736 granting such locks to it in the first place if possible. Locks
3737 are taken by proxy on the OST during the read or write and
3738 dropped immediately afterward. Add connect flags handling, do
3739 not grant locks to liblustre clients for glimpse ASTs.
3741 Severity : enhancement
3743 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3744 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3745 sequential reads and behaves badly with applications doing random
3746 reads. Improve it by reading ahead at least read region, and
3747 avoiding excessive large RPC for small reads.
3749 Severity : enhancement
3751 Description: Creating more than 1000 files for a single job may cause a load
3752 imbalance on the OSTs if there are also a large number of OSTs.
3753 Details : qos_prep_create() uses an OST index reseed value that is an
3754 even multiple of the number of available OSTs so that if the
3755 reseed happens in the middle of the object allocation it will
3756 still utilize the OSTs as uniformly as possible.
3761 Description: OST or MDS may oops in ping_evictor_main()
3762 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3763 but doesn't restart at beginning of obd_exports_timed list
3766 Severity : enhancement
3768 Description: improve by-nid export eviction on the MDS and OST
3769 Details : allow multiple exports with the same NID to be evicted at one
3770 time without re-searching the exports list.
3773 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3775 Description: MDS may oops in groups_free()
3776 Details : in rare race conditions a newly allocated group_info struct is
3777 freed again, and this can be NULL. The 2.4 compatibility code
3778 for groups_free() checked for a NULL pointer, but 2.6 did not.
3781 Frequency : common for liblustre clients doing little filesystem IO
3782 Bugzilla : 9352, 7313
3783 Description: server may evict liblustre clients accessing contended locks
3784 Details : if a client is granted a lock or receives a completion AST
3785 with a blocking AST already set it would not reply to the AST
3786 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3789 Frequency : lfs setstripe, only systems with more than 160 OSTs
3791 Description: unable to set striping with a starting offset beyond OST 160
3792 Details : llapi_create_file() incorrectly limited the starting stripe
3793 index to the maximum single-file stripe count.
3796 Frequency : LDAP users only
3798 Description: lconf did not handle in-kernel recovery with LDAP properly
3799 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3801 Severity : enhancement
3803 Description: bind OST threads to NUMA nodes to improve performance
3804 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3805 node and do their allocations there to localize memory access
3807 Severity : enhancement
3809 Description: llmount can determine client NID directly from Myrinet (GM)
3810 Details : the client NID code from gmnalnid was moved directly into
3811 llmount, removing the need to use this or specifying the
3812 client NID explicitly when mounting GM clients with zeroconf
3815 Frequency : if client is started with down MDS
3817 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3818 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3819 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3820 ptlrpc_import_delayed_req() because we didn't check if the
3821 request was interrupted, and we also didn't break out of the
3822 event loop if there was no timeout
3827 Description: data loss during non-page-aligned writes to a single file from
3828 both multiple nodes and multiple threads on one node at same time
3829 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3830 inconsistency led to false short-reads, that were cached and later
3831 used by ->prepare_write() to fill in partially written page,
3832 leading to data loss.
3835 Frequency : always, if lconf --abort_recovery used
3837 Description: lconf --abort_recovery fails with 'Operation not supported'
3838 Details : lconf was attempting to abort recovery on the MDT device and not
3841 Severity : enhancement
3843 Description: remove cleanup logs
3844 Details : replace lconf-generated cleanup logs with lustre internal
3845 cleanup routines. Eliminates the need for client-cleanup and
3848 Severity : enhancement
3850 Description: add support for EAs (user and system) on lustre filesystems
3851 Details : it is now possible to store extended attributes in the Lustre
3852 client filesystem, and with the user_xattr mount option it
3853 is possible to allow users to store EAs on their files also
3855 Severity : enhancement
3857 Description: Add possibility (config option) to show minimal available OST free
3859 Details : When compiled with --enable-mindf configure option, statfs(2)
3860 (and so, df) will return least minimal free space available from
3861 all OSTs as amount of free space on FS, instead of summary of
3862 free spaces of all OSTs.
3864 Severity : enhancement
3866 Description: do not expand extent locks acquired on OST-side
3867 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3868 by server: they are not cached anyway.
3871 Frequency : when mmap is used/binaries executed from Lustre
3873 Description: Unmmap pages before throwing them away from read cache.
3874 Details : llap_shrink cache now attempts to unmap pages before discarding
3875 them (if unmapping failed - do not discard). SLES9 kernel has
3876 extra checks that trigger if this unmapping is not done first.
3881 Description: lconf didn't resolve symlinks before checking to see whether a
3882 given mountpoint was already in use
3885 Frequency : when migrating failover services
3886 Bugzilla : 6395, 9514
3887 Description: When migrating a subset of services from a node (e.g. failback
3888 from a failover service node) the remaining services would
3889 time out and evict clients.
3890 Details : lconf --force (implied by --failover) sets the global obd_timeout
3891 to 5 seconds in order to quickly disconnect, but this caused
3892 other RPCs to time out too quickly. Do not change the global
3893 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3895 Severity : enhancement
3896 Frequency : if MDS is started with down OST
3897 Bugzilla : 9439,5706
3898 Description: Allow startup/shutdown of an MDS without depending on the
3899 availability of the OSTs.
3900 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3901 Add appropriate locking and lov-osc refcounts for safe
3902 cleaning. Add osc abort_inflight calls in case the
3903 synchronize never started.
3906 Frequency : occasional (Cray XT3 only)
3908 Description: root not authorized to access files in CRAY_PORTALS environment
3909 Details : The client process capabilities were not honoured on the MDS in
3910 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3911 been accessed by an authorized user then root was able to access
3912 the file on the local client also. The root user capabilities
3913 are now allowed on the MDS, as this environment has secure UID.
3916 Frequency : occasional
3918 Description: ldiskfs "too long searching" message happens too often
3919 Details : A debugging message (otherwise harmless) prints too often on
3920 the OST console. This has been reduced to only happen when
3921 there are fragmentation problems on the filesystem.
3926 Description: Division by zero in statfs when all OSCs are inactive
3927 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3928 lov_statfs divides by value returned by lov_get_stripecnt().
3932 Bugzilla : 9489, 3273
3933 Description: First write from each client to each OST was only 4kB in size,
3934 to initialize client writeback cache, which caused sub-optimal
3935 RPCs and poor layout on disk for the first writen file.
3936 Details : Clients now request an initial cache grant at (re)connect time
3937 and so that they can start streaming writes to the cache right
3938 away and always do full-sized RPCs if there is enough data.
3939 If the OST is rebooted the client also re-establishes its grant
3940 so that client cached writes will be honoured under the grant.
3945 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3946 Details : Now I/O RPCs go to different portal number and (presumably) fast
3947 lock requests (and glimses) and other RPCs get their own service
3948 threads pool that should be able to service those RPCs
3951 Severity : enhancement
3953 Description: Ability to exchange lustre version between client and servers and
3954 issue warnings at client side if client is too old. Also for
3955 liblustre clients there is ability to refuse connection of too old
3957 Details : New 'version' field is added to connect data structure that is
3958 filled with version info. That info is later checked by server and
3962 Frequency : rare, liblustre only.
3963 Bugzilla : 9296, 9581
3964 Description: Two simultaneous writes from liblustre at offset within same page
3965 might proceed at the same time overwriting eachother with stale
3967 Details : I/O lock withing llu_file_prwv was released too early, before data
3968 actually was hitting the wire. Extended lock-holding time until
3969 server acknowledges receiving data.
3972 Frequency : extremely rare. Never observed in practice.
3974 Description: avoid generating lustre_handle cookie of 0.
3975 Details : class_handle_hash() generates handle cookies by incrementing
3976 global counter, and can hit 0 occasionaly (this is unlikely, but
3977 not impossible, because initial value of cookie counter is
3978 selected randonly). Value of 0 is used as a sentinel meaning
3979 "unassigned handle" --- avoid it. Also coalesce two critical
3980 sections in this function into one.
3982 Severity : enhancement
3984 Description: allow liblustre clients to delegate truncate locking to OST
3985 Details : To avoid overhead of locking, liblustre client instructs OST to
3986 take extent lock in ost_punch() on client's behalf. New connection
3987 flag is added to handle backward compatibility.
3989 Severity : enhancement
3990 Bugzilla : 4928, 7341, 9758
3991 Description: allow number of OST service threads to be specified
3992 Details : a module parameter allows the number of OST service threads
3993 to be specified via "options ost ost_num_threads={N}" in the
3994 OSS's /etc/modules.conf or /etc/modprobe.conf.
3998 Bugzilla : 6146, 9635, 9895
3999 Description: servers crash with bad pointer in target_handle_connect()
4000 Details : In rare cases when a client is reconnecting it was possible that
4001 the connection request was the last reference for that export.
4002 We would temporarily drop the export reference and get a new
4003 one, but this may have been the last reference and the export
4004 was just destroyed. Get new reference before dropping old one.
4006 Severity : enhancement
4007 Frequency : if client is started with failover MDS
4009 Description: Allow multiple MDS hostnames in the mount command
4010 Details : Try to read the configuration from all specified MDS
4011 hostnames during a client mount in case the "primary"
4014 Severity : enhancement
4016 Description: Stop sending data to evicted clients as soon as possible.
4017 Details : Check if the client we are about to send or are sending data to
4018 was evicted already. (Check is done every second of waiting,
4019 for which l_wait_event interface was extended to allow checking
4020 of exit condition at specified intervals).
4023 Frequency : rare, normally only when NFS exporting is done from client
4025 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4027 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4030 Frequency : always, if srand() is called before liblustre initialization
4032 Description: Liblustre uses system PRNG disturbing its usage by user application
4033 Details : Introduce internal to lustre fast and high-quality PRNG for
4034 lustre usage and make liblustre and some other places in generic
4035 lustre code to use it.
4037 Severity : enhancement
4038 Bugzilla : 9477, 9557, 9870
4039 Description: Verify that the MDS configuration logs are updated when xml is
4040 Details : Check if the .xml configuration logs are newer than the config
4041 logs stored on the MDS and report an error if this is the case.
4042 Request --write-conf, or allow starting with --old_conf.
4044 Severity : enhancement
4046 Description: Handle symlinks in the path when checking if Lustre is mounted.
4047 Details : Resolve intermediate symlinks when checking if a client has
4048 mounted a filesystem to avoid duplicate client mounts.
4053 Description: lconf can hit an error exception but still return success.
4054 Details : The lconf command catches the Command error exception at the top
4055 level script context and will exit with the associated exit
4056 status, but doesn't ensure that this exit status is non-zero.
4061 Description: failure of ptlrpc thread startup can cause oops
4062 Details : Starting a ptlrpc service thread can fail if there are a large
4063 number of threads or the server memory is very fragmented.
4064 Handle this without oopsing.
4067 Frequency : always, only if liblustre and non-default acceptor port was used
4069 Description: liblustre cannot connect to servers with non-default acceptor port
4070 Details : tcpnal_set_default_params() was not called and was therefore
4071 ignoring the environment varaible TCPNAL_PORT, as well as other
4072 TCPNAL_ environment variables
4077 Description: two objects could be created on the same OST for a single file
4078 Details : If an OST is down, in some cases it was possible to create two
4079 objects on a single OST for a single file. No problems other
4080 than potential performance impact and spurious error messages.
4084 Bugzilla : 5681, 9562
4085 Description: Client may oops in ll_unhash_aliases
4086 Details : Client dcache may become inconsistent in race condition.
4087 In some cases "getcwd" can fail if the current directory is
4093 Description: Inode refcounting problems in NFS export code
4094 Details : link_raw functions used to call d_instantiate without obtaining
4095 extra inode reference first.
4099 Bugzilla : 9942, 9903
4100 Description: Referencing freed requests leading to crash, memleaks with NFS.
4101 Details : We used to require that call to ll_revalidate_it was always
4102 followed by ll_lookup_it. Also with revalidate_special() it is
4103 possible to call ll_revalidate_it() twice for the same dentry
4104 even if first occurence returned success. This fix changes semantic
4105 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4106 reference on a request referred from the intent.
4107 ll_intent_release() then releases such a request.
4110 Frequency : rare, normally benchmark loads only
4112 Description: unlinked inodes were kept in memory on the client
4113 Details : If a client is repeatedly creating and unlinking files it
4114 can accumulate a lot of stale inodes in the inode slab cache.
4115 If there is no other client load running this can cause the
4116 client node to run out of memory. Instead flush old inodes
4117 from client cache that have the same inode number as a new inode.
4120 Frequency : SLES9 2.6.5 kernel and long filenames only
4121 Bugzilla : 9969, 10379
4122 Description: utime reports stale NFS file handle
4123 Details : SLES9 uses out-of-dentry names in some cases, which confused
4124 the lustre dentry revalidation. Change it to always use the
4128 Frequency : rare, unless heavy write-truncate concurrency is continuous
4129 Bugzilla : 4180, 6984, 7171, 9963, 9331
4130 Description: OST becomes very slow and/or deadlocked during object unlink
4131 Details : filter_destroy() was holding onto the parent directory lock
4132 while truncating+unlinking objects. For very large objects this
4133 may block other threads for a long time and slow overall OST
4134 responsiveness. It may also be possible to get a lock ordering
4135 deadlock in this case, or run out of journal credits because of
4136 the combined truncate+unlink. Solution is to do object truncate
4137 first in one transaction without parent lock, and then do the
4138 final unlink in a new transaction with the parent lock. This
4139 reduces the lock hold time dramatically.
4142 Frequency : rare, 2.4 kernels only
4144 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4145 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4146 before unmounting their filesystem in order to avoid deadlock.
4147 The kernel_locked() function in 2.4 kernels only checks whether
4148 the kernel lock is held, not whether it is this process that is
4149 holding it as 2.6 kernels do.
4154 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4155 Details : The client ptlrpc code may be trying to reconnect to a down
4156 server before a previous connection attempt has timed out.
4157 Increase the reconnect interval to be longer than the connection
4158 timeout interval to avoid sending duplicate connections to
4162 Frequency : echo_client brw_test command
4164 Description: fix echo_client to work with OST preallocated code
4165 Details : OST preallocation code (5137) didn't take echo_client IO path
4166 into account: echo_client calls filter methods outside of any
4167 OST thread and, hence, there is no per-thread preallocated
4168 pages and buffers to use. Solution: hijack pga pages for IO. As
4169 a byproduct, this avoids unnecessary data copying.
4173 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4174 Description: Client can oops in mdc_commit_close() after open replay
4175 Details : It was possible for the MDS to return an open request with no
4176 transaction number in mds_finish_transno() if the client was
4177 evicted, but without actually returning an error. Clients
4178 would later try to replay that open and may trip an assertion
4179 Simplify the client close codepath, and always return an error
4180 from the MDS in case the open is not successful.
4183 Frequency : rare, 2.6 OSTs only
4185 Description: OST may deadlock under high load on fragmented files
4186 Details : If there was a heavy load and highly-fragmented OST filesystems
4187 it was possible to have all the OST threads deadlock waiting on
4188 allocation of biovecs, because the biovecs were not released
4189 until the entire RPC IO was completed. Instead, release biovecs
4190 as soon as they are complete to ensure forward IO progress.
4192 Severity : enhancement
4194 Description: Support for specifying external journal device at mount
4195 Details : If an OST or MDS device is formatted with an external journal
4196 device, this device major/minor is stored in the ext3 superblock
4197 and may not be valid for failover. Allow detecting and
4198 specifying the external journal at mount time.
4203 Description: Mounting an MDS with pending unlinked files may cause oops
4204 Details : target_finish_recovery() calls mds_postrecov() which returned
4205 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4206 considers this an error and immediately begins cleaning up the
4207 lov, just after starting the mds_lov process
4209 Severity : enhancement
4211 Description: Implement 'lfs df' to report actual free space on per-OST basis
4212 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4213 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4214 usage of objects; '-h' to report in human readable format.
4216 ------------------------------------------------------------------------------
4218 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4225 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4226 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4227 at startup during orphan unlinking. Instead, skip buddy update
4228 before setup, as it will be regenerated after recovery is complete.
4233 Description: If an OST is inactive, its locks might reference stale inodes.
4234 Details : lov_change_cbdata() must iterate over all namespaces, even if
4235 they are inactive to clear inode references from the lock.
4237 Severity : enhancement
4238 Frequency : occasional, if non-standard max_dirty_mb used
4240 Description: Client will block write RPCs if not enough grant
4241 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4242 then the client will block writes while waiting for another RPC
4243 to complete instead of consuming its dirty limit. With change
4244 we get improved performance when max_dirty_mb is small.
4246 Severity : enhancement
4247 Bugzilla : 3389, 6253
4248 Description: Add support for supplementary groups on the MDS.
4249 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4250 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4251 lookups for user supplementary groups into a cache.
4255 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4256 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4257 for liblustre O_EXCL is enforced by MDS.
4261 Description: readdir never returns NULL in liblustre.
4262 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4263 dentry in struct dirent.
4267 Frequency : liblustre only, depends on application IO pattern
4268 Description: liblustre clients evicted if not contacting servers
4269 Details : Don't put liblustre clients into the ping_evictor list, so
4270 they will not be evicted by the pinger ever.
4272 Severity : enhancement
4274 Description: Add ability to evict clients by NID from MDS.
4275 Details : By echoing "nid:$NID" string into
4276 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4277 $NID would be instantly evicted from this MDS and from all active
4278 OSTs connected to it.
4282 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4283 Details : lookup_it_finish() used to query file size from OSTs that was not
4288 Description: service threads change working directory to that of init
4289 Details : Starting lustre service threads may pin the working directory
4290 of the parent thread, making that filesystem busy. Threads
4291 now change to the working directory of init to avoid this.
4295 Frequency : during shutdown only
4296 Description: shutdown with a failed MDS or OST can cause unmount to hang
4297 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4302 Frequency : 2.6 only
4303 Description: chmod/chown may include an extra supplementary group
4304 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4305 supplementary group and if none is specified this is used.
4308 Bugzilla : 5479 (6816)
4310 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4311 Details : If lookup is in progress on a file that is unlinked we might try
4312 to revalidate the inode and fail in revalidate after lookup is
4313 complete and ll_file_open() enqueues the open again but
4314 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4317 Frequency : always, if lconf --abort_recovery used
4319 Description: lconf --abort_recovery fails with 'Operation not supported'
4320 Details : lconf was attempting to abort recovery on the MDT device and not
4323 ------------------------------------------------------------------------------
4325 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4330 Frequency : rare (only unsupported configurations with a node running as an
4332 Bugzilla : 6514, 5137
4333 Description: Mounting a Lustre file system on a node running as an OST could
4335 Details : OSTs now preallocates memory needed to write out data at
4336 startup, instead of when needed, to avoid having to
4337 allocate memory in possibly low memory situations.
4338 Specifically, if the file system is mounted on on OST,
4339 memory pressure could force it to try to write out data,
4340 which it needed to allocate memory to do. Due to the low
4341 memory, it would be unable to do so and the node would
4342 become unresponsive.
4344 Severity : enhancement
4346 Description: Addition of lconf --service command line option
4347 Details : lconf now accepts a '--service <arg>' option, which is
4348 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4350 Severity : enhancement
4352 Description: Failover mode is now the default for OSTs.
4353 Details : By default, OSTs will now run in failover mode. To return to
4354 the old behaviour, add '--failout' to the lmc line for OSTs.
4356 Severity : enhancement
4358 Description: Health checks are now provided for MDS and OSTs
4359 Details : Additional detailed health check information on MSD and OSTs
4360 is now provided through the procfs health_check value.
4363 Frequency : occasional, depends on IO load
4365 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4366 after numerous create/delete cycles
4367 Details : The ext3 inode allocation policy would not allocate new inodes
4368 very well on the OSTs because there are no new directories
4369 being created. Instead we look for groups with free space if
4370 the parent directories are nearly full.
4375 Description: Network or server problems during mount may cause partially
4376 mounted clients instead of returning an error.
4377 Details : The config llog parsing code may overwrite the error return
4378 code during mount error handling, returning success instead
4384 Description: MDS can fail to allocate large reply buffers
4385 Details : After long uptimes the MDS can fail to allocate large reply
4386 buffers (e.g. zconf client mount config records) due to memory
4387 fragmentation or consumption by the buffer cache. Preallocate
4388 some large reply buffers so that these replies can be sent even
4389 under memory pressure.
4393 Frequency : rare (liblustre)
4394 Description: fsx running with liblustre complained that using truncate() to
4395 extend the file doesn't work. This patch corrects that issue.
4396 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4397 ATTR_SIZE and lsm use in llu_setattr_raw.
4401 Frequency : rare, only 2.6 kernels
4402 Description: Unusual file access patterns on the MDS may result in inode
4403 data being lost in very rare circumstances.
4404 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4405 code in the 2.6 kernel caused the inode and/or EA data not to
4406 be read from disk, causing single-file corruption.
4410 Frequency : rare, only 2.6 filesystems using extents
4411 Description: Heavy concurrent write and delete load may cause data corruption.
4412 Details : It was possible under high-load situations to have an extent
4413 metadata block in the block device cache from a just-unlinked
4414 file overwrite a newly-allocated data block. We now unmap any
4415 metadata buffers that alias just-allocated data blocks.
4419 Frequency : filesystems with default stripe_count larger than 77
4420 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4421 Details : lconf specifies an inode size of 4096 bytes when the default
4422 stripe_count is larger than 77. This conflicts with the default
4423 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4424 this case to avoid pinning too much memory for large EAs.
4426 ------------------------------------------------------------------------------
4428 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4433 Frequency : rare (extremely heavy IO load with hundreds of clients)
4435 Description: Client is evicted, gets IO error writing to file
4436 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4437 caused clients to be evicted for AST timeouts. The fixes in
4438 bug 5192 mean we no longer need to have such short AST timeouts
4439 so ldlm_timeout has been increased.
4442 Frequency : occasional during --force or --failover shutdown under load
4443 Bugzilla : 5949, 4834
4444 Description: Server oops/LBUG if stopped with --force or --failover under load
4445 Details : a collection of import/export refcount and cleanup ordering
4446 issues fixed for safer force cleanup
4449 Frequency : only filesystems larger than 120 OSTs
4450 Bugzilla : 5990, 6223
4451 Description: lfs getstripe would oops on a very large filesystem
4452 Details : lov_getconfig used kfree on vmalloc'd memory
4455 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4456 Bugzilla : 6242, 6243
4457 Description: reading from files that had been truncated to a non-zero size
4458 but never opened returned no data
4459 Details : ll_file_read() reads zeros from no-object files to EOF
4464 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4465 Details : There's little harm in aborting MDS/OSS recovery and letting it
4466 try again, so I removed the LASSERT and return an error instead.
4468 Severity : enhancement
4470 Description: New debugging infrastructure for tracking down data corruption
4471 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4472 (b) cover more of the client-side code path, and (c) try to narrow
4473 down where problems occurred
4477 Bugzilla : 3819, 4364, 4397, 6313
4478 Description: Racing close and eviction MDS could cause assertion in mds_close
4479 Details : It was possible to get multiple mfd references during close and
4480 client eviction, leading to one thread referencing a freed mfd.
4482 Severity: : enhancement
4483 Bugzilla : 3262, 6359
4484 Description: Attempts to reconnect to servers are now more aggressive.
4485 Details : This builds on the enhanced upcall-less recovery that was added
4486 in 1.4.2. When trying to reconnect to servers, clients will
4487 now try each server in the failover group every 10 seconds. By
4488 default, clients would previously try one server every 25 seconds.
4493 Description: After recovery, certain operations trigger a failed
4494 assertion on a client.
4495 Details : Failing over an mds, using lconf -d --failover, while a
4496 client was doing a readdir() call would cause the client to
4497 LBUG after recovery completed and the readdir() was resent.
4499 Severity : enhancement
4501 Description: Default groups are now added by lconf
4502 Details : You can now run lconf --group <servicename> without having to
4503 manually add groups with lmc.
4506 Frequency : occasional
4508 Description: Nodes with an elan id of 0 trigger a failed assertion
4511 Frequency : always when accessing e.g. tty/console device nodes
4513 Description: tty and some other devices nodes cannot be used on lustre
4514 Details : file's private_data field is used by device data and lustre
4515 values in there got lost. New field was added to struct file to
4516 store fs-specific private data.
4519 Frequency : when exporting Lustre via NFS
4521 Description: NFSD failed occasionally when looking up a path component
4522 Details : NFSD is looking up ".." which was broken in ext3 directories
4523 that had grown large enough to become hashed.
4526 Frequency : Clusters with multiple interfaces not on the same subnet
4528 Description: Nodes will repeatedly try to reconnect to an interface which it
4529 cannot reach and report an error to the log.
4530 Details : Extra peer list entries will be created by lconf with some peers
4531 unreachable. lconf now validates the peer before adding it.
4534 Frequency : Only if a default stripe is set on the filesystem root.
4536 Description: Setting a default stripe on the filesystem root prevented the
4537 filesystem from being remounted.
4538 Details : The client was sending extra request flags in the root getattr
4539 request and did not allocate a reply buffer for the dir EA.
4542 Frequency : occasional, higher if lots of files are accessed by one client
4543 Bugzilla : 6159, 6097
4544 Description: Client trips assertion regarding lsm mismatch/magic
4545 Details : While revalidating inodes the VFS looks up inodes with ifind()
4546 and in rare cases can find an inode that is being freed.
4547 The ll_test_inode() code will free the lsm during ifind()
4548 when it finds an existing inode and then the VFS later attaches
4549 this free lsm to a new inode.
4553 Bugzilla : 6422, 7030
4554 Description: MDS deadlock between mkdir and client eviction
4555 Details : Creating a new file via mkdir or mknod (starting a transaction
4556 and getting the ns lock) can deadlock with client eviction
4557 (gets ns lock and trying to finish a synchronous transaction).
4560 Frequency : occasional
4561 Description: While starting a server, the fsfilt_ext3 module could not be
4563 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4564 kernels. Previously, lconf would still use the ext3 name
4565 when trying to load modules. Now, it will correctly use
4566 ext3 on 2.4 and ldiskfs on 2.6.
4568 Severity : enhancement
4569 Description: The default stripe count has been changed to 1
4570 Details : The interpretation of the default stripe count (0, to lfs
4571 or lmc) has been changed to mean striping across a single
4572 OST, rather than all available. For general usage we have
4573 found a stripe count of 1 or 2 works best.
4575 Severity : enhancement
4576 Description: Add support for compiling against Cray portals.
4577 Details : Conditional compiling for some areas that are different
4581 Frequency : occasional
4582 Bugzilla : 6409, 6834
4583 Description: Creating files with an explicit stripe count may lead to
4584 a failed assertion on the MDS
4585 Details : If some OSTs are full or unavailable, creating files may
4586 trigger a failed assertion on the MDS. Now, Lustre will
4587 try to use other servers or return an error to the
4591 Frequency : occasional
4593 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4594 caused lock timeout during readahead (since 1.4.2).
4595 Details : Processes doing readahead might match a lock that hasn't been
4596 granted yet if there are overlapping and conflicting lock
4597 requests. The readahead process waits on ungranted lock
4598 (original lock is CBPENDING), while OST waits for that process
4599 to cancel CBPENDING read lock and eventually evicts client.
4601 Severity : enhancement
4603 Description: Initial enabling of flock support for clients
4604 Details : Implements fcntl advisory locking and file status functions.
4605 This feature is provided as an optional mount flag (default
4606 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4607 locking are implemented yet, and those that are are not guaranteed
4608 to be completely correct in production environments.
4609 mount -t lustre -o [flock|noflock] ...
4612 Frequency : occasional
4614 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4615 assertion in the ext3 JBD (journaling) layer.
4616 Details : The b_committed_data struct is protected by the big kernel lock
4617 in 2.4 kernels, serializing journal_commit_transaction() and
4618 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4619 access to this struct. In 2.6 kernels there is finer grained
4620 locking to improve SMP performance of the JBD layer.
4624 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4629 Description: merge and cleanup kernel patches.
4632 -----------------------------------------------------------------------------
4634 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4636 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4637 versions of Lustre. Please update all servers and clients to
4638 version 1.4.2 or later at the same time. You must also run
4639 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4640 to update the configuration logs.
4642 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4643 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4644 - swab llog records of type '0' so we get proper header size/idx (5861)
4645 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4646 - fix rename of one directory over another leaking an inode (5953)
4647 - avoid SetPageDirty on 2.6 (5981)
4648 - don't re-add just-being-destroyed locks to the waiting list (5653)
4649 - when creating new directories, inherit the parent's custom
4650 striping settings if present parent (3048)
4651 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4652 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4653 - don't hold client locks on temporary worklist from l_lru (5666)
4654 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4655 - automatically evict dead clients (5921)
4656 - Update file size properly in create+truncate+fstat case (6196)
4657 - Do not unhash mountpoint dentries, do not allow removal of
4659 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4660 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4661 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4662 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4663 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4664 - Correctly handle reads of files with no objects (6243)
4665 - lctl recover will also mark a device active if deactivate used (5933)
4667 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4668 - allow --write-conf on an MDS with different nettype than client (5619)
4669 - don't write config llogs to MDS for mounts not from that MDS (5617)
4670 - lconf should create multiple TCP connections from a client (5201)
4671 - init scripts are now turned off by default; run chkconfig --on
4672 lustre and chkconfig --on lustrefs to use them
4673 - upcalls are no longer needed for clients to recover to failover
4675 - add --abort-recovery option to lconf to abort recovery on device
4677 - add support for an arbitrary number of OSTs (3026)
4678 - Quota support protocol changes.
4679 - forward compatibility changes to wire structs (6007)
4680 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4681 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4682 - improved "lustre status" script
4683 - initialize blocksize for non-regular files (6062)
4684 - added --disable-server and --disable-client configure options (5782)
4685 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4686 - Vanilla 2.4.29 support
4687 - increase maximum number of obd devices to 520 (6242)
4688 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4689 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4690 SLES 9 2.6.5 kernels
4691 - limit stripes per file to 160 (the maximum EA size) (6093)
4693 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4696 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4697 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4698 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4699 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4700 - small ext3 extents cleanups and fixes (5733)
4701 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4702 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4703 - avoid needless client->OST connect, fix handle mismatch (5317)
4704 - fix DLM error path that led to out-of-sync client, long delays (5779)
4705 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4706 - fix several locking issues related to i_size (5492,5624,5654,5672)
4707 - don't move pending lock onto export if it is already evicted (5683)
4708 - fix kernel oops when creating .foo in unlinked directory (5548)
4709 - fix deadlock in obdfilter statistics vs. object create (5811)
4710 - use time_{before,after} to avoid timer jiffies wrap (5882)
4711 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4712 - Do not leak request if server was not able to process it (5154)
4713 - If mds_open unable to find parent dir, make that negative lookup(5154)
4714 - don't create new directories with extent-mapping (5909, 5936)
4716 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4717 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4718 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4719 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4720 - support for RHEL4 kernel on i686 (5773)
4721 - provide error messages when incompatible logs are encountered (5898)
4723 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4724 * version 1.4.0.10 (1.4.1 release candidate 1)
4726 - don't keep a lock reference when lock is not granted (4238)
4727 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4728 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4729 - drop import inflight refcount on signal_completed_replay error (5255)
4730 - unlock page after async write error during send (3677)
4731 - handle missing objects in filter_preprw_read properly (5265)
4732 - no transno return for symlink open, don't save no-trasno open (3440)
4733 - don't try to complete elan receive that already failed (4012)
4734 - free RPC server reply state on error (5406)
4735 - clean up thread from ptlrpc_start_thread() on error (5160)
4736 - readahead could read extra page into cache that wasn't ejected (5388)
4737 - prevent races in class_attach/setup/cleanup/detach (5260)
4738 - don't dereference de->d_inode after l_dput of de (5458)
4739 - use "int" for stripe value returned from lock_to_stripe (5544)
4740 - mballoc allocation and error-checking fixes in 2.6 (5504)
4741 - block device patches to fix I/O request sizes in 2.6 (5482)
4742 - look up hostnames for IB nals (5602)
4743 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4744 - don't start multiple acceptors for the same port (5277)
4745 - fix incorrect LASSERT in mds_getattr_name (5635)
4746 - export a proc file for general "ping" checking (5628)
4747 - fix "lfs check" to not block when the MDS is down (5628)
4749 - service request history (4965)
4750 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4751 - create an "evict_client" /proc entry on OSTs, like the MDS has
4752 - fix mount usage message, return errors per mount(8) (5168)
4753 - change grep [] to grep "[]" in tests so they work in more UMLs
4754 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4755 - remove ancient LOV_MAGIC_V0 EA support (5047)
4756 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4757 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4758 - allow readahead allocations to fail when low on memory (5383)
4759 - mmap locking landed again, after considerable improvement (2828)
4760 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4762 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4765 - send OST transaction number in read/write reply to free req (4966)
4766 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4767 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4768 - the watchdog thread now runs as interruptible (5246)
4769 - flock/lockf fixes (but it's still disabled, pending 5135)
4770 - don't use EXT3 constants in llite code (5094)
4771 - memory shortage at startup could cause assertion (5176)
4773 - reorganization of lov code
4774 - single portals codebase
4776 - add extents/mballoc support (5025)
4777 - direct I/O reads in the obdfilter (4048)
4778 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4780 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4783 - send OST transaction number in read/write reply to free req (4966)
4784 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4785 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4786 - the watchdog thread now runs as interruptible (5246)
4787 - handle missing objects in filter_preprw_read properly (5265)
4788 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4789 - drop import inflight refcount on signal_completed_replay error (5255)
4790 - unlock page after async write error during send (3677)
4791 - return original error code on reconstructed replies (3761)
4792 - no transno return for symlink open, don't save no-trasno open (3440)
4794 - add pid to ldlm debugging output (4922)
4795 - bump the watchdog timeouts -- we can't handle 30sec yet
4796 - extra debugging for orphan dentry/inode bug (5259)
4798 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4801 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4802 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4803 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4804 - fix NULL dereference in /proc/sys/portals/routes (4827)
4805 - allow failed mdc_close() operations to be interrupted (4561)
4806 - stop precreate on OST before MDS would time out on it (4778)
4807 - don't send partial-page writes before EOF from client (4410)
4808 - discard client grant for sub-page writes on large-page clients (4520)
4809 - don't free dentries not owned by NFS code, check generation (4806)
4810 - fix lsm leak if mds_create_objects() fails (4801)
4811 - limit debug_daemon file size, always print CERROR messages (4789)
4812 - use transno after validating reply (3892)
4813 - process timed out requests if import state changes (3754)
4814 - update mtime on OST during writes, return in glimpse (4829)
4815 - add mkfsoptions to LDAP (4679)
4816 - use ->max_readahead method instead of zapping global ra (5039)
4817 - don't interrupt __l_wait_event() during strace
4819 - add software watchdogs to catch hung threads quickly (4941)
4820 - make lustrefs init script start after nfs is mounted
4821 - fix CWARN/ERROR duplication (4930)
4822 - return async write errors to application if possible (2248)
4823 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4824 - print NAL number in %x format (4645)
4825 - update barely-supported suse-2.4.21-171 series (4842)
4826 - support for sles 9 %post scripts
4827 - support for building 2.6 kernel-source packages
4828 - support for sles km_* packages
4830 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4833 - ignore -ENOENT errors in osc_destroy (3639)
4834 - notify osc create thread that OSC is being cleaned up (4600)
4835 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4836 - reconstruct ost_handle() like mds_handle() (4657)
4837 - create a new thread to do import eviction to avoid deadlock (3969)
4838 - let lconf resolve symlinked-to devices (4629)
4839 - don't unlink "objects" from directory with default EA (4554)
4840 - hold socknal file ref over connect in case target is down (4394)
4841 - allow more than 32000 subdirectories in a single directory (3244)
4842 - fix blocks count for O_DIRECT writes (3751)
4843 - OST returns ENOSPC from object create when no space left (4539)
4844 - don't send truncate RPC if file size isn't changing (4410)
4845 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4846 - bind to privileged port in socknal and tcpnal (3689)
4848 - rate limit CERROR/CWARN console message to avoid overload (4519)
4849 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4850 - basic mmap support (3918)
4851 - kernel patch series update from b1_4 (4711)
4853 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4856 - avoid crash during MDS cleanup with OST shut down (2775)
4857 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4858 - don't use bad inodes on the MDS (3744)
4859 - dynamic object preallocation to improve recovery speed (4236)
4860 - don't hold spinlock over lock dumping or change debug flags (4401)
4861 - don't zero obd_dev when it is force cleaned (3651)
4862 - print grants to console if they go negative (4431)
4863 - "lctl deactivate" will stop automatic recovery attempts (3406)
4864 - look for existing locks in ldlm_handle_enqueue() (3764)
4865 - don't resolve lock handle twice in recovery avoiding race (4401)
4866 - revalidate should check working dir is a directory (4134)
4868 - don't always mark "slow" obdfilter messages as errors (4418)
4870 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4873 - don't close LustreDB during write_conf until it is done (3860)
4874 - fix typo in lconf for_each_profile (3821)
4875 - allow dumping logs from multiple threads at one time (3820)
4876 - don't allow multiple threads in OSC recovery (3812)
4877 - fix debug_size parameters (3864)
4878 - fix mds_postrecov to initialize import for llog ctxt (3121)
4879 - replace config semaphore with spinlock (3306)
4880 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4881 - don't allow enqueue to complete on a destroyed export (3822)
4882 - down write_lock before checking llog header bitmap (3825)
4883 - recover from lock replay timeout (3764)
4884 - up llog sem before sending rpc (3652)
4885 - reduce ns lock hold times when setting kms (3267)
4886 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4887 - fix NULL deref and obd_dev leak on setup error (3312)
4888 - replace some LBUG about llog ops with error handling (3841)
4889 - don't match INVALID dentries from d_lookup and spin (3784)
4890 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4891 - fix invalid assertion in ptlrpc_set_wait (3880)
4893 - add libwrap support for the TCP acceptor (3996)
4894 - add /proc/sys/portals/routes for non-root route listing (3994)
4895 - allow setting MDS UUID in .xml (2580)
4896 - print the stack of a process that LBUGs (4228)
4898 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4901 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4902 - make sure to unset replay flag from failed open requests (3440)
4903 - if default stripe count is 0, use OST count for inode size (3636)
4904 - update parent mtime/ctime on client for create/unlink (2611)
4905 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4906 - free recovery state on server during a forced cleanup (3571)
4907 - unregister_reply for resent reqs (3063)
4908 - loop back devices mounting and status check on 2.6 (3563)
4909 - fix resource-creation race that can provoke i_size == 0 (3513)
4910 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4911 - more debugging for page-accounting assertion (3746)
4912 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4913 - avoid deadlock after precreation failure (3758)
4914 - fix race and lock order deadlock in orphan handling (3450, 3750)
4915 - add validity checks when grabbing inodes from l_ast_data (3599)
4917 - add /proc/.../recovery_status to obdfilter (3428)
4918 - lightweight CDEBUG infrastructure, debug daemon (3668)
4919 - change default OSC RPC parameters to be better on small clusters
4920 - turn off OST read cache for files smaller than 32MB
4921 - install man pages and include them in rpms (3100)
4922 - add new init script for (un)mounting lustre filesystems (2593)
4923 - run chkconfig in %post for init scripts (3701)
4924 - drop scimac NAL (unmaintained)
4926 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4929 - clean kiobufs before and after use (3485)
4930 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4931 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4932 - change init script to fail more clearly if not run as root (1528)
4933 - allow clients to reconnect during replay (1742)
4934 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4935 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4936 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4937 - fix cond_resched (3554)
4938 - only evict unfinished clients after recovery (3515)
4939 - allow bulk resend, prevent data loss (3570)
4940 - dynamic ptlrpc request buffer allocation (2102)
4941 - don't allow unlinking open directory if it isn't empty (2904)
4942 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4943 - remove extraneous obd dereference causing LASSERT failure (3334)
4944 - don't use get_cycles() when creating temp. files on the mds (3156)
4945 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4946 - handle EEXIST for set-stripe, set proper directory name (3336)
4948 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4949 - fix ksocknal_fmb_callback() error messages (2918)
4951 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4954 - don't copy lvb into (possibly NULL) reply on error (2983)
4955 - don't deref dentry after dput, don't free lvb on error (2922)
4956 - use the kms to determine writeback rpc length (2947)
4957 - increment oti_logcookies when osc is inactive (2948)
4958 - update client's i_blocks count via lvb messages (2543)
4959 - handle intent open/close of special files properly (1557)
4960 - mount MDS with errors=remount-ro, like obdfilter (2009)
4961 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4962 - don't use cancelling-locks' kms values (2947)
4963 - use highest lock extent for kms, not last one (2925)
4964 - don't dereference ERR_PTR() dentry in error handling path (3107)
4965 - fix thread race in portals_debug_dumplog() (3122)
4966 - create lprocfs device entries at setup instead of at attach (1519)
4967 - common AST error handler, don't evict client on completion race (3145)
4968 - zero nameidata in detach_mnt in 2.6 (3118)
4969 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4970 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4971 - initialize RPC timeout timer earlier for 2.6 (3219)
4972 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4973 - print nal/nid for unknown nid (3258)
4974 - additional checks for oscc recovery before doing precreate (3284)
4975 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4976 - don't crash in mdc_close for bad permissions on open (3285)
4977 - zero i_rdev for non-device files (3147)
4978 - clear page->private before handing to FS, better assertion (3119)
4979 - tune the read pipeline (3236)
4980 - fix incorrect decref of invalidated dentry (2350)
4981 - provide read-ahead stats and refine rpc in flight stats (3328)
4982 - don't hold journal transaction open across create RPC (3313)
4983 - update atime on MDS at close time (3265)
4984 - close LDAP connection when recovering to avoid server load (3315)
4985 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4986 - don't leak open file on MDS after open resend (3325)
4987 - serialize filter_precreate and filter_destroy_precreated (3329)
4988 - loop device shouldn't call sync_dev() for nul device (3092)
4989 - clear page cache after eviction (2766)
4990 - resynchronize MDS->OST in background (2824)
4991 - refuse to mount the same filesystem twice on same mountpoint (3394)
4992 - allow llmount to create routes for mounting behind routers (3320)
4993 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4994 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4995 - fix rare problem with rename on htree directories (3417)
4997 - allow default OST striping configuration per directory (1414)
4998 - fix compilation for qswnal for 2.6 kernels (3125)
4999 - increase maximum number of MDS request buffers for large systems
5000 - change liblustreapi to be useful for external progs like lfsck (3098)
5001 - increase local configuration timeout for slow disks (3353)
5002 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5004 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5007 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5008 - don't overwrite extent policy data in reply if lock was blocked (2901)
5009 - drop filter export grants atomically with removal from device (2663)
5010 - del obd_self_export from work_list in class_disconnect_exports (2908)
5011 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5012 - swab reply message in mdc_close, other PPC fixes (2464)
5013 - fix destroying of named logs (2325)
5014 - overwrite old logs when running lconf --write_conf (2264)
5015 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5016 - fix race in target_handle_connect (2898)
5017 - mds_reint_create() should take same inode create lock (2926)
5018 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5019 - don't close files for self_export to avoid uninitialized obd (2936)
5020 - allow MDS with the same name as client node (2939)
5021 - hold dentry reference for closed log files for unlink (2325)
5022 - reserve space for all logs during transactions (2059)
5023 - don't evict page beyond end of stripe extent (2925)
5024 - don't oops on a deleted current working directory (2399)
5025 - handle hard links to targets without a parent properly (2517)
5026 - don't dereference NULL lock when racing during eviction (2867)
5027 - don't grow lock extents when lots of conflicting locks (2919)
5029 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5032 - account for cache space usage on clients to avoid data loss (974)
5033 - lfsck support in lustre kernel code (2349)
5034 - reduce journal credits needed for BRW writes (2370)
5035 - orphan handling to avoid losing space on client/server crashes
5036 - ptlrpcd can be blocked, stopping ALL progress (2477)
5037 - use lock value blocks to assist in proper KMS, faster stat (1021)
5038 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5039 - recovery for initial connections (2355)
5040 - fixes for mds_cleanup_orphans (1934)
5041 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5042 - block all file creations until orphan recovery completes (1901)
5043 - client remove rq_connection from request struct (2423)
5044 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5045 - recovery timer race (2670)
5046 - mdc_close recovey bug (2532)
5047 - ptlrpc cleanup bug (2710)
5048 - mds timeout on local locks (2588)
5049 - namespace lock held during RPCs (2431)
5050 - handle interrupted sync write properly (2503)
5051 - don't try to handle a message that hasn't been replied to (2699)
5052 - client assert failure during cleanup after abort recovery (2701)
5053 - leak mdc device after failed mount (2712)
5054 - ptlrpc_check_set allows timedout requests to complete (2714)
5055 - wait for inflight reqs when ptlrpcd finishes (2710)
5056 - make sure unregistered services are removed from the srv_list
5057 - reset bulk XID's when resending them (caught by 1138 test)
5058 - unregister_bulk after timeout
5059 - fix lconf error (2694)
5060 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5061 - readahead locks pages, leaves pending causing memory pressure (2673)
5062 - increase OST request buffers to 4096 on large machines (2729)
5063 - fix up permission of existing directories in simple_mkdir (2661)
5064 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5065 - don't assign transno to errored transactions (2742)
5066 - don't delete objects on OST if given a bogus objid from MDS (2751)
5067 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5068 - if rq_no_resend, then timeout request after recovery (2432)
5069 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5070 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5071 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5072 - MDS recovery shouldn't delete objects that it has given out (2730)
5073 - if enqueue arrives after completion, don't clobber LVB (2819)
5074 - don't unlock pages twice when trigger_group_io returns error (2814)
5075 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5076 - don't write pages to disk if there was an error (1450)
5077 - don't ping imports that have recovery disabled (2676)
5078 - take buffered bytes into account when balancing socknal conn (2817)
5079 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5080 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5081 - remove little-endian swabbing of llog records (1987)
5082 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5083 - retry reposting request buffers if they fail (1191)
5084 - grow extent at grant time to avoid granting a revoked lock (2809)
5085 - lock revoke doesn't evict page if covered by a second lock (2765)
5086 - disable VM readahead to avoid reading outside lock extents (2805)
5088 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5089 - updated kernel patches for hp-2.4.20 kernel (2681)
5091 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5094 - fix truncated write corruption (2366)
5095 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5097 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5098 - ensure len(uuid) < 37 in lmc (1171)
5099 - fix ia64 OOPS in llog_test (2255)
5100 - zero end of page at obdfilter for partial page writes (2648)
5101 - don't leave stale dentries around after renames (bug 2428)
5102 - fix timeouts when evicting a client with a single lock held (2642)
5103 - set deadline for the initial HELLO message to drain (2634)
5104 - print out dotted-quad IP addresses in the socknal (2302)
5106 - additional debugging for MDS client eviction problem (2443)
5107 - fix mkfsoptions support for osts (2603, 2604)
5109 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5112 - add series for the vanilla 2.6.0 kernel
5113 - add series for the vanilla 2.4.24 kernel
5114 - add series for a cray x86/64 UL kernel drop
5115 - fix xattr patches for the vanilla 2.4.19 series
5117 - generate true UUIDs in lmc (1171)
5118 - have portals stack dumping break in UML (2466)
5119 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5120 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5121 - wake callback waiting threads on client eviction (2460)
5122 - Add --ptldebug and --subsystem to lmc (1719)
5123 - update assertion to allow safe interrupt allocation
5124 - set rq_no_resend for cancel requests (2432)
5125 - recalculate ptlrpcd timeout after resend (2494)
5126 - call vfs_rmdir when removing pending directories (2368)
5127 - fix renaming a file to itself (2429)
5128 - lmc creates a default one-stripe lov (2454)
5129 - expand procfs space to handle large clusters (2326)
5130 - increase UML stack to avoid overflow
5131 - update lconf's list of debug and subsystem masks
5132 - fix lfs find --obd (2510)
5133 - /proc tunable for disabling filter read caching (2591)
5134 - stop rpm packages from altering slapd.conf (2301)
5135 - disable nagle in the socknal under 0conf (2578)
5136 - choose mds inode size based on stripe count (2572)
5137 - fix kernel-source rpm problems (2516)
5139 - add --disable-doc to avoid pdf generation (2421)
5140 - update documentation, tests, type-os, comments
5141 - avoid format warnings on ia64
5142 - remove the TOE NAL
5143 - tiny code cleanups by removing unused fields
5145 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5148 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5149 - remove the most blatant lies from BUILDING file (2371)
5150 - change default debug level to reasonable production setting
5151 - reduce client side cache size to reduce cache flush time
5152 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5153 - make TCP zerocopy and pinger support enabled by default (2476)
5154 - sync writes completed after process exits caused crashes (2319)
5155 - maintain correct mount count on the MDS (2356)
5156 - backout 1557, because 2316 wasn't really fixed
5157 - better file I/O statistics gathering in /proc
5158 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5159 - another kernel patch to fix zero-copy TCP function export
5160 - don't take duplicate lock when processing re-sent getattr (2420)
5161 - lctl uses obd_self_export instead of creating new conn (2353)
5162 - MDS/OST recovery case which requires object creation asserted (2425)
5163 - move lfs from /usr/sbin to /usr/bin in packages
5164 - fix race between mds_client_add and mds_client_free (2417)
5165 - use kmalloc instead of slabs in portals (2430)
5166 - don't create duplicate records when a failover MDS is present (2442)
5167 - remove unnecessary mount age check (2332)
5168 - don't remove directory inodes from locks prematurely (2451)
5169 - don't break if MDS service name is the same as hostname (2103)
5170 - fix races in client write RPC generation when cache full (2482)
5172 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5175 - remove now-unused request->rq_obd (278)
5176 - if an allocation fails, print out how much memory we've used (1933)
5177 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5178 - add the "configurable stack size" patch to most series files (1256)
5179 - ability to write large log records, for 100+ OST configs (2306)
5180 - fix NULL deref when filter_prep fails (2314)
5181 - fix operator precedence error in filter_sync
5182 - dynamic allocation of socknal TX descriptors (2315)
5183 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5184 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5185 - ensure that CWARN messages in Portals always get to the syslog
5186 - __init/__exit are not for prototype decls (ldlm_init/exit)
5187 - x86-64 compile warning fixes
5188 - fix gateway LMC keyword conflict (2318)
5189 - fix MDS lock inversions in getattr/reint paths (1844)
5190 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5191 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5192 - DLM race condition prevented some lock evictions (2328)
5193 - ENOMEM detection and retry on socknal sends (2230)
5194 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5195 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5196 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5197 - better error messages when a client is rejected during recovery (1505)
5198 - avoid cancelling locks which were never granted, after failure (2330)
5199 - fix i_sem/journal inversion in mds_client_add (2333)
5200 - fix truncate/getattr lock cycle deadlock (2334)
5201 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5202 - fix two rare exit paths which could leak an l_lock() ref (2321)
5203 - fencepost error in MDS/OST orphan recovery (2226)
5204 - make log record alignment 8 bytes (1988)
5205 - lstripe now fails when requested offset > ost_count (2237)
5206 - ensure that all kernel series have a complete list.h (1607)
5207 - fix crashes in special-file operations (2316)
5208 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5210 - allow configurable automake binary, for testing new versions
5211 - small update to the lfs documentation
5213 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5215 * fix negative export reference count in fsfilt_sync (2312)
5217 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5218 * release candidate 0.9.1
5220 - orphans are moved into the PENDING directory for possible recovery
5221 - replayed opens now open by fid for orphan/rename safety (1042)
5222 - last close of an orphan inode generates a transno (683)
5223 - chdir() and mount() now pin the directory entry (1020)
5224 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5225 - discard very old requests without processing them (1502)
5226 - remove some common, well-understood CERRORs (1505)
5227 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5228 - clear "grant" flags in OST replies until OST grant code lands (1644)
5229 - fix read performance by not clobbering i_blksize on client (1598)
5230 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5231 - make LRU size a /proc tunable, clears locks when reduced (707)
5232 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5233 - server locks take references on exports now (1558)
5234 - build fixes for 2.4.20-rh trees (1663)
5235 - return an error from lov_create if all OSCs are inactive (1751)
5236 - fix import levels when a reconnect happens without a timeout (1597)
5237 - exit early from mds_open if we get a lookup error (1749)
5238 - partial page read at EOF wouldn't wait for disk before sending (1642)
5239 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5240 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5241 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5242 - write/truncate lock inversion (1639)
5243 - Don't auto-load obdclass, portals modules during cleanup (1495)
5244 - fix timestamps from jumping to "now" (1763)
5245 - extra journal assertions (1648)
5246 - add an extra multiunlink test (1771)
5247 - fix read_record/write_record API (1776)
5248 - fix leak of offset_extent, possible incorrect i_size later (1772)
5249 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5250 - Debugging for the kqswnal_get_idle_tx problems (1820)
5251 - Allow recovery to be attempted multiple times (1536)
5252 - Write out MDS last_rcvd file after it is first created (1600)
5253 - Fix tx_descriptor leak in failed transmit situations (1827)
5254 - ext3 journaling fixes for assertion failure after IO error (1871)
5255 - class_export_put() on freed export after completion AST error (1896)
5256 - Fix revalidate looping in VFS (1322)
5257 - Don't access a freed export during MDS_REINT timeout (1521)
5258 - Add open-unlink recovery support on the MDS (1673,1764)
5259 - Return an error if no MDS data was read from last_rcvd (1946)
5260 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5261 - Don't setup a disk device that doesn't match exported UUID (317)
5262 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5263 - avoid committing NULL handle in force close
5264 - local.sh is now a one-stripe LOV configuration
5265 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5266 - fix ext3 htree duplicate directory entry corruption (1516)
5267 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5268 - update to new LOV EA format (2097)
5269 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5271 2003-06-15 Phil Schwan <phil@clusterfs.com>
5274 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5275 - per-import/export recovery handling (958, 931, 959)
5276 - multiple last-rcvd slots, for serving multiple FSes (949)
5277 - connections are again shared between multiple imp/exports (963, 964)
5278 - "umount -f" would hang if any requests needed to be sent (393, 978)
5279 - avoid pinning large req buffer by copying for queued messages (989)
5280 - add "uuid" to "lctl device" command to help upcalls (991)
5281 - "open" RPCs with transnos would confuse recovery counters (1037)
5282 - do proper endian conversion of all wire messages (288, 340, 891)
5283 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5284 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5285 - fix ptlrpc_connection leak in target_handle_connect (1174)
5286 - fix import refcounting bug in OST and MDS cleanup (1134)
5287 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5288 - fix very unlikely obd_types race condition (501)
5289 - remove osc_open hack for echo_client (1187)
5290 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5291 - a failure in read_inode2 leads to deadlock (1139)
5292 - cancel ack-locks as soon as transaction is committed (1072)
5293 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5294 - make sure to commitrw after any preprw to avoid deadlock (1162)
5295 - failing to execute a file in a lustre FS would lock inode (1203)
5296 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5297 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5298 - obd_uuid2tgt would walk off the end of the list (1255)
5299 - on IA64 the osc would give portals incorrect bulk size (1258)
5300 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5301 - fix lock inversion caused by new llite matching code (1282)
5302 - limit the number of dirty pages on a client to 10MB (1286)
5303 - timed out locks were not being corrected cancelled (1289)
5304 - fix O_DIRECT above 4GB on IA-32 (1292)
5305 * major user-visible changes
5306 - fail out/fail over policy now controlled by the upcall (993)
5308 - add OBD_PING to check server availability and failure (954)
5309 - lustre messages are now sent in sending host order (288, 340, 891)
5310 - add eadatalen to MDS getattr reply (340)
5311 - OST read replies may contain second buffer, with per-page status (593)
5313 2003-03-11 Phil Schwan <phil@clusterfs.com>
5316 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5317 - failed open()s could cause deadlock; fixed (867, 869)
5318 - stop cancelling OST locks when files are closed (481)
5319 - overlapping XID spaces caused network corruption (851, 853)
5320 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5321 - setattr_raw added, to do single-RPC, server-side setattrs
5322 - lmc/lconf syntax change for OST UUIDs
5323 - fix crashy race condition between ptlrpc_free_req and osc_close
5324 - don't use request in mdc_enqueue if we hit a timeout (889)
5325 - don't set the inode i_size for regular files from the MDS (896)
5326 - handle out of order completion AST (842)
5327 - don't LBUG if a lock request times out after receiving AST (913)
5328 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5329 - if a bad lock AST arrives, send an error instead of dropping entirely
5330 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5331 - fix leak in bulk IO when only partially completed (899, 900, 926)
5332 - fix O_DIRECT for ia64 (55)
5333 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5334 - C-z after timeout could hang a process forever; fixed (977)
5336 - client-side I/O cache (678, 924, 929, 941, 970)
5338 - READPAGE and SETATTRs which don't take server-side locks get
5341 2003-02-11 Phil Schwan <phil@clusterfs.com>
5344 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5345 - Fix fsfilt_extN_readpage() to read a full page of directory
5346 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5347 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5348 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5349 - Most intent operations were removed, in favour of a new RPC mode
5350 that does a single RPC to the server and bypasses most of the VFS
5351 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5352 - Aggressively cancel local locks on DLM servers
5353 - mds_reint_unlink sends EA to the client if it's the last nlink.
5354 client uses that EA to unlink OST objects.
5355 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5356 - recursive symlinks were fixed (439)
5357 - fixed NULL deref in DEBUG_REQ
5358 - filter_update_lastobjid no longer calls sync, which annoyed extN
5359 - fixed multi-client small-writes to a single file problem (445)
5360 - fixed mtime updates during file writes (607)
5361 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5362 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5363 - fixed Linux OST size reporting problem (444, 656)
5364 - OST now updates object mtime with writes or setattr (607, 619)
5365 - client verifies file size before zeroing page past EOF (445)
5366 - OST now writes last allocated objid to disk with allocation (108)
5367 - LOV on echo now works (409)
5369 - mds_reint_unlink sends a new buffer, with the EA included. this
5370 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5371 set if a regular file was being unlinked, and it was the last link
5372 - use PtlGet from the target for bulk writes (315)
5373 - OST now updates object mtime with writes or setattr (607, 619)
5374 - LDLM now has a grant-time callback to revalidate locked items, if
5376 - Many MDS operations were reorganized to combat race conditions
5378 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5379 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5381 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5384 - Fully reactivate OST imports after reconnection (512, others)
5385 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5386 - More graceful error handling for truncating on dead OST (515)
5387 - Don't error out unless we're actually accessing dead stripes (474)
5388 - Fix garbage sizes when stripes are missing (410)
5389 - LRU counters were broken, causing constant lock purge (433, 432)
5390 - garbage on read from stripes with failed OSTs (441)
5391 - mark OSCs as active before reconnecting during recovery (438)
5392 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5393 - lfind did not preserve OST order in output (443)
5394 - symlinks cause hung clients, incorrect data (439)
5395 - stop dereferencing request after dropping refcount (457)
5396 - don't LASSERT(spin_is_locked) on non-SMP (455)
5397 - fixes for many rename() bugs
5398 - fstat didn't correctly synchronize attributes (399)
5399 - server must handle lock cancellation during blocking AST prep (487)
5400 - bulk descriptors were free()d too soon (511)
5401 - fix paths in lconf, which would load incorrect modules (451, 507)
5402 - fix confusing lconf 'host not found' error message (386)
5403 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5404 - fix race condition in mdc_blocking_ast() for inode access (526)
5405 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5406 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5407 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5408 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5409 - fix lstripe in several ways (kernel side) (527)
5410 - fix request leak in ldlm_cli_enqueue (262)
5411 - incorrect OSC was marked inactive after OST failure
5412 - call mds_fs_cleanup before unmounting filesystem (524)
5413 - fix races between taking ns_lock and ldlm_lock_change_resource
5414 - fix races updating LOV export open file list
5415 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5416 - fix recovery NULL deref in ldlm_cli_cancel_unused
5417 - fix some DLM races by using new hash table for lock handles (419)
5418 - permit the client to specify desired inodes, at replay
5419 - duplicate requests when we queue them for replay reintegration
5420 - fix last_rcvd offset calculation
5421 - sync after each recovered transaction, so we always make progress
5422 - never, not always, ERESTART requests without transnos
5423 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5425 - skip replay if the MDS says that the client is already connected
5426 - don't check for a recovery-enabled export to match lctl's UUID
5427 - don't INC_USE_COUNT for phantom exports
5428 - don't crash when cleaning up phantom exports (567)
5429 - don't double-finish or set replay data for errored mdc_open requests
5430 - abort requests when they time out, so we don't get old replies
5431 - send/receive replies for AST messages again
5432 - if the client says that it doesn't have the lock, cancel it on the
5434 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5435 mark it as destroyed, it will all work out when decref is called
5436 - fix module use counts (22, 581)
5438 - ASTs now expect a reply (server cancels lock on error reply)
5440 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5443 - fix many simultaneous client startup (392)
5444 - fix dentry->d_it clobbering
5445 - credentials weren't being shipped for readdir/getattr operations
5446 - remove invalid assertions triggered during some concurrent MD
5448 - proper Lustre versions added (336, 389)
5449 - fix memory leak for create error case (398)
5450 - fix LOV locking bug that would get cli/srv out of sync
5451 - fix echo client over LOV (409)
5452 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5453 - fix double-O_EXCL intent crash (424)
5454 - avoid sending multiple lock CANCELs (352)
5456 - MDS can do multi-client recovery (modulo bugs in new code)
5458 - many updates, edits, cleanups
5460 2002-11-18 Phil Schwan <phil@clusterfs.com>
5463 - fix null d_it dereference (346)
5464 - fix full OST/dbench hang (333)
5465 - fix permission problem with file removal (286)
5466 - fix removal of OSCs from LOV when they fail
5467 - fix NULL deref during bulk timeout (214)
5468 - fix problems related to multiple filesystems on one MDS (241)
5469 - fixed serious subtle metadata locking bugs
5470 - free locks on clients when inodes are removed due to memory
5472 - fix inode pointer in lock data (285)
5473 - partial support for multiple MDS on a single host (241)
5474 - data locks weren't cancelled at clear_inode time (290, 311)
5475 - intent locks could lead to unbounded lock growth (205)
5476 - added a maximum lock count, an LRU list, and a flusher
5477 - fix multiple rename (365)
5478 - properly abstracted the echo client
5479 - OSC locked 1 byte too many; fixed
5480 - rewrote brw callback code:
5481 - fixed recovery bugs related to LOVs (306)
5482 - fixed too-many-pages-in-one-write crash (191)
5483 - fixed (again) crash in sync_io_timeout (214)
5484 - probably fixed callback-related race (385)
5486 - Add capability to MDS protocol
5487 - LDLM cancellations and callbacks on different portals
5489 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5492 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5493 - defer open object destruction to close time (601981, 138)
5494 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5495 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5496 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5497 - don't call obd_setattr in ll_file_release for destroyed objects
5498 * protocol change to lustre_msg: move |version| and add |flags|
5499 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5500 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5501 * added replay of create, unlink, link and rename operations during
5502 MDS failover; recovery should be much more robust now
5503 * remove failed OSCs from LOVs (only lov_create uses this so far)
5504 * the lustre-HOWTO was brought (more) up to date (582544)
5506 2002-10-23 Phil Schwan <phil@clusterfs.com>
5509 - in-use dentries weren't being reused properly (617851)
5510 - prevent multiple LDLM setup (599178)
5511 - fix LOV size calculations for truncate (617853)
5512 - fix client handling of MDS intent errors (POSIX)
5513 - fix permission bug in lovstripe.c test (624321)
5514 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5515 - truncate past end of file could corrupt data
5516 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5517 - a race in recovery could return ETIMEDOUT to apps (623947)
5518 - building outside the source directory was fixed
5519 * the lustre-HOWTO was brought (more) up to date (582544)
5520 * major progress was made on recovery functionality
5522 2002-10-10 Phil Schwan <phil@clusterfs.com>
5525 - recovery deadlock fix
5526 - rm -rf causes LBUG fix (617817)
5527 - file open by multiple tasks fix (618962)
5528 - directory permissions bugs (602707 and 620007)
5529 - journal_stop fixed with locking (611313)
5530 - O_APPEND failures resolved (618273, perhaps 614459)
5531 - lconf PATH fix (619770)
5532 - IA64 build fix (621450)
5533 - RPC buffer sizes scale with amount of memory
5535 2002-10-01 Phil Schwan <phil@clusterfs.com>
5538 - locks would be cancelled without throwing away data pages,
5539 resulting in inconsistent data (605627)
5540 - inode attributes were not always being refreshed (605627, 612449)
5541 - lconf now continues to cleanup after lctl reports an error
5542 - MDS now enforces user permissions (602707)
5543 - lprocfs cleanup fixed, but not yet enabled (614157)
5544 - fixed infinite server hang, should a client not respond to an AST
5545 - avoid going into recovery if user calls readlink() with a buffer
5546 that's too small (613941)
5547 - AST RPCs no longer require replies (614867) -- this may be changed
5548 - don't crash server if client sends an IOV that's too big (611336)
5549 - fixed lock conversion deadlock (611892)
5550 - fixed the following of symlinks (614622)
5551 * recovery: the server can remove locks from a client that dies, other
5552 clients can make progress
5553 * more extN patch fixes
5554 * compile-time configurable ptlrpc buffer allocations
5556 - collaborative read cache document
5557 - Lustre Lite Performance CDR document-in-progress
5559 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5562 - fix typo in patch-2.4.18
5564 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5567 - clear ptlrpc request each time in handle_incoming_request()
5568 - unlink of files now destroys the object on the OST
5570 2002-09-19 Peter Braam <braam@clusterfs.com>
5572 * add hard link support
5573 * change obdfile creation method
5574 * kernel patch changed
5576 2002-09-19 Peter Braam <braam@clusterfs.com>
5579 - stack overflow bug in extN fixed
5581 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5583 * documentation updates
5584 - add man pages for config tools
5585 - update tests/README to describe testing with new config tools
5586 - finish metadata API descriptions
5587 * bug fixes and cleanups
5588 - statfs workaround for 16TB limit
5589 - LOV stripe allocation improved, can stripe on subset of OSTs
5590 - LOV file size/IO offset was wrong for files > 4GB in size
5591 - object EA data was being dropped, caused files to be unreadable
5592 - memory overflow with non-LOV OST caused memory corruption
5593 - fixed regression tests to work with new config tools, obdfilter
5594 - fixed bug when directory size became larger than 1 block
5595 - fixed bug (for single client case) when PWD was deleted
5596 - invalidate local directory pages when doing intent-based ops
5597 - avoid LDLM oops when lock callback contained bad data
5599 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5601 * documentation updates
5602 * bug fixes and cleanups
5603 - configuration tools
5606 - 64-bit compile warnings
5607 - 64-bit internal statfs data
5609 * test_brw on persistent OST devices
5611 * lprocfs (disabled)
5613 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5615 * documentation updates
5616 * bug fixes and cleanups
5617 * configuration tools
5619 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5622 * many small fixes to 0_5_4
5623 * io/network handling
5624 * thinkos in MDS operations
5626 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5629 * crucial basic fixes to 0.5.3
5630 * IOR, Iozone work over Elan
5633 2002-08-07 Phil Schwan <phil@clusterfs.com>
5634 * version 0_5_3, our first alpha
5635 * we use the new Portals iovs
5636 * documentation updates
5637 * bug fixes and cleanups
5638 * small changes in the DLM wire protocol
5640 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5641 * version 0_5_1 with some initial stability,
5642 * locking on MD and file I/O.
5643 * documentation updates
5644 * several bug fixes since 0.5.0
5645 * small changes in wire protocol
5647 2002-07-18 Phil Schwan <phil@clusterfs.com>
5649 * delivered as Lustre Light Alpha
5650 * fixed a crash after handling invalid MDS requests
5651 * fixed directory pages for architectures with non-4k pages sizes
5653 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5654 * release version v0_4_4
5655 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5657 * Fixes a number of interruption problems with OST operations.
5658 * Update documentation for portals header changes
5659 * Move all wire protocol structs/defines to lustre_idl.h
5660 * Fixes symlink length bug.
5661 * Add tcpdump to repository.
5663 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5664 * release version v0_4_3
5665 * Fixes statfs for inodes on extN.
5666 * Fixes bug in runtests which would delete /etc/hosts.
5667 * Use 64-bit object IDs wherever possible (not into VFS though)
5668 Remove ost_get_info, which is unused by lustre, and out of date.
5670 2002-07-03 Peter Braam <braam@clusterfs.com>
5671 * release version v0_4_2 Fixes a lookup error (type not passed)
5672 * move forward to head of Portals
5673 * move forward to latest Lustre kernel
5675 2002-06-25 Peter Braam <braam@clusterfs.com>
5676 * release version v0_4_1. Hopefully stable on single node use.