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 : rare, need acl's on inode.
19 Description: client can't handle ost additional correctly
20 Details : if ost was added after client connected to mds client can have
21 hit lnet_try_match_md ... to big messages to wide striped files.
22 in this case need teach client to handle config events about add
23 lov target and update client max ea size at that event.
25 Severity : enhancement
27 Description: Changelogs
28 Details : Changelogs are a lightweight mechanism to track filesystem
29 metadata and namespace changes. The changelog is recorded
30 permanently on the MDTs, and is periodically "consumed" / purged
31 when records are no longer needed.
33 Severity : enhancement
35 Description: compact fld format with extents
36 Details : Store range of seq rather than every seq in FLD. Seq
37 controller update FLD rather than clients. In Case of CMD, mdt0
38 has FLD, all other metadata server act as non persistent proxy
39 for FLD queries and cache fld entries in fld cache.
44 Description: don't skip ost target if they assigned to file
45 Details : Drop slow OSCs if we can, but not for requested start idx.
46 This means "if OSC is slow and it is not the requested
47 start OST, then it can be skipped, otherwise skip it only
48 if it is inactive/recovering/out-of-space.
52 Description: more cleanup in mds_lov
53 Details : not send LOV EA under replay, we can't know about they size at this
54 time. Don't allow client connect to mds before any ost connected,
55 for avoid problems with LOV EA size and returning EIO to client.
57 Severity : enhancement
59 Description: Interoperability at server side (Disk interoperability)
61 Severity : enhancement
63 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
65 Severity : enhancement
67 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
69 Severity : enhancement
71 Description: New client IO stack (CLIO).
73 Severity : enhancement
75 Description: Commit on sharing. Eliminate inter-client dependencies between
76 uncommitted transactions by doing transaction commits.
77 Thereby clients may recovery independently.
80 Frequency : Create a symlink file with a very long name
82 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
83 Details : If there is no extra space in the request for early cancels,
84 ldlm_req_handles_avail() returns 0 instead of a negative value.
86 Severity : enhancement
88 Description: Add /proc entry for import status
89 Details : The mdc, osc, and mgc import directories now have
90 an import directory that contains useful import data for debugging
93 Severity : enhancement
95 Description: Re-disable certain /proc logging
96 Details : Enable and disable client's offset_stats, extents_stats and
97 extents_stats_per_process stats logging on the fly.
100 Frequency : Only on FC kernels 2.6.22+
102 Description: oops in statahead
103 Details : Do not drop reference count for the dentry from VFS when lookup,
104 VFS will do that by itself.
106 Severity : enhancement
108 Description: Generic /proc file permissions
109 Details : Set /Proc file permissions in a more generic way to enable non-
110 root users operate on some /proc files.
114 Description: Hitting mdc_commit_close() ASSERTION
115 Details : Properly handle request reference release in
116 ll_release_openhandle().
120 Description: quota recovery deadlock during mds failover
121 Details : This patch includes att18982, att18236, att18237 in bz14840.
123 1. fix osts hang when mds does failover with quotaon
124 2. prevent watchdog storm when osts threads wait for the
129 Frequency : only patchless client
130 Description: add workaround for race between add/remove dentry from hash
132 Severity : enhancement
134 Description: Allow OST glimpses to return PW locks
138 Description: LBUG when llog conf file is full
139 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
143 Description: Prevent import from entering FULL state when server in recovery
147 Description: service mount cannot take device name with ":"
148 Details : Only when device name contains ":/" will mount treat it as
154 Description: replace ptlrpcd with the statahead thread to interpret the async
155 statahead RPC callback
159 Frequency : on recovery
160 Description: I/O failures after umount during fail back
161 Details : if client reconnected to restarted server we need join to recovery
162 instead of find server handler is changed and process self
163 eviction with cancel all locks.
165 Severity : enhancement
167 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
169 Severity : enhancement
171 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
173 Severity : enhancement
175 Description: Upcall on Lustre log has been dumped
176 Details : Allow for a user mode script to be called once a Lustre log has
177 been dumped. It passes the filename of the dumped log to the
178 script, the location of the script can be specified via
179 /proc/sys/lnet/debug_log_upcall.
184 Description: avoid idr_remove called for id which is not allocated.
185 Details : Move assigment s_dev for clustered nfs to end of initialization,
186 to avoid problem with error handling.
191 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
192 Details : When connection is reused this not moved from CONN_UNUSED_HASH
193 into CONN_USED_HASH and this prodice warning when put connection
194 again in unused hash.
197 Severity : enhancement
199 Description: File striping can now be set to use an arbitrary pool of OSTs.
201 Severity : enhancement
203 Description: Export bytes_read/bytes_write count on OSC/OST.
207 Description: Early reply size mismatch, MGC loses connection
208 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
209 the connect flags are properly negotiated.
214 Description: quotas are not honored with O_DIRECT
215 Details : all writes with the flag O_DIRECT will use grants which leads to
216 this problem. Now using OBD_BRW_SYNC to guard this.
220 Description: add quota statistics
221 Details : 1. sort out quota proc entries and proc code.
222 2. add quota statistics
224 Severity : enhancement
226 Description: enable quota support for HEAD.
230 Description: Properly propagate oinfo flags from lov to osc for statfs
231 Details : restore missing copy oi_flags to lov requests.
233 Severity : enhancement
235 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
236 : l_getgroups(8), lst(8), routerstat(8)
237 Details : included man pages for llobdstat(8), llstat(8),
238 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
240 Severity : enhancement
242 Description: configure's --enable-quota should check the
243 : kernel .config for CONFIG_QUOTA
244 Details : configure is terminated if --enable-quota is passed but
245 : no quota support is in kernel
249 Description: Remove portals compatibility
250 Details : Remove portals compatibility, not interoperable with releases
255 Description: Resolve device initialization race
256 Details : Prevent proc handler from accessing devices added to the
257 obd_devs array but yet be intialized.
259 Severity : enhancement
261 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
263 Severity : enhancement
265 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
270 Description: Using wrong pointer in osc_brw_prep_request
271 Details : Access to array[-1] can produce panic if kernel compiled with
272 CONFIG_PAGE_ALLOC enabled
276 Description: Client runs out of low memory
277 Details : Consider only lowmem when counting initial number of llap pages
281 Description: *optional* service tags registration
282 Details : if the "service tags" package is installed on a Lustre node
283 When the filesystem is mounted, a local-node service tag will
284 be created. See http://inventory.sun.com/ for more information
285 about the Service Tags asset management system.
289 Description: Kernel BUG tries to release flock
290 Details : Lustre does not destroy flock lock before last reference goes
291 away. So always drop flock locks when client is evicted and
292 perform unlock regardless of successfulness of speaking to MDS.
296 Description: add refcount protection for osc callbacks, avoid panic on shutdown
300 Description: sanity test 65a fails if stripecount of -1 is set
301 Details : handle -1 striping on filesystem in ll_dirstripe_verify
306 Description: ASSERTION(CheckWriteback(page,cmd)) failed
307 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
310 Severity : enhancement
312 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
316 Description: do not process already freed flock
317 Details : flock can possibly be freed by another thread before it reaches
318 to ldlm_flock_completion_ast.
322 Description: LBUG during stress test
323 Details : Need properly lock accesses the flock deadlock detection list.
327 Description: oops in page fault handler
328 Details : kernel page fault handler can return two special 'pages' in error
329 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
333 Description: timeout with invalidate import.
334 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
335 should be handled by ptlrpcd. This produce long age waiting and
336 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
338 Severity : enhancement
340 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
344 Description: enable MGS and MDT services start separately
345 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
346 a co-located MGS without starting the MGS, which is a complement
347 to 'nosvc' mount option.
351 Frequency : after recovery
352 Description: precreate to many object's after del orphan.
353 Details : del orphan st in oscc last_id == next_id and this triger growing
354 count of precreated objects. Set flag LOW to skip increase count
355 of precreated objects.
359 Frequency : rare, on clear nid stats
360 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
361 Details : when clean nid stats sometimes try destroy live entry,
362 and this produce panic in free.
366 Description: Stack overflow during MDS log replay
367 ease stack pressure by using a thread dealing llog_process.
371 Description: wait until IO finished before start new when do lock cancel.
372 Details : VM protocol want old IO finished before start new, in this case
373 need wait until PG_writeback is cleared until check dirty flag and
374 call writepages in lock cancel callback.
376 Severity : enhancement
378 Description: using special macro for print time and cleanup in includes.
382 Description: mds_mfd_close() ASSERTION(rc == 0)
383 Details : In mds_mfd_close(), we need protect inode's writecount change
384 within its orphan write semaphore to prevent possible races.
388 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
392 Frequency : rare, on shutdown ost
393 Description: don't hit live lock with umount ost.
394 Details : shrink_dcache_parent can be in long loop with destroy dentries,
395 use shrink_dcache_sb instead.
399 Description: don't panic with use echo client
400 Details : echo client pass NULL as client nid pointer and this produce null
405 Description: fix build on ppc32
406 Details : compile code with -m64 flag produce wrong object file for ppc32.
410 Description: add message levels for liblustreapi
414 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
415 Details : Move the 'good_osts' check before the 'total_bavail' check. This
416 will result in an -EAGAIN and in the exit call path we call
417 alloc_rr() which will with increasing aggressiveness attempt to
418 aquire precreated objects on the minimum number of required OSCs.
422 Description: Use old size assignment to avoid deadlock
423 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
424 the scheduling while holding a spinlock. We do not need locking
425 for size in ll_update_inode() because size is only updated from
426 the MDS for directories or files without objects, so there is no
427 other place to do the update, and concurrent access to such inodes
428 are protected by the inode lock.
432 Description: resolve "_IOWR redefined" build error on SLES10
436 Description: dump the memory debugging after all modules are unloaded to
437 suppress false negative in conf_sanity test 39
439 Severity : enhancement
441 Description: build kernel-ib packages for OFED 1.3 in our release cycle
446 Description: fix SLES kernel versioning
447 Details : the kernel version for our SLES 10 kernel did not include a "-"
448 before the "smp" at the end. while this was not a problem in
449 general, it did mean that software trying to use the kernel
450 version to try to detect a vendor specific kernel would fail.
451 this was most evident by the OFED build scripts.
455 Description: Don't update lov_desc members until making sure they are valid
456 Details : When updating lov_desc members via proc fs, need fix their
457 validities before doing the real update.
461 Description: don't put request into delay list while invalidate in flight.
462 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
463 import in flight. this produce timeout for invalidate and sometimes
464 can cause stale data.
468 Frequency : on ppc only
469 Description: not convert ost objects for directory because it's not exist.
470 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
474 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
476 Severity : enhancement
478 Description: Update to support 2.6.22.14 vanilla kernel.
482 Frequency : rare, on recovery
483 Description: read procfs can produce deadlock in some situation
484 Details : Holding lprocfs lock which send rpc can produce block for destroy
485 obd objects and this also block reconnect with -EALREADY. This
486 isn't fix all lprocfs bugs - but make it rare.
488 Severity : enhancement
490 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
493 Frequency : frequent on X2 node
495 Description: mdc_set_open_replay_data LBUG
496 Details : Set replay data for requests that are eligible for replay.
500 Description: lustre_mgs: operation 101 on unconnected MGS
501 Details : When MGC is disconnected from MGS long enough, MGS will evict the
502 MGC, and late on MGC cannot successfully connect to MGS and a lot
503 of the error messages complaining that MGS is not connected.
507 Frequency : on network error
508 Description: panic with double free request if network error
509 Details : mdc_finish_enqueue is finish request if any network error ocuring,
510 but it's true only for synchronus enqueue, for async enqueue
511 (via ptlrpcd) this incorrect and ptlrpcd want finish request
514 Severity : enhancement
516 Description: client-side metadata stat-ahead during readdir(directory readahead)
517 Details : perform client-side metadata stat-ahead when the client detects
518 readdir and sequential stat of dir entries therein
521 Frequency : on start mds
523 Description: Implement get_info(last_id) in obdfilter.
526 Frequency : occasional
528 Description: Correctly check stale fid, not start epoch if ost not support SOM
529 Details : open with flag O_CREATE need set old fid in op_fid3 because
530 op_fid2 was overwritten with new generated fid, but mds can answer
531 with one of these two fids and both is not stale. Setattr
532 incorrectly started an epoch and assume will be called
533 done_writing, but without SOM done_writing ever being called.
536 Frequency : rare, depends on device drivers and load
538 Description: MDS or OSS nodes crash due to stack overflow
539 Details : Code changes in 1.8.0 increased the stack usage of some functions.
540 In some cases, in conjunction with device drivers that use a lot
541 of stack the MDS (or possibly OSS) service threads could overflow
542 the stack. One change which was identified to consume additional
543 stack has been reworked to avoid the extra stack usage.
546 Frequency : occasional
548 Description: Do not fail import if osc_interpret_create gets -EAGAIN
549 Details : If osc_interpret_create got -EAGAIN it immediately exits and
550 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
551 oscc_has_objects and see OSC has no objests and call
552 oscc_internal_create to resend create request.
554 Severity : enhancement
556 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
558 Severity : enhancement
560 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
563 Frequency : very rare
565 Description: Fix replay if there is an un-replied request and open
566 Details : In some cases, older replay request will revert the
567 mcd->mcd_last_xid on MDS which is used to record the client's
570 Severity : enhancement
572 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
574 Severity : enhancement
576 Description: Add rhel5 support to HEAD.
578 Serverity : enhancement
580 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
585 Description: Don't allow skipping OSTs if index has been specified.
586 Details : Don't allow skipping OSTs if index has been specified, make
587 locking in internal create lots better.
591 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
592 Details : access to bitfield in structure is always rounded to long
593 and this produce problem with not atomic change any bit.
597 Description: Lustre make rpms failed.
598 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
599 builds Lustre rpms from tarball.
602 Frequency : rare on shutdown ost
604 Description: If llog cancel was not send before clean_exports phase, this can
605 produce deadlock in llog code.
606 Details : If llog thread has last reference to obd and call class_import_put
607 this produce deadlock because llog_cleanup_commit_master wait when
608 last llog_commit_thread exited, but this never success because was
609 called from llog_commit_thread.
613 Description: allow userland application know is lost one of stripes.
614 Details : fill lvb_blocks with error code on ost and return it to
615 application if error flag found.
619 Description: NULL lov_tgts causing MDS oops
620 Details : more safe checks for NULL lov_tgts for avoid oops.
622 Severity : enhancement
624 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
628 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
629 Details : If an OST is down the MDS will hang indefinitely in
630 obd_statfs_rqset() waiting for the statfs data. While for
631 MDS QOS usage of statfs, it should not stuck in waiting.
633 Severity : enhancement
635 Description: remote_acl support
636 Details : Support ACL-based permission check for remote user.
637 Support setfacl/getfacl for remote user with the utils
638 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
639 the system "{s,g}etfacl" utils.
641 Severity : enhancement
643 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
645 Severity : enhancement
647 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
651 Description: make mgs_setparam() handle fsname containing dash
652 Details : fsname containing a dash does not work with lctl conf_param
654 Severity : enhancement
656 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
658 Severity : enhancement
660 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
664 Frequency : rare, at shutdown
665 Description: access already free / zero obd_namespace.
666 Details : if client_disconnect_export was called without force flag set,
667 and exist connect request in flight, this can produce access to
668 NULL pointer (or already free pointer) when connect_interpret
669 store ocd flags in obd_namespace.
673 Frequency : only at startup
674 Description: not alloc memory with spinlock held.
675 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
676 if any spinlock held.
678 Severity : enhancement
680 Description: make lustre randomly fail allocating memory
681 Details : Make lustre randomly failed allocating memory for testing purpose.
683 Severity : enhancement
685 Description: lost problems with lov objid file
686 Details : Fixes some scability and access to not inited memory problems
687 in work with lov objdid file.
692 Description: lfs find does not continue on file error
693 Details : Continue other files processing when a file/dir is absent.
697 Description: Inconsistent usage of lustre_pack_reply()
698 Details : Standardize the usage of lustre_pack_reply() such that it
699 always generate a CERROR on failure.
702 Frequency : occasional
704 Description: MDS hang or stay in waiting lock
705 Details : If client receive lock with CBPENDING flag ldlm need send lock
706 cancel as separate rpc, to avoid situation when cancel request
707 can't processed due all i/o threads stay in wait lock.
711 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
715 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
719 Description: Update to SLES9 kernel-2.6.5-7.287.3.
723 Description: Update to RHEL4 latest kernel.
725 Severity : enhancement
727 Description: Build SLES10 patchless client fails
728 Details : The configure was broken by run ./configure with
729 --with-linux-obj=.... argument for patchless client. When the
730 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
731 can't search header adequately. Use absolute path such as
732 -I($LINUX)/include instead.
736 Description: interrupt oig_wait produce painc on resend.
737 Details : brw_redo_request can be used for resend requests from ptlrpcd and
738 private set, and this produce situation when rq_ptlrpcd_data not
739 copyed to new allocated request and triggered LBUG on assert
740 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
741 ptlrpcd set if request is changed and can be safety changed to use
746 Description: LASSERT_{REQ,REP}SWAB macros are buggy
747 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
748 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
749 these macros and replace them with their difinitions instead.
753 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
754 Details : Update which_patch & target file for SLES10 latest kernel.
757 Description: Remove client patches from SLES 10 kernel.
758 Details : This causes SLES 10 clients to behave as patchless clients
759 even on a Lustre-patched (server) kernel.
761 Severity : enhancement
763 Description: self-adjustable client's lru lists
764 Details : use adaptive algorithm for managing client cached locks lru
765 lists according to current server load, other client's work
766 pattern, memory activities, etc. Both, server and client
767 side namespaces provide number of proc tunables for controlling
770 Severity : enhancement
772 Description: light-weight GSS support
773 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
774 and reduce performance overhead.
776 Severity : enhancement
778 Description: Linux keyring support
779 Details : Support using service of Linux keyring for Lustre GSS internal
780 context refresh/cache mechanism.
784 Description: Fix errors in lfs documentation
785 Details : Fixes man pages
789 Description: don't use GFP_* in generic Lustre code.
790 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
794 Description: obdclass is limited by single OBD_ALLOC(idarray)
795 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
799 Description: warnings with build patchless client with vanila 2.6.19 and up
800 Details : change old ctl_table style and replace ctl_table/ctl_table_header
801 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
805 Frequency : Only for SLES
806 Description: sanity_quota fail test_1
807 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
808 file, which makes TSTID[2] inunique.
812 Frequency : Only for SLES9
813 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
814 Details : Update target/ChangeLog/which_patch .
818 Frequency : Only for test_10 in sanity-quota.sh
819 Description: a bug in quota test code
820 Details : checking a test flag in a wrong place
824 Frequency : Only for test_13 in sanity-quota.sh
825 Description: a bug in quota test script
826 Details : shouldn't check overfull allocated quota
830 Description: jbd statistics
831 Details : Port older jbd statistics patch for sles10
833 Severity : enhancement
835 Description: Add printing inode info into message about error in writepage.
840 Description: reply_lock_interpret crash due to race with it and lock cancel.
841 Details : Do not replay locks that are being cancelled. Do not reference
842 locks by their address during replay, just by their handle.
846 Frequency : When flocks are used.
847 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
848 Details : Flock locks might destroy just granted lock if it could be merged
849 with another existing flock, this is done in completion handler,
850 so teach ldlm_cli_enquque_fini that this is a valid case for
856 Description: Oops in read and write path when failing to allocate lock.
857 Details : Check if lock allocation failed and return error back.
861 Description: lstripe command fails for valid OST index
862 Details : The stripe offset is compared to lov->desc.ld_tgt_count
863 instead of lov->desc.ld_active_tgt_count.
867 Description: sanity.sh failed test 103
868 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
873 Description: df doesn't work properly if diskfs blocksize != 4K
874 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
878 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
879 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
881 Severity : enhancement
883 Description: Adaptive timeouts
884 Details : RPC timeouts adapt to changing server load and network
885 conditions to reduce resend attempts and improve recovery time.
889 Description: llapi_file_create() does not allow some changes
890 Details : add llapi_file_open() that allows specifying the mode and
891 open flags, and also returns an open file handle.
895 Description: merge and cleanup kernel patches.
896 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
900 Description: Add journal checksum support.(Kernel part)
901 Details : The journal checksum feature adds two new flags i.e
902 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
903 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
904 indicates that the commit block contains the checksum for
905 the blocks described by the descriptor blocks. Now commit
906 record can be sent to disk without waiting for descriptor
907 blocks to be written to disk. This behavior is controlled
908 using JBD2_FEATURE_ASYNC_COMMIT flag.
912 Description: OSS needs mutliple precreate threads
913 Details : Add ability to start more than one create thread per OSS.
917 Description: Sanity.sh test_65k failed: setstripe should have succeeded
918 Details : execute lfs setstripe on client
922 Description: mds_obd_create error creating tmp object
923 Details : When the user sets quota on root, llog will be affected and can't
924 create files and write files.
927 Frequency : Always on ia64 patchless client, and possibly others.
929 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
930 Details : This allows the patchless client to be loaded on architectures
935 Description: RedHat Update kernel for RHEL5
936 Details : Add the kernel config file for RHEL5.
940 Description: RedHat Update kernel for RHEL5
941 Details : Modify the kernel config file more closer RHEL5.
945 Description: Build failure against Centos5 (RHEL5)
946 Details : Define PAGE_SIZE when it isn't present.
950 Description: test_7 on sanity-quota.sh can't be used on separate machines
951 Details : add facet to handle it
955 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
956 Details : with NFS, the anon dentry's parent was set to itself in
957 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
958 corresponding dentry other than use rec->ur_name.
960 Severity : enhancement
962 Description: lfs setstripe enhancement
963 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
967 Description: enable data checksumming by default
968 Details : enable checksum by default, allow --disable-checksum
969 configure option and "-o nochecksum" mount option. Checksums
970 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
974 Description: lustre support for RHEL5
975 Details : Add support for RHEL5.
979 Description: Client eviction due to failover config
980 Details : after a connection loss, the lustre client should attempt to
981 reconnect to the last active server first before trying the
982 other potential connections.
986 Description: when mds and osts use different quota unit(32bit and 64bit),
987 quota will be released repeatly.
988 Details : void sending multiple quota reqs to mds, which will keep the
989 status between the reqs.
993 Description: osts not allocated evenly to files
994 Details : change the condition to increase offset_idx
997 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
999 Description: Data corruption for OSTs that are formatted larger than 2TB
1001 Details : When generating the bio request for lustre file writes the
1002 sector number would overflow a temporary variable before being
1003 used for the IO. The data reads correctly from Lustre (which
1004 will overflow in a similar manner) but other file data or
1005 filesystem metadata may be corrupted in some cases.
1009 Description: Tune the kernel for good SCSI performance.
1010 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1011 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1016 Description: rewrite ext2-derived code in obdclass/uuid.c
1017 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1018 from scratch preserving functionality.
1022 Description: "lfs find -obd UUID" prints directories
1023 Details : "lfs find -obd UUID" will return all directory names instead
1024 of just file names. It is incorrect because the directories
1025 do not reside on the OSTs.
1029 Description: use i_size_read and i_size_write in 2.6 port
1030 Details : replace inode->i_size access with i_size_read/write()
1033 Frequency : when removing large files
1035 Description: scheduling issue during removal of large Lustre files
1036 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1037 It causes scheduling issues when removing large files (17TB in the
1041 Frequency : only with liblustre clients on XT3
1043 Description: evictions taking too long
1044 Details : allow llrd to evict clients directly on OSTs
1047 Frequency : only on ppc
1049 Description: /proc/fs/lustre/devices broken on ppc
1050 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1051 We should make sure the type of 'index' is loff_t and then cast
1052 explicitly as needed below. Do not assign an explicitly cast
1056 Frequency : only for rhel5
1058 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1059 Details : Modify the target file & which_kernel.
1061 Severity : enhancement
1063 Description: omit set fsid for export NFS
1064 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1065 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1069 Frequency : Always, for kernels after 2.6.16
1070 Description: Fix warning idr_remove called for id=.. which is not allocated.
1071 Details : Last kernels save old s_dev before kill super and not allow
1072 to restore from callback - restore it before call kill_anon_super.
1076 Description: buffer overruns could theoretically occur
1077 Details : llapi_semantic_traverse() modifies the "path" argument by
1078 appending values to the end of the origin string, and a
1079 overrun may occur. Adding buffer overrun check in liblustreapi.
1083 Description: Fix error on 'ls .' at the top of the Lustre mount.
1084 Details : Don't revalidate dentry if it is a root dentry.
1088 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1089 Details : Modify vm-tunables-rhel4.patch.
1093 Description: Kernel config for 2.6.18-vanilla.
1094 Details : Modify targets/2.6-vanilla.target.in.
1095 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1096 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1097 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1098 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1102 Description: improve handling recoverable errors
1103 Details : if request processig with error which can be recoverable on server
1104 request should be resend, otherwise page released from cache and
1109 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1110 Details : Modify target file & which_patch.
1111 A flaw was found in the IA32 system call emulation provided
1112 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1113 value could be stored in the %RAX register, which could trigger an
1114 out-of-bounds system call table access. An untrusted local user
1115 could exploit this flaw to run code in the kernel
1116 (ie a root privilege escalation). (CVE-2007-4573).
1120 Description: O_DIRECT bypasses client statistics.
1121 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1122 read_ahead_stats were updated. copied the stats section from
1123 osc_send_oap_rpc() into async_internal().
1127 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1131 Description: change order of libsysio includes
1132 Details : '#include sysio.h' should always come before '#include xtio.h'
1136 Description: handle "serious error: objid * already exists" more gracefully
1137 Details : If LAST_ID value on disk is smaller than the objects existing in
1138 the O/0/d* directories, it indicates disk corruption and causes an
1139 LBUG(). If the object is 0-length, then we should use the existing
1140 object. This will help to avoid a full fsck in most cases.
1142 Severity : enhancement
1144 Description: adapt the lustre_config script to support the upgrade case
1145 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1146 server targets to 1.6 in parallel.
1150 Description: To avoid grant space > avaible space when the disk is almost
1151 full. Without this patch you might see the error "grant XXXX >
1152 available" or some LBUG about grant, when the disk is almost
1154 Details : In filter_check_grant, for non_grant cache write, we should
1155 check the left space by if (*left > ungranted + bytes), instead
1156 of (*left > ungranted), because only we are sure the left space
1157 is enough for another "bytes", then the ungrant space should be
1158 increase. In client, we should update cl_avail_grant only there
1159 is OBD_MD_FLGRANT in the reply.
1163 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1164 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1165 described in CVE-2007-4573. This problem could allow untrusted
1166 local users to gain root access.
1169 Frequency : when using O_DIRECT and quotas
1171 Description: Incorrect file ownership on O_DIRECT output files
1172 Details : block usage reported by 'lfs quota' does not take into account
1173 files that have been written with O_DIRECT.
1178 Description: touch file failed when fs is not full
1179 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1180 otherwise we can get ENOSP while fs is not full.
1184 Description: parallel lock callbacks
1185 Details : Instead of sending blocking and completion callbacks as separated
1186 requests, adding them to a set and sending in parallel.
1189 Frequency : only for Cray XT3
1190 Bugzilla : 12829/13455
1191 Description: Changing primary group doesn't change the group lustre assigns to
1193 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1194 overridden with the primary group provided by the group upcall,
1195 whereas the supplied fsgid can be trusted if it is in the list of
1196 supplementary groups returned by the group upcall.
1198 Severity : enhancement
1200 Description: Allow masking D_WARNING, D_ERROR messages from console
1201 Details : Console messages can now be disabled via lnet.printk.
1205 Description: User code with malformed file open parameter crashes client node
1206 Details : Before packing join_file req, all the related reference should be
1207 checked carefully in case some malformed flags cause fake
1208 join_file req on client.
1212 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1213 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1214 mode is equal to the granted mode. However, it should also check
1215 that LDLM_FL_CP_REQD is not set.
1219 Description: Heavy nfs access might result in deadlocks
1220 Details : After ELC code landed, it is now improper to enqueue any mds
1221 locks under och_sem, because enqueue might want to decide to
1222 cancel open locks for same inode we are holding och_sem for.
1226 Description: Client eviction while running blogbench
1227 Details : A lot of unlink operations with concurrent I/O can lead to a
1228 deadlock causing evictions. To address the problem, the number of
1229 oustanding OST_DESTROY requests is now throttled to
1230 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1231 callbacks are processed in priority.
1235 Description: enable ACLs on MDS by default
1236 Details : ACLs must be enabled on MDS by default.
1239 Frequency : PPC/PPC64 only
1241 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1242 Details : fix duplicated definitions between asm-ppc64/types.h and
1243 lustre_types.h on PPC.
1246 Frequency : PPC/PPC64 only
1248 Description: asm-ppc/segment.h does not exist
1249 Details : fix compile issue on PPC.
1253 Description: better handle error messages in extents code
1256 Frequency : RHEL4 only
1258 Description: mkfs is very slow on IA64/RHEL4
1259 Details : A performance regression has been discovered in the MPT Fusion
1260 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1261 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1262 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1264 Severity : enhancement
1266 Description: SNMP support enhancement
1267 Details : Adding total number of sampled request for an MDS node in snmp
1270 Severity : enhancement
1272 Description: Optimize ldlm waiting list processing for PR extent locks
1273 Details : When processing waiting list for read extent lock and meeting read
1274 lock that is same or wider to it that is not contended, skip
1275 processing rest of the list and immediatelly return current
1276 status of conflictness, since we are guaranteed there are no
1277 conflicting locks in the rest of the list.
1281 Description: Time out and refuse to reconnect
1282 Details : When the failover node is the primary node, it is possible
1283 to have two identical connections in imp_conn_list. We must
1284 compare not conn's pointers but NIDs, otherwise we can defeat
1285 connection throttling.
1289 Description: port llog fixes from b1_6 into HEAD
1290 Details : Port llog reference couting and some llog cleanups from b1_6
1291 (bug 10800) into HEAD, for protect from panic and access to already
1292 free llog structures.
1296 Description: Detect stride IO mode in read-ahead
1297 Details : When a client does stride read, read-ahead should detect that and
1298 read-ahead pages according to the detected stride pattern.
1302 Description: data checksumming impacts single node performance
1303 Details : add support for several checksum algorithm. Currently, only CRC32
1304 and Adler-32 are supported. The checksum type can be changed on
1305 the fly via /proc/fs/lustre/osc/*/checksum_type.
1309 Description: use adler32 for page checksums
1310 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1315 Description: build for x2 fails
1316 Details : fix compile issue on Cray systems.
1320 Description: Properly match for duplicate locks
1321 Details : Due to different lock order from skiplists code, we need to
1322 traverse entire list for now
1325 Frequency : only on PPC/SLES10
1327 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1328 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1329 be defined. Add a hack in mkfs_lustre.c to work around this compile
1334 Description: LASSERT on MDS when client holding flock lock dies
1335 Details : ldlm pool logic depends on number of granted locks equal to
1336 number of released locks which is not true for flock locks, so
1337 just exclude such locks from consideration.
1341 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1342 Details : Use fsfilt_sync() for both the whole filesystem sync and
1343 individual file sync to eliminate dangerous inode locking
1344 with I_LOCK that can lead to a deadlock.
1348 Description: performance in 1.6.3
1349 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1350 sure that 1MB requests are not fragmented by the block layer.
1352 Severity : enhancement
1354 Description: organize the server-side client stats on per-nid basis
1355 Details : Change the structure of stats under obdfilter and mds to
1363 The "uuid"s file would list the uuids of _active_ exports.
1364 And the clear entry is to clear all stats and stale nids.
1366 Severity : enhancement
1368 Description: eliminate client locks in face of contention
1369 Details : file contention detection and lockless i/o implementation
1370 for contended files.
1374 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1377 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1379 Description: MDS may lose file striping (and hence file data) in some cases
1380 Details : If there are additional extended attributes stored on the MDS,
1381 in particular ACLs, SELinux, or user attributes (if user_xattr
1382 is specified for the client mount options) then there is a risk
1383 of attribute loss. Additionally, the Lustre file striping
1384 needs to be larger than default (e.g. striped over all OSTs),
1385 and an additional attribute must be stored initially in the
1386 inode and then increase in size enough to be moved to the
1387 external attribute block (e.g. ACL growing in size) for file
1392 Description: skiplist implementation simplification
1393 Details : skiplists are used to group compatible locks on granted list
1394 that was implemented as tracking first and last lock of each
1395 lock group the patch changes that to using doubly linked lists
1399 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1400 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1401 because LASSERT on both the data supplied by a client, and the
1402 data on disk is dangerous and incorrect.
1404 Severity : enhancement
1406 Description: Slow truncate/writes to huge files at high offsets.
1407 Details : Directly associate cached pages to lock that protect those pages,
1408 this allows us to quickly find what pages to write and remove
1409 once lock callback is received.
1413 Description: more ldlm soft lockups
1414 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1415 starve other threads from the resource lock for a long time in
1416 case of long waiting queue, so change the debug level from
1417 D_OTHER to the less frequently used D_INFO.
1419 Severity : enhancement
1421 Description: add -gid, -group, -uid, -user options to lfs find
1425 Description: Hung threads in invalidate_inode_pages2_range
1426 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1427 once one is completed. As a result, some RPCs are stuck in the
1428 queue and are never sent.
1432 Description: filter threads hungs on waiting journal commit
1433 Details : Cleanup filter group llog code, then only filter group llog will
1434 be only created in the MDS/OST syncing process.
1438 Description: Procfs and llog threads access destoryed import sometimes.
1439 Details : Sync the import destoryed process with procfs and llog threads by
1440 the import refcount and semaphore.
1442 Severity : enhancement
1444 Description: openlock cache of b1_6 port to HEAD
1449 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1450 Details : If insertion of an extent fails, then discard the inode
1451 preallocation and free data blocks else it can lead to duplicate
1456 Description: don't always update ctime in ext3_xattr_set_handle()
1457 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1458 In some cases the ctime should not be updated, for example for
1459 2.0->1.8 compatibility it is necessary to delete an xattr and it
1460 should not update the ctime.
1464 Bugzilla : 15713/16362
1465 Description: Assertion in iopen_connect_dentry in 1.6.3
1466 Details : looking up an inode via iopen with the wrong generation number can
1467 populate the dcache with a disconneced dentry while the inode
1468 number is in the process of being reallocated. This causes an
1469 assertion failure in iopen since the inode's dentry list contains
1470 both a connected and disconnected dentry.
1474 Description: assertion failure in ldlm_handle2lock()
1475 Details : fix a race between class_handle_unhash() and class_handle2object()
1476 introduced in lustre 1.6.5 by bug 13622.
1481 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1482 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1487 Description: oops in sd_iostats_seq_show()
1488 Details : unloading/reloading the scsi low level driver triggers a kernel
1489 bug when trying to access the sd iostat file.
1494 Description: Kernel panics during QLogic driver reload
1495 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1496 patch, causing memory corruption.
1501 Description: journal_dev option does not work in b1_6
1502 Details : pass mount option during pre-mount.
1504 Severity : enhancement
1506 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1507 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1508 extent information of a file. It can be used to map logical blocks
1509 in a file to physical blocks in the block device.
1513 Description: LDLM soft lockups - improvement
1514 Details : It is be possible to send the lock handle along with each read
1515 or write request because the client is already doing a lock match
1516 itself so there isn't any reason the OST should have to re-do that
1520 Frequency : only on Cray X2
1522 Description: X2 build failures
1523 Details : fix build failures on Cray X2.
1527 Description: xid & resent requests
1528 Details : Initialize RPC XID from clock at startup (randomly if clock is
1531 Severity : enhancement
1533 Description: Add lustre_start utility to start or stop multiple Lustre servers
1538 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1539 Details : In case of memory pressure, list_del() can be called twice on
1540 req->rq_history_list, causing a kernel oops.
1544 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1545 Details : fix stack overflow in the distributed lock manager by defering
1546 export eviction after a failed AST to the elt thread instead of
1547 handling it in the dlm interpret routine.
1551 Description: Convert some comments to new format.
1552 Details : Update documenting comments to match doxygen conventions.
1556 Description: Grammar fixes.
1557 Details : A couple of trivial spelling fixes.
1561 Description: OSD_COUNTERS-mandatory
1562 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1563 to simplify the code.
1567 Description: simplify cmm_device freeing logic.
1568 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1572 Description: Add lockdep support to dt_object_operations locking interface.
1573 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1574 parameter indicating lock ordering. Update mdd code to use new
1579 Description: Introduce failloc constants for lockless IO tests.
1580 Details : Add two new failloc constants to test lockless IO. Only one of
1581 them in implemented---another is checked in yet to be landed
1586 Description: Add lockdep support for inode mutex.
1587 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1588 in the situations where Lustre has to lock more than one inode
1593 Description: Add optional invariants checking support.
1594 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1595 configure switch. This macro is to be used for consistency and
1596 sanity checks that are too expensive to be left in `production'
1601 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1602 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1607 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1608 Details : Use spin_lock_nested() in (the only) situation where more than
1609 one ldlm_lock is locked simultaneously. Also, fix possible
1610 dead-lock in ldlm_lock_change_resource() by enforcing particular
1615 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1616 Details : Instead of specifying each ldlm_lock call-back through separate
1617 parameter, wrap them into struct ldlm_callback_suite.
1621 Description: Kill join_lru obd method and its callers.
1622 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1623 in memory---a requirement implemented through ->o_join_lru() obd
1624 method in HEAD. Remove this method and its users.
1628 Description: Add asynchronous ldlm ENQUEUE completion handler.
1629 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1630 case that stock ldlm_completion_ast() cannot handle as it waits
1631 until lock is granted. Introduce new ldlm_completion_ast_async()
1632 for this. Also comment ldlm_completion_ast().
1636 Description: ldlm_error <-> errno conversion.
1637 Details : Add functions to map (rather arbitrary) between LDLM error codes
1638 and standard errno values. CLIO needs this to prevent LDLM specific
1639 constants from escaping ldlm and osc.
1643 Description: Kill unused ldlm_handle2lock_ns() function.
1644 Details : Kill unused ldlm_handle2lock_ns() function.
1648 Description: Add lu_ref support to ldlm_lock
1649 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1650 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1651 and ldlm_resource. LDLM interface has to be changed, because code that
1652 releases a reference on a lock, has to "know" what reference this is.
1653 In the most frequent case
1655 lock = ldlm_handle2lock(handle);
1657 LDLM_LOCK_PUT(lock);
1659 no changes are required. When any other reference (received _not_ from
1660 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1661 instead of LDLM_LOCK_PUT().
1663 Arguably, changes are pervasive, and interface requires some discipline
1664 for proper use. On the other hand, it was very instrumental in finding
1665 a few leaked lock references.
1669 Description: Add ldlm_lock_addref_try().
1670 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1671 attempts to addref a lock that might be being canceled
1676 Description: Add ldlm_weigh_callback().
1677 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1678 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1679 hard-coded `number of pages' logic.
1683 Description: Add lockdep annotations to llog code.
1684 Details : Use appropriately tagged _nested() locking calls in the places
1685 where llog takes more than one ->lgh_lock lock.
1689 Description: Add loi_kms_set().
1690 Details : Wrap kms updates into a helper function.
1694 Description: Constify instances of struct lsm_operations.
1695 Details : Constify instances of struct lsm_operations.
1699 Description: lu_conf support.
1700 Details : On a server, a file system object is uniquely identified
1701 by a fid, which is sufficient to locate and load all object
1702 state (inode). On a client, on the other hand, more data are
1703 necessary instantiate an object. Change lu_object_find() and
1704 friends to take additional `lu_conf' argument describing object.
1705 Typically this includes layout information.
1709 Description: lu_context fixes.
1710 Details : Introduce new lu_context functions that are needed on the client
1711 side, where some system threads (ptlrpcd) are shared by multiple
1712 modules, and so cannot be stopped during module shutdown.
1716 Description: Add start and stop methods to lu_device_type_operations.
1717 Details : Introduce two new methods in lu_device_type_operations, that are
1718 invoked when first instance of a given type is created and last one
1719 is destroyed respectively. This is need by CLIO.
1723 Description: Add lu_ref support to struct lu_device.
1724 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1725 track leaked references.
1729 Description: Introduce lu_kmem_descr.
1730 Details : lu_kmem_descr and its companion interface allow to create
1731 and destroy a number of kmem caches at once.
1735 Description: Fix lu_object finalization race.
1736 Details : Fix a race between lu_object_find() finding an object and its
1737 concurrent finalization. This race is (most likely) not possible
1738 on the server, but might happen on the client.
1742 Description: Introduce lu_ref interface.
1743 Details : lu_ref is a debugging module allowing to track references to
1744 a given object. It is quite cpu expensive, and has to be
1745 explicitly enabled with --enable-lu_ref. See usage description
1750 Description: Factor lu_site procfs stats into a separate function.
1751 Details : Separate lu_site stats printing code into a separate function
1752 that can be reused on a client.
1756 Description: Constify instances of struct {lu,dt,md}_device_operations.
1757 Details : Constify instances of struct {lu,dt,md}_device_operations.
1761 Description: Introduce struct md_site and move meta-data specific parts of
1762 struct lu_site here.
1763 Details : Move md-specific fields out of struct lu_site into special struct
1764 md_site, so that lu_site can be used on a client.
1768 Description: Kill mdd_lov_destroy().
1769 Details : Remove unused mdd code.
1773 Description: Add st_block checking to multistat.c.
1774 Details : Add st_block checking to multistat.c.
1778 Description: Add lu_ref support to struct obd_device.
1779 Details : Add lu_ref tracking to obd_device.
1783 Description: Kill obd_set_fail_loc().
1784 Details : Remove unused code.
1788 Description: Add special type for ptlrpc_request interpret functions.
1789 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1790 there. Actual usage will be in CLIO.
1794 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1795 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1796 doesn't work with lockdep.
1800 Description: Add rwv.c test program.
1801 Details : New testing program exercising readv(2) and writev(2) (Qian).
1805 Description: Add sendfile.c test program.
1806 Details : New testing program exercising sendfile(2) (Jay).
1810 Description: Ratelimit a message that can be very frequent.
1811 Details : Ratelimit a memory allocation failure message that can
1816 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1817 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1821 Description: Liblustre build fixes.
1822 Details : Liblustre build fixes.
1826 Description: libcfs: add cfs_{need,cond}_resched() interface.
1827 Details : libcfs: add cfs_{need,cond}_resched() definition and
1828 implementations for Linux, NT, and liblustre.
1830 Severity : enhancement
1832 Description: More exported tunables for mballoc
1833 Details : Add support for tunable preallocation window and new tunables for
1834 large/small requests
1838 Description: Detect corruption of block bitmap and checking for preallocations
1839 Details : Checks validity of on-disk block bitmap. Also it does better
1840 checking of number of applied preallocations. When corruption is
1841 found, it turns filesystem readonly to prevent further corruptions.
1845 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1846 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1847 window to 0. If it does hit cache miss, and read-pattern is still
1848 stride-io mode, does not reset the stride window, but also does not
1849 increase the stride window length in this case.
1853 Frequency : only for big-endian servers
1854 Description: Check if system is big-endian while mounting fs with extents feature
1855 Details : Mounting a filesystem with extents feature will fail on big-endian
1856 systems since ext3-based ldiskfs is not supported on big-endian
1857 systems. This can be over-riden with "bigendian_extents" mount option.
1859 Severity : enhancement
1861 Description: The root squash functionality
1862 Details : A security feature, which is to prevent users from being able
1863 to mount lustre on their desktop, run as root, and delete
1864 all of the files in the filesystem. The goal is accomplished by
1865 remapping user id (UID) and group id (GID) of the root user to
1866 a UID and GID specified by the system administartor via Lustre
1867 configuration management server (MGS). The functionality also
1868 allows to specify sets of clients for which the remapping does
1873 Description: Excessive recovery window
1874 Details : With AT enabled, the recovery window can be excessively long (6000+
1875 seconds). To address this problem, we no longer use
1876 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1877 timeout no longer depends on the service time, it is set to
1878 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1879 time via pb_service_time.
1881 --------------------------------------------------------------------------------
1883 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1885 * Support for kernels:
1886 2.6.5-7.283 (SLES 9),
1887 2.6.9-55.EL (RHEL 4),
1888 2.6.16.46-0.14 (SLES 10),
1889 2.6.18.8 vanilla (kernel.org)
1890 * Client support for unpatched kernels:
1891 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1892 2.6.16 - 2.6.22 vanilla (kernel.org)
1893 * Due to recently discovered recovery problems, we do not recommend
1894 using patchless RHEL 4 clients with this or any earlier release.
1895 * Recommended e2fsprogs version: 1.39.cfs8
1896 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1897 * Starting with this release, the ldiskfs backing filesystem required
1898 by Lustre is now in its own package, lustre-ldiskfs. This package
1899 should be installed. It is versioned separately from Lustre and
1900 may be released separately in future.
1904 Description: block reactivating mgc import until all deactivates complete
1905 Details : Fix race when failing back MDT/MGS to itself (testing)
1908 Frequency : at statup only
1910 Description: mds_lov_synchronize race leads to various problems
1911 Details : simultaneous MDT->OST connections at startup can cause the
1912 sync to abort, leaving the OSC in a bad state.
1914 Severity : enhancement
1916 Description: add optional extra BUILD_VERSION info
1917 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1918 to override the lustre version.
1921 Frequency : 2.6.18 servers only
1923 Description: ll_kern_mount() doesn't release the module reference
1924 Details : The ldiskfs module reference count never drops down to 0
1925 because ll_kern_mount() doesn't release the module reference.
1930 Description: server LBUG when using old ost_num_threads parameter
1931 Details : Accept the old ost_num_threads parameter but warn that it
1932 is deprecated, and fix an off-by-one error that caused an LBUG.
1937 Description: Transient SCSI error results in persistent IO issue
1938 Details : iobuf->dr_error is not reinitialized to 0 between two
1942 Frequency : sometimes when underlying device returns I/O errors
1944 Description: OSTs not going read-only during write failures
1945 Details : OSTs are not remounted read-only when the journal commit threads
1946 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1947 instead of the ext3 wrappers.
1951 Description: poor connect scaling with increasing client count
1952 Details : Don't run filter_grant_sanity_check for more than 100 exports
1953 to improve scaling for large numbers of clients.
1956 Frequency : SLES10 only
1958 Description: sanity-quota.sh quotacheck failed: rc = -22
1959 Details : Quotas cannot be enabled on SLES10.
1962 Frequency : liblustre clients only
1964 Description: getdirentries does not give error when run on compute nodes
1965 Details : getdirentries does not fail when the size specified as an argument
1966 is too small to contain at least one entry
1968 Severity : enhancement
1970 Description: Add LNET router traceability for debug purposes
1971 Details : If a checksum failure occurs with a router as part of the
1972 IO path, the NID of the last router that forwarded the bulk data
1973 is printed so it can be identified.
1978 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1979 Details : Due to a race condition, liblustre clients were occasionally
1980 evicted incorrectly.
1982 Severity : enhancement
1984 Description: lfs setstripe use optional parameters instead of postional
1987 Severity : enhancement
1989 Description: Nanosecond timestamp support for ldiskfs
1990 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1991 resolution timestamps. There is not yet support for this at
1992 the Lustre filesystem level.
1995 Frequency : during server recovery
1997 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1998 Details : request with rq_no_resend flag not awake l_wait_event if they get
2002 Frequency : nfs export on patchless client
2004 Description: connectathon hang when test nfs export over patchless client
2005 Details : Disconnected dentry cannot be found with lookup, so we do not need
2006 to unhash it or make it invalid
2009 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2010 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2011 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2016 Description: open req refcounting wrong on reconnect
2017 Details : If reconnect happened between getting open reply from server and
2018 call to mdc_set_replay_data in ll_file_open, we will schedule
2019 replay for unreferenced request that we are about to free.
2020 Subsequent close will crash in variety of ways.
2021 Check that request is still eligible for replay in
2022 mdc_set_replay_data().
2027 Description: disable writes to filesystem when reading health_check file
2028 Details : the default for reading the health_check proc file has changed
2029 to NOT do a journal transaction and write to disk, because this
2030 can cause reads of the /proc file to hang and block HA state
2031 checking on a healthy but otherwise heavily loaded system. It
2032 is possible to return to the previous behaviour during configure
2033 with --enable-health-write.
2035 Severity : enhancement
2037 Description: 64-bit inode version
2038 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2039 to the inode. This will be required for version-based recovery.
2044 Description: MDS fails to start if a duplicate client export is detected
2045 Details : in some rare cases it was possible for a client to connect to
2046 an MDS multiple times. Upon recovery the MDS would detect this
2047 and fail during startup. Handle this more gracefully.
2049 Severity : enhancement
2051 Description: Add -o localflock option to simulate old noflock
2053 Details : This will achieve local-only flock/fcntl locks
2059 Description: log_commit_thread vs filter_destroy race leads to crash
2060 Details : Take import reference before releasing llog record semaphore
2065 Description: Wrong request locking in request set processing
2066 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2067 imp_delayed_list, in this place should be used imp_lock.
2070 Frequency : when reconnection
2072 Description: Grant Leak when osc reconnect to OST
2073 Details : When osc reconnect ost, OST(filter) should check whether it
2074 should grant more space to client by comparing fed_grant and
2075 cl_avail_grant, and return the granted space to client instead
2076 of "new granted" space, because client will call osc_init_grant
2077 to update the client grant space info.
2080 Frequency : when client reconnect to OST
2082 Description: Grant Leak when osc do resend and replay bulk write
2083 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2084 bulk write request, because the grant info while be sync between
2085 OSC and OST when reconnect, and we should ignore the grant info
2086 these of resend/replay write req.
2091 Description: Grant space more than avaiable left space sometimes.
2092 Details : When then OST is about to be full, if two bulk writing from
2093 different clients came to OST. Accord the avaliable space of the
2094 OST, the first req should be permitted, and the second one
2095 should be denied by ENOSPC. But if the seconde arrived before
2096 the first one is commited. The OST might wrongly permit second
2097 writing, which will cause grant space > avaiable space.
2100 Frequency : when client is evicted
2102 Description: Grant might be wrongly erased when osc is evicted by OST
2103 Details : when the import is evicted by server, it will fork another
2104 thread ptlrpc_invalidate_import_thread to invalidate the
2105 import, where the grant will be set to 0. While the original
2106 thread will update the grant it got when connecting. So if
2107 the former happened latter, the grant will be wrongly errased
2108 because of this race.
2113 Description: Checking Stale with correct fid
2114 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2115 to check whether it is stale, because sometimes, we want the
2116 enqueue happened anyway, and op_data.fid2 will not be initialized.
2118 Severity : enhancement
2120 Description: update patchless client
2121 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2124 Frequency : only with 2.4 kernel
2126 Description: random memory corruption
2127 Details : size of struct ll_inode_info is to big for union inode.u and this
2128 can be cause of random memory corruption.
2133 Description: Memory leak in recovery
2134 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2135 It should also check obd_fail before fsfilt_start, otherwise if
2136 fsfilt_start return -EROFS,(failover mds during mds recovery).
2137 then the req will return with repmsg->transno = 0 and rc = EROFS.
2138 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2139 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2140 be freed no matter whether fsfilt_commit success or not.
2143 Frequency : only with huge count clients
2145 Description: Prevents from taking the superblock lock in llap_from_page for
2147 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2148 for llap_from_page call in ll_removepage prevents from taking the
2149 superblock lock for a soon died page.
2154 Description: Not check open intent error before release open handle
2155 Details : in some rare cases, the open intent error is not checked before
2156 release open handle, which may cause
2157 ASSERTION(open_req->rq_transno != 0), because it tries to release
2158 the failed open handle.
2163 Description: Set cat log bitmap only after create log success.
2164 Details : in some rare cases, the cat log bitmap is set too early. and it
2165 should be set only after create log success.
2169 Description: Accessing a block bevice can re-enable I/O when Lustre is
2170 tearing down a device.
2171 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2176 Description: service threads may hog cpus when there are a lot of requests
2177 Details : Insert cond_resched to give other threads a chance to use some CPU
2182 Description: the cat log was not initialized in recovery
2183 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2184 unlink log on mds will not be initialized until mds post
2185 recovery. And also in mds post recovery, the unlink log will
2186 initialization will be done asynchronausly, so there will be race
2187 between add unlink log and unlink log initialization.
2191 Description: brw_stats were being printed incorrectly
2192 Details : brw_stats were being printed as log2 but all of them were not
2193 recorded as log2. Also remove some code duplication arising from
2194 filter_tally_{read,write}.
2198 Frequency : rare, only in recovery.
2199 Description: ASSERTION(req->rq_type != LI_POISON) failed
2200 Details : imp_lock should be held while iterating over imp_sending_list for
2201 prevent destroy request after get timeout in ptlrpc_queue_wait.
2205 Description: replay-single.sh test 52 fails
2206 Details : A lock's skiplist need to be cleanup when it being unlinked
2207 from its resource list.
2211 Description: Short directio read returns full requested size rather than
2213 Details : Direct I/O operations should return actual amount of bytes
2214 transferred rather than requested size.
2216 Severity : enhancement
2218 Description: metadata RPC reduction (e.g. for rm performance)
2219 Details : decrease the amount of synchronous RPC between clients and servers
2220 by canceling conflicing lock before the operation on the client
2221 and packing thier handles into the main operation RPC to server.
2223 Severity : enhancement
2225 Description: Async OSC create to avoid the blocking unnecessarily.
2226 Details : If a OST has no remain object, system will block on the creating
2227 when need to create a new object on this OST. Now, ways use
2228 pre-created objects when available, instead of blocking on an
2229 empty osc while others are not empty. If we must block, we block
2230 for the shortest possible period of time.
2234 Description: improve handling recoverable errors
2235 Details : if request processig with error which can be recoverable on server
2236 request should be resend, otherwise page released from cache and
2239 Severity : enhancement
2241 Description: refine locking for avoid write wrong info into lov_objid file
2242 Details : fix possible races with add new target and write/update data in
2245 --------------------------------------------------------------------------------
2247 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2252 Frequency : on some architectures
2254 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2255 Details : Uninitialized flags sometimes cause configuration commands to
2259 Frequency : patchless clients only
2261 Description: missing __iget() symbol export
2262 Details : The __iget() symbol export is missing. To avoid the need for
2263 this on patchless clients the deathrow inode reaper is turned
2264 off, and we depend on the VM to clean up old inodes. This
2265 dependency was during via the fix for bug 12181.
2267 --------------------------------------------------------------------------------
2269 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2271 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2272 INTEROPERATE with older versions automatically. In many cases a
2273 special upgrade step is needed. Please read the
2274 user documentation before upgrading any part of a live system.
2275 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2276 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2277 * WARNING: Lustre configuration and startup changes are required with
2278 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2282 Severity : enhancement
2284 Description: MountConf
2285 Details : Lustre configuration is now managed via mkfs and mount
2286 commands instead of lmc and lconf. New obd types (MGS, MGC)
2287 are added for dynamic configuration management. See
2288 https://mail.clusterfs.com/wikis/lustre/MountConf for
2291 Severity : enhancement
2293 Description: dynamic OST addition
2294 Details : OSTs can now be added to a live filesystem
2296 Severity : enhancement
2298 Description: startup order invariance
2299 Details : MDTs and OSTs can be started in any order. Clients only
2300 require the MDT to complete startup.
2302 Severity : enhancement
2304 Description: parallel, asynchronous orphan cleanup
2305 Details : orphan cleanup is now performed in separate threads for each
2306 OST, allowing parallel non-blocking operation.
2308 Severity : enhancement
2310 Description: optimized stripe assignment
2311 Details : stripe assignments are now made based on ost space available,
2312 ost previous usage, and OSS previous usage, in order to try
2313 to optimize storage space and networking resources.
2315 Severity : enhancement
2317 Description: Permanently set tunables
2318 Details : All writable /proc/fs/lustre tunables can now be permanently
2319 set on a per-server basis, at mkfs time or on a live system.
2321 Severity : enhancement
2323 Description: Lustre message v2
2324 Details : Add lustre message format v2.
2326 Severity : enhancement
2328 Description: client OST exclusion list
2329 Details : Clients can be started with a list of OSTs that should be
2330 declared "inactive" for known non-responsive OSTs.
2334 Description: SPEC SFS validation failure on NFS v2 over lustre.
2335 Details : Changes the blocksize for regular files to be 2x RPC size,
2336 and not depend on stripe size.
2338 Severity : enhancement
2340 Description: Multiple MD RPCs in flight.
2341 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2342 To avoid overly-overloading MDS, introduce a limit on number of
2343 MDS RPCs in flight for a single client and add /proc controls
2344 to adjust this limit.
2346 Severity : enhancement
2348 Description: client read/write statistics
2349 Details : Add client read/write call usage stats for performance
2350 analysis of user processes.
2351 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2352 file access. extents_stats shows chunk size distribution.
2353 extents_stats_per_process show chunk size distribution per
2356 Severity : enhancement
2358 Description: mds statistics
2359 Details : Add detailed mds operations statistics in
2360 /proc/fs/lustre/mds/*/stats.
2364 Description: Failure of copying files with lustre special EAs.
2365 Details : Client side always return success for setxattr call for lustre
2366 special xattr (currently only "trusted.lov").
2371 Description: Refcount LNET uuids
2372 Details : The global LNET uuid list grew linearly with every startup;
2373 refcount repeated list entries instead of always adding to
2376 Severity : enhancement
2378 Description: Dynamic service threads
2379 Details : Within a small range, start extra service threads
2380 automatically when the request queue builds up.
2383 Frequency : mixed-endian client/server environments
2385 Description: mixed-endian crashes
2386 Details : The new msg_v2 system had some failures in mixed-endian
2389 Severity : enhancement
2391 Description: Easy OST removal
2392 Details : OSTs can be permanently deactivated with e.g. 'lctl
2393 conf_param lustre-OST0001.osc.active=0'
2395 Severity : enhancement
2397 Description: MGS proc entries
2398 Details : Added basic proc entries for the MGS showing what filesystems
2401 Severity : enhancement
2403 Description: provide MGS failover
2404 Details : Added config lock reacquisition after MGS server failover.
2406 Severity : enhancement
2408 Description: add Linux 2.4 support
2409 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2413 Description: a large application tries to do I/O to the same resource and dies
2414 in the middle of it.
2415 Details : Check the req->rq_arrival time after the call to
2416 ost_brw_lock_get(), but before we do anything about
2417 processing it & sending the BULK transfer request. This
2418 should help move old stale pending locks off the queue as
2419 quickly as obd_timeout.
2422 Frequency : when an incorrect nid is specified during startup
2424 Description: ptlrpc connect to non-existant node causes kernel crash
2425 Details : LNET can't be re-entered from an event callback, which
2426 happened when we expire a message after the export has been
2427 cleaned up. Instead, hand the zombie cleanup off to another
2430 Severity : enhancement
2432 Description: plain/inodebits lock performance improvement
2433 Details : Grouping plain/inodebits in granted list by their request modes
2434 and bits policy, thus improving the performance of search through
2438 Frequency : only if OST filesystem is corrupted
2440 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2441 Details : for a short time RPCs with bulk IO are in the replay list,
2442 but replay of bulk IOs is unimplemented. If the OST filesystem
2443 is corrupted due to disk cache incoherency and then replay is
2444 started it is possible to trip an assertion. Avoid putting
2445 committed RPCs into the replay list at all to avoid this issue.
2448 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2450 Description: System hang on startup
2451 Details : This bug allowed the liblustre (e.g. catamount) client to
2452 return to the app before handling all startup RPCs. This
2453 could leave the node unresponsive to lustre network traffic
2454 and manifested as a server ptllnd timeout.
2456 Severity : enhancement
2458 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2459 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2460 boolean to control whether to print peer debug info when a
2461 client's RPC times out.
2464 Frequency : only for kernels with patches from Lustre below 1.4.3
2466 Description: Remove old rdonly API
2467 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2470 Frequency : only for devices with external journals
2472 Description: Set external device read-only also
2473 Details : During a commanded failover stop, we set the disk device
2474 read-only while the server shuts down. We now also set any
2475 external journal device read-only at the same time.
2478 Frequency : when upgrading from 1.4 while trying to change parameters
2480 Description: The wrong (new) MDC name was used when setting parameters for
2481 upgraded MDT's. Also allows changing of OSC (and MDC)
2482 parameters if --writeconf is specified at tunefs upgrade time.
2485 Frequency : when setting specific ost indicies
2487 Description: QOS code breaks on skipped indicies
2488 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2489 created with --index need not be sequential.
2491 Severity : enhancement
2493 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2494 Details : The uninit_groups feature works in conjunction with the kernel
2495 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2496 up the pass1 processing of e2fsck. This is a read-only feature
2497 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2498 mount filesystems that have had this feature enabled.
2500 Severity : enhancement
2502 Description: Improve multi-block allocation algorithm to avoid fragmentation
2503 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2504 allocation locality and avoid filesystem fragmentation.
2506 ------------------------------------------------------------------------------
2508 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2510 * Support for kernels:
2511 2.4.21-47.0.1.EL (RHEL 3)
2512 2.6.5-7.283 (SLES 9)
2513 2.6.9-42.0.10.EL (RHEL 4)
2514 2.6.12.6 vanilla (kernel.org)
2515 2.6.16.27-0.9 (SLES 10)
2516 * Recommended e2fsprogs version: 1.39.cfs5
2518 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2522 Frequency : occasional, depends on client load and configuration
2523 Bugzilla : 12181, 12203
2524 Description: data loss for recently-modified files
2526 Details : In some cases it is possible that recently written or created
2527 files may not be written to disk in a timely manner (this should
2528 normally be within 30s unless client IO load is very high).
2529 The problem appears as zero-length files or files that are a
2530 multiple of 1MB in size after a client crash or client eviction
2531 that are missing data at the end of the file.
2533 This problem is more likely to be hit on clients where files are
2534 repeatedly created and unlinked in the same directory, clients
2535 have a large amount of RAM, have many CPUs, the filesystem has
2536 many OSTs, the clients are rebooted frequently, and/or the files
2537 are not accessed by other nodes after being written.
2539 The presence of the problem can be detected by looking at
2540 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2541 smaller than the second (nr_unused) then dirty files will not
2542 be flushed automatically to disk. "sync; sleep 10" should be
2543 run several times on the node before unmounting it to update
2544 Lustre (this is also safe to run on nodes without this problem).
2546 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2547 that can cause this same problem, so customers using that kernel
2548 also need to update the kernel in addition to Lustre. In order
2549 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2551 It is normal that files written just before a client crash (less
2552 than 30s) may not yet have been flushed to disk, even for local
2556 Frequency : frequent on thin XT3 nodes
2558 Description: UUID collision on thin XT3 Linux nodes
2559 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2560 randomly, since we relied on an insufficiently-seeded PRNG.
2565 Description: OSS hangs after "All ost request buffers busy"
2566 Details : A deadlock between quota and journal operations caused OSS
2567 hangs after printing "All ost request buffers busy."
2570 Frequency : always on liblustre builds
2572 Description: Cleanup compiler warnings on liblustre
2575 Frequency : always on liblustre builds on XT3
2577 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2581 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2582 > 32000 subdirectories
2583 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2584 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2585 subdirectory count crosses 32000. This will aid e2fsck to
2586 correctly handle more than 32000 subdirectories.
2589 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2591 Description: System hang on startup
2592 Details : This bug allowed the liblustre (e.g. catamount) client to
2593 return to the app before handling all startup RPCs. This
2594 could leave the node unresponsive to lustre network traffic
2595 and manifested as a server ptllnd timeout.
2597 Severity : enhancement
2599 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2600 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2601 boolean to control whether to print peer debug info when a
2602 client's RPC times out.
2607 Description: make O_SYNC working on 2.6 kernels
2608 Details : 2.6 kernels use different method for mark pages for write,
2609 so need add a code to lustre for O_SYNC work.
2614 Description: Failure to close file and release space on NFS
2615 Details : Put inode details into lock acquired in ll_intent_file_open.
2616 Use mdc_intent_lock in ll_intent_open to properly
2617 detect all kind of errors unhandled by mdc_enqueue.
2622 Description: proc file read during shutdown sometimes raced obd removal,
2624 Details : Add lock to prevent obd access after proc file removal.
2627 Frequency : Only for files larger than 4GB on 32-bit clients.
2629 Description: improperly doing page alignment of locks
2630 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2631 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2636 Description: RPCs being resent when they shouldn't be
2637 Details : Some RPCs that should not be resent are being resent. This
2638 can cause inconsistencies in the RPC state machine. Do not
2639 resend such requests.
2642 Frequency : rare, only with NFS export
2644 Description: Crash on NFS re-export node
2645 Details : under very unusual load conditions an assertion is hit in
2646 ll_intent_file_open()
2649 Frequency : only if OST filesystem is corrupted
2651 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2652 Details : for a short time RPCs with bulk IO are in the replay list,
2653 but replay of bulk IOs is unimplemented. If the OST filesystem
2654 is corrupted due to disk cache incoherency and then replay is
2655 started it is possible to trip an assertion. Avoid putting
2656 committed RPCs into the replay list at all to avoid this issue.
2661 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2662 Details : Large single O_DIRECT read and write calls can fail to allocate
2663 a sufficiently large buffer to process the request. In case of
2664 allocation failure the allocation is retried with a smaller
2665 buffer and broken into smaller requests.
2667 Severity : enhancement
2669 Description: Add -o localflock option to simulate old noflock behaviour.
2670 Details : This will achieve local-only flock/fcntl locks coherentness.
2675 Description: versioning check is incomplete
2676 Details : Checking the version difference of client vs. server, report
2677 error if the gap is too big.
2682 Description: add support PG_writeback bit
2683 Details : add support for PG_writeback bit for Lustre, for more carefull
2684 work with page cache in 2.6 kernel. This also fix some deadlocks
2685 and remove hack for work O_SYNC with 2.6 kernel.
2687 Severity : enhancement
2689 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2690 Details : The uninit_groups feature works in conjunction with the kernel
2691 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2692 up the pass1 processing of e2fsck. This is a read-only feature
2693 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2694 mount filesystems that have had this feature enabled.
2696 Severity : enhancement
2698 Description: Improve multi-block allocation algorithm to avoid fragmentation
2699 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2700 allocation locality and avoid filesystem fragmentation.
2702 ------------------------------------------------------------------------------
2704 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2706 * Support for kernels:
2707 2.6.9-42.0.3.EL (RHEL 4)
2708 2.6.5-7.276 (SLES 9)
2709 2.4.21-47.0.1.EL (RHEL 3)
2710 2.6.12.6 vanilla (kernel.org)
2711 2.6.16.21-0.8 (SLES10)
2712 * Recommended e2fsprogs version: 1.39.cfs2-0
2714 * The backwards-compatible /proc/sys/portals symlink has been removed
2715 in this release. Before upgrading, please ensure that you change
2716 any configuration scripts or /etc/sysctl.conf files that access
2717 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2718 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2719 in advance of the upgrade on any system running Lustre 1.4.6 or
2720 newer, since /proc/sys/lnet was added in that version.
2721 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2725 Frequency : only when quota is used
2727 Description: avoid scanning export list for quota master
2728 Details : Change the algorithms to avoid scanning export list in order
2729 to improve the efficiency.
2732 Frequency : MDS failover only, very rarely
2734 Description: "went back in time" messages on mds failover
2735 Details : The greatest transno may be lost when the current operation
2736 finishes with an error (transno==0) and the client's last_rcvd
2737 record is over-written. Save the greatest transno in the
2738 mds_last_transno for this case.
2741 Frequency : always for specific kernels and striping counts
2743 Description: client may get "Matching packet too big" without ACL support
2744 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2745 when trying to access files in certain configurations. The
2746 clients should in fact be denied when mounting because they do
2747 not understand ACLs.
2750 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2752 Description: many clients connecting with IO in progress causes connect timeouts
2753 Details : Avoid synchronous journal commits to avoid delays caused by many
2754 clients connecting/disconnecting when bulk IO is in progress.
2755 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2756 OST_IO_PORTAL to avoid delays behind potentially many pending
2760 Frequency : occasionally with multiple writers to a single file
2762 Description: shared writes to file may result in wrong size reported by stat()
2763 Details : Allow growing of kms when extent lock is cancelled
2766 Frequency : always with random mmap IO to multi-striped file
2768 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2769 Details : Only if the hole is at the end of OST object so that kms is too
2770 small. Fix is to increase kms accordingly in ll_nopage.
2773 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2775 Description: writes to a missing object would leak memory on the OST
2776 Details : If there is an inconsistency between the MDS and OST filesystems,
2777 such that the MDS references an object that doesn't exist, writes
2778 to that object will leak memory due to incorrect cleanup in the
2779 error handling path, eventually running out of memory on the OST.
2784 Description: Creating too long symlink causes lustre errors
2785 Details : Check symlink and name lengths before sending requests to MDS.
2788 Frequency : only if flock is enabled (not on by default)
2790 Description: posix locks not released on fd closure on 2.6.9+
2791 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2792 kernels, this caused such locks not to be released on fd close and
2793 then assertions on fs unmount about still used locks.
2796 Frequency : MDS failover only, very rarely
2798 Description: clients may get ASSERTION(granted_lock != NULL)
2799 Details : When request was taking a long time, and a client was resending
2800 a getattr by name lock request. The were multiple lock requests
2801 with the same client lock handle and
2802 mds_getattr_name->fixup_handle_for_resent_request found one of the
2803 lock handles but later failed with ASSERTION(granted_lock != NULL).
2808 Description: handle->h_buffer_credits > 0, assertion failure
2809 Details : h_buffer_credits is zero after truncate, causing assertion
2810 failure. This patch extends the transaction or creates a new
2814 Frequency : NFS re-export or patchless client
2815 Bugzilla : 11179, 10796
2816 Description: Crash on NFS re-export node (__d_move)
2817 Details : We do not want to hash the dentry if we don't have a lock.
2818 But if this dentry is later used in d_move, we'd hit uninitialised
2819 list head d_hash, so we just do this to init d_hash field but
2820 leave dentry unhashed.
2823 Frequency : NFS re-export or patchless client
2825 Description: NFS exports has problem with symbolic link
2826 Details : lustre client didn't properly install dentry when re-exported
2827 to NFS or running patchless client.
2830 Frequency : NFS re-export or patchless client
2832 Description: Various nfs/patchless fixes.
2833 Details : fixes reuse disconected alias for lookup process - this fixes
2834 warning "find_exported_dentry: npd != pd",
2835 fix permission error with open files at nfs.
2836 fix apply umask when do revalidate.
2839 Frequency : occasional
2841 Description: Crash on NFS re-export node
2842 Details : calling clear_page() on the wrong pointer triggered oops in
2843 generic_mapping_read().
2846 Frequency : rarely, using O_DIRECT IO
2848 Description: unaligned directio crashes client with LASSERT
2849 Details : check for unaligned buffers before trying any requests.
2852 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2854 Description: stale data returned from RAID cache
2855 Details : If only a small amount of IO is done to the RAID device before
2856 reading it again it is possible to get stale data from the RAID
2857 cache instead of reading it from disk.
2860 Frequency : always for sles10 kernel
2862 Description: sles10 support
2863 Details : ll_follow_link: compile fixes and using of nd_set_link
2864 under newer kernels.
2867 Frequency : depends on arch, kernel and compiler version, always on sles10
2870 Description: recursive or deep enough symlinks cause stack overflow
2871 Details : getting rid of large stack-allocated variable in
2875 Frequency : depends on hardware
2877 Description: lustre write performance loss in the SLES10 kernel
2878 Details : the performance loss is caused by using of write barriers in the
2879 ext3 code. The SLES10 kernel turns barrier support on by
2880 default. The fix is to undo that change for ldiskfs.
2882 ------------------------------------------------------------------------------
2884 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2886 * Support for kernels:
2887 2.6.9-42.0.3EL (RHEL 4)
2888 2.6.5-7.276 (SLES 9)
2889 2.4.21-47.0.1.EL (RHEL 3)
2890 2.6.12.6 vanilla (kernel.org)
2894 Frequency : quota enabled and large files being deleted
2896 Description: releasing more than 4GB of quota at once hangs OST
2897 Details : If a user deletes more than 4GB of files on a single OST it
2898 will cause the OST to spin in an infinite loop. Release
2899 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2904 Description: statfs data retrieved from /proc may be stale or zero
2905 Details : When reading per-device statfs data from /proc, in the
2906 {kbytes,files}_{total,free,avail} files, it may appear
2907 as zero or be out of date.
2910 Frequency : systems with MD RAID1 external journal devices
2912 Description: lconf's call to blkid is confused by RAID1 journal devices
2913 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2914 of returning all block devices that match the journal UUID.
2917 Frequency : always, for aggregate stripe size over 4GB
2919 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2920 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2921 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2924 Frequency : always if "lfs find" used on a local file/directory
2926 Description: "lfs find" segfaults if used on a local file/directory
2927 Details : The case where a directory component was not specified wasn't
2928 handled correctly. Handle this properly.
2931 Frequency : always on ppc64
2933 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2934 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2935 perform endianness conversion.
2938 Frequency : rarely (truncate to non-zero file size after write under load)
2939 Bugzilla : 10730, 10687
2940 Description: Files padded with zeros to next 4K multiple
2941 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2942 it is possible that files that are truncated to a non-zero size
2943 immediately after being written are filled with zero bytes beyond
2944 the truncated size. No file data is lost.
2946 Severity : enhancement
2948 Description: Allow recovery/failover for liblustre clients.
2949 Details : liblustre clients were unaware of failover configurations until
2952 Severity : enhancement
2954 Description: user file locks should fail when not mounting with flock option
2955 Details : Set up an error-returning stub in ll_file_operations.lock field
2956 to prevent incorrect behaviour when client is mounted without
2957 flock option. Also, set up properly f_op->flock field for
2961 Frequency : always on ia64
2963 Description: "lfs df" loops on printing out MDS statfs information
2964 Details : The obd_ioctl_data was not initialized and in some systems
2965 this caused a failure during the ioctl that did not return
2966 an error. Initialize the struct and return an error on failure.
2969 Frequency : SLES 9 only
2971 Description: Error of copying files with lustre special EAs as root
2972 Details : Client side always return success for setxattr call for lustre
2973 special xattr (currently only "trusted.lov").
2976 Frequency : rarely on clusters with both ia64+i386 clients
2978 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2979 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2980 clients are doing concurrent writes to the same file it is
2981 possible that the ia64 clients may LASSERT because the OST
2982 extent locks are not PAGE_SIZE aligned. Ensure that grown
2983 locks are always aligned on the request boundary.
2986 Frequency : specific use, occasional
2988 Description: Overwriting in use executable truncates on-disk binary image
2989 Details : If one node attempts to overwrite an executable in use by
2990 another node, we now correctly return ETXTBSY instead of
2991 truncating the file.
2993 Severity : enhancement
2995 Description: Async OSC create to avoid the blocking unnecessarily.
2996 Details : If a OST has no remain object, system will block on the creating
2997 when need to create a new object on this OST. Now, ways use
2998 pre-created objects when available, instead of blocking on an
2999 empty osc while others are not empty. If we must block, we block
3000 for the shortest possible period of time.
3005 Description: chmod on Lustre root is propagated to other clients
3006 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3007 invalid by the follow_mount time.
3010 Frequency : liblustre clients only
3012 Description: Race in 'instant cancel' lock handling could lead to such locks
3013 never to be granted in case of SMP MDS
3014 Details : Do not destroy not yet granted but cbpending locks in
3018 Frequency : replay/resend of open
3020 Description: non null lock assetion failure in mds_intent_policy
3021 Details : Trying to replay/resend lockless open requests resulted in
3022 mds_open() returning 0 with no lock. Now it sets a flag if
3023 a lock is going to be returned.
3025 Severity : enhancement
3027 Description: Checksum enhancements
3028 Details : New checksum enhancements allow for resending RPCs that failed
3031 Severity : enhancement
3033 Description: Tunables on number of dirty pages in cacche
3034 Details : Allow to set limit on number of dirty pages cached.
3039 Description: client crash on unmount - lock still has references
3040 Details : In some error handling cases it was possible to leak a lock
3041 reference on a client while accessing a file. This was not
3042 harmful to the client during operation, but would cause the
3043 client to crash when the filesystem is unmounted.
3046 Frequency : specific case, rare
3048 Description: ETXTBSY on mds though file not in use by client
3049 Details : ETXTBSY is no longer incorrectly returned when attempting to
3050 chmod or chown a directory that the user previously tried to
3051 execute or a currently-executing binary.
3054 Frequency : extremely rare except on liblustre-based clients
3056 Description: Lustre space not freed when files are deleted
3057 Details : Clean up open-unlinked files after client eviction. Previously
3058 the unlink was skipped and the files remained as orphans.
3063 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3064 Details : In some cases it was possible to send a blocking callback to
3065 a client doing a glimpse, even though that client didn't get
3066 a lock granted. When the glimpse lock is cancelled on the OST
3067 the freed lock is left on the waiting list and corrupted the list.
3070 Frequency : all core dumps
3072 Description: Broke core dumps to lustre
3073 Details : Negative dentry may be unhashed if parent does not have UPDATE
3074 lock, but some callers, e.g. do_coredump, expect dentry to be
3075 hashed after successful create, hash it in ll_create_it.
3077 ------------------------------------------------------------------------------
3079 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3081 * Support for kernels:
3082 2.6.9-42.0.2.EL (RHEL 4)
3083 2.6.5-7.276 (SLES 9)
3084 2.4.21-40.EL (RHEL 3)
3085 2.6.12.6 vanilla (kernel.org)
3089 Frequency : always on RHEL 3
3091 Description: Number of open files grows over time
3092 Details : The number of open files grows over time, whether or not
3093 Lustre is started. This was due to a filp leak introduced
3094 by one of our kernel patches.
3096 ------------------------------------------------------------------------------
3098 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3100 * Support for kernels:
3101 2.6.9-42.EL (RHEL 4)
3102 2.6.5-7.276 (SLES 9)
3103 2.4.21-40.EL (RHEL 3)
3104 2.6.12.6 vanilla (kernel.org)
3109 Bugzilla : 5719, 9635, 9792, 9684
3110 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3111 Details : If a server is under heavy load and cannot reply to new
3112 connection requests before the client resends the (re)connect,
3113 the connection handling code can behave badly if two service
3114 threads are concurrently handing separate (re)connections from
3115 the same client. Add better locking to the connection handling
3116 code, and ensure that only a single connection will be processed
3117 for a given client UUID, even if the lock is dropped.
3119 Severity : enhancement
3121 Description: add TCP zero-copy support to kernel
3122 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3123 sends if the hardware supports scatter-gather and checksumming.
3124 This allows socklnd to do client-write and server-read more
3125 efficiently and reduce CPU utilization from skbuf copying.
3128 Frequency : only if NFS exporting from client
3130 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3131 Details : If setting attributes on a file created under NFS that had
3132 never been opened it would be possible to oops the client
3133 if the file had no objects.
3136 Frequency : always for liblustre
3138 Description: liblustre client does MDS+OSTs setattr RPC for each write
3139 Details : When doing a write from a liblustre client, the client
3140 incorrectly issued an RPC to the MDS and each OST the file was
3141 striped over in order to update the timestamps. When writing
3142 with small chunks and many clients this could overwhelm the MDS
3143 with RPCs. In all cases it would slow down the write because
3144 these RPCs are unnecessary.
3146 Severity : enhancement
3148 Description: allow number of MDS service threads to be changed at module load
3149 Details : It is now possible to change the number of MDS service threads
3150 running. Adding "options mds mds_num_threads={N}" to the MDS's
3151 /etc/modprobe.conf will set the number of threads for the next
3152 time Lustre is restarted (assuming the "mds" module is also
3153 reloaded at that time). The default number of threads will
3154 stay the same, 32 for most systems.
3159 Description: OST crash if filesystem is unformatted or corrupt
3160 Details : If an OST is started on a device that has never been formatted
3161 or if the filesystem is corrupt and cannot even mount then the
3162 error handling cleanup routines would dereference a NULL pointer.
3167 Description: NULL pointer deref in llap_from_page.
3168 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3169 incorrect) mapping if the page was truncated/reclaimed while it was
3170 searched for. Check for this condition and skip such pages when
3171 doing readahead. Introduce extra check to llap_from_page() to
3172 verify page->mapping->host is non-NULL (so page is not anonymous).
3175 Frequency : Sometimes when using sys_sendfile
3177 Description: "page not covered by a lock" warnings from ll_readpage
3178 Details : sendfile called ll_readpage without right page locks present.
3179 Now we introduced ll_file_sendfile that does necessary locking
3180 around call to generic_file_sendfile() much like we do in
3184 Frequency : with certain MDS communication failures at client mount time
3186 Description: NULL pointer deref after failed client mount
3187 Details : a client connection request may delayed by the network layer
3188 and not be sent until after the PTLRPC layer has timed out the
3189 request. If the client fails the mount immediately it will try
3190 to clean up before the network times out the request. Add a
3191 reference from the request import to the obd device and delay
3192 the cleanup until the network drops the request.
3195 Frequency : occasionally during client (re)connect
3197 Description: assertion failure during client (re)connect
3198 Details : processing a client connection request may be delayed by the
3199 client or server longer than the client connect timeout. This
3200 causes the client to resend the connection request. If the
3201 original connection request is replied in this interval, the
3202 client may trip an assertion failure in ptlrpc_connect_interpret()
3203 which thought it would be the only running connect process.
3206 Frequency : only with obd_echo servers and clients that are rebooted
3208 Description: kernel BUG accessing uninitialized data structure
3209 Details : When running an obd_echo server it did not start the ping_evictor
3210 thread, and when a client was evicted an uninitialized data
3211 structure was accessed. Start the ping_evictor in the RPC
3212 service startup instead of the OBD startup.
3214 Severity : enhancement
3215 Bugzilla : 10193 (patchless)
3216 Description: Remove dependency on various unexported kernel interfaces.
3217 Details : No longer need reparent_to_init, exit_mm, exit_files,
3218 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3221 Frequency : rare (only users of deprecated and unsupported LDAP config)
3223 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3224 Details : LDAP apparently contains 'lustreName' attributes instead of
3225 'name'. A simple remapping of the name is sufficient.
3228 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3230 Description: waiting_locks_callback trips kernel BUG if client is evicted
3231 Details : Running with the dump_on_timeout debug flag turned on makes
3232 it possible that the waiting_locks_callback() can try to dump
3233 the Lustre kernel debug logs from an interrupt handler. Defer
3234 this log dumping to the expired_lock_main() thread.
3236 Severity : enhancement
3238 Description: Support NFS exporting on 2.6 kernels.
3239 Details : Implement non-rawops metadata methods for NFS server to use without
3240 changing NFS server code.
3243 Frequency : very rare (synthetic metadata workload only)
3245 Description: two racing renames might cause an MDS thread to deadlock
3246 Details : Running the "racer" program may cause one MDS thread to rename
3247 a file from being the source of a rename to being the target of
3248 a rename at exactly the same time that another thread is doing
3249 so, and the second thread has already enqueued these locks after
3250 doing a lookup of the target and is trying to relock them in
3251 order. Ensure that we don't try to re-lock the same resource.
3254 Frequency : only very large systems with liblustre clients
3256 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3257 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3258 all OSTs in parallel. This allows the eviction of stale liblustre
3259 clients to proceed much faster than if they were done in series,
3260 and also offers similar improvements for other set_info RPCs.
3265 Description: excessive CPU usage during initial read phase on client
3266 Details : During the initial read phase on a client, it would agressively
3267 retry readahead on the file, consuming too much CPU and impacting
3268 performance (since 1.4.5.8). Improve the readahead algorithm
3269 to avoid this, and also improve some other common cases (read
3270 of small files in particular, where "small" is files smaller than
3271 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3276 Description: MDS crash when receiving packet with unknown intent.
3277 Details : Do not LBUG in unknown intent case, just return -EFAULT
3279 Severity : enhancement
3280 Bugzilla : 9293, 9385
3281 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3282 Details : Do not serialize getattr (non-intent version) and statfs.
3285 Frequency : occasional, when OST network is overloaded/intermittent
3287 Description: client evicted by OST after bulk IO timeout
3288 Details : If a client sends a bulk IO request (read or write) the OST
3289 may evict the client if it is unresposive to its data GET/PUT
3290 request. This is incorrect if the network is overloaded (takes
3291 too long to transfer the RPC data) or dropped the OST GET/PUT
3292 request. There is no need to evict the client at all, since
3293 the pinger and/or lock callbacks will handle this, and the
3294 client can restart the bulk request.
3297 Frequency : Always when mmapping file with no objects
3299 Description: client crashes when mmapping file with no objects
3300 Details : Check that we actually have objects in a file before doing any
3301 operations on objects in ll_vm_open, ll_vm_close and
3307 Description: Request leak when working with deleted CWD
3308 Details : Introduce advanced request refcount tracking for requests
3309 referenced from lustre intent.
3311 Severity : Enhancement
3313 Description: Cache open file handles on client.
3314 Details : MDS now will return special lock along with openhandle, if
3315 requested and client is allowed to hold openhandle, even if unused,
3316 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3317 closing files for every read/write openration.
3319 Severity : Enhancement
3321 Description: Cache open negative dentries on client when possible.
3322 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3323 negative dentries on lock revocation.
3328 Description: Remounting a client read-only wasn't possible with a zconf mount
3329 Details : It wasn't possible to remount a client read-only with llmount.
3331 Severity : enhancement
3332 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3333 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3334 MPICH is included to add Lustre-specific ADIO interfaces.
3335 This is based closely on the UFS ADIO layer and only differs
3336 in file creation, in order to allow the OST striping to be set.
3337 This is user-contributed code and not supported by CFS.
3342 Description: extended inode attributes (immutable, append-only) work improperly
3343 when 2.4 and 2.6 kernels are used on client/server or vice versa
3344 Details : Introduce kernel-independent values for these flags.
3346 Severity : enhancement
3349 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3350 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3351 granularity. Now, Fractional MB granularity is be supported,
3352 this is very useful for low memory system.
3354 Severity : enhancement
3356 Description: Getattr by fid
3357 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3358 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3359 have locks and avoids getattr rpc requests for it on every stat.
3362 Frequency : Always, for filesystems larger than 2TB
3364 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3365 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3366 filesystems larger than 2TB to be mounted with mballoc enabled.
3369 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3371 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3372 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3373 filesystems larger than 2TB is mounted then the kernel will
3374 silently corrupt the start of the filesystem. CONFIG_LBD is
3375 enabled for all CFS-supported kernels, but the possibility of
3376 this happening with a modified kernel config exists.
3378 Severity : enhancement
3380 Description: add client O_DIRECT support for 2.6 kernels
3381 Details : It is now possible to do O_DIRECT reads and writes to files
3382 in the Lustre client mountpoint on 2.6 kernel clients.
3384 Severity : enhancement
3386 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3387 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3388 parallel, not waiting for response from every OST before sending
3389 a rpc to the next OST.
3394 Description: setattr vs write race when updating file timestamps
3395 Details : Client processes that update a file timestamp into the past
3396 right after writing to the file (e.g. tar) it is possible that
3397 the updated file modification time can be reset to the current
3398 time due to a race between processing the setattr and write RPC.
3400 Severity : enhancement
3402 Description: Bring 'lfs find' closer in line with regular Linux find.
3403 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3404 -print0 options and obtains all the needed info through the lustre
3407 Severity : enhancement
3409 Description: support up to 1024 configured devices on one node
3410 Details : change obd_dev array from statically allocated to dynamically
3411 allocated structs as they are first used to reduce memory usage
3416 Description: Flush dirty partially truncated pages during truncate
3417 Details : Immediatelly flush partially truncated pages in filter_setattr,
3418 this way we completely avoid having any pages in page cache on OST
3419 and can retire ugly workarounds during writes to flush such pages.
3424 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3425 Details : protect inode from truncation within vfs_unlink() context
3426 just take a reference before calling vfs_unlink() and release it
3427 when parent's i_sem is free.
3432 Description: last_id value checked outside lock on OST caused LASSERT failure
3433 Details : If there were multiple MDS->OST object precreate requests in
3434 flight, it was possible that the OST's last object id was checked
3435 outside a lock and incorrectly tripped an assertion. Move checks
3436 inside locks, and discard old precreate requests.
3439 Frequency : always, if extents are used on OSTs
3441 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3442 Details : OSTs using the extents format would not zero the high 16 bits of
3443 the index physical block number. This is not a problem for any
3444 OST filesystems smaller than 16TB, and no kernels support ext3
3445 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3446 new/modified files) and can be fixed for existing filesystems
3447 with e2fsprogs-1.39-cfs1.
3452 Description: import connection selection may be incorrect if timer wraps
3453 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3454 import connections to be ignored if the 32-bit jiffies counter
3455 wraps. Use a 64-bit jiffies counter.
3458 Frequency : very large clusters immediately after boot
3460 Description: LNET request buffers exhausted under heavy short-term load
3461 Details : If a large number of client requests are generated on a service
3462 that has previously never seen so many requests it is possible
3463 that the request buffer growth cannot keep up with the spike in
3464 demand. Instead of dropping incoming requests, they are held in
3465 the LND until the RPC service can accept more requests.
3468 Frequency : Sometimes during replay
3470 Description: Assertion failure in ll_local_open after replay.
3471 Details : If replay happened on an open request reply before we were able
3472 to set replay handler, reply will become not swabbed tripping the
3473 assertion in ll_local_open. Now we set the handler right after
3474 recognising of open request
3477 Frequency : very rare
3479 Description: kernel reports "badness in vsnprintf"
3480 Details : Reading from the "recovery_status" /proc file in small chunks
3481 may cause a negative length in lprocfs_obd_rd_recovery_status()
3482 call to vsnprintf() (which is otherwise harmless). Exit early
3483 if there is no more space in the output buffer.
3485 Severity : enhancement
3487 Description: clear OBD RPC statistics by writing to them
3488 Details : It is now possible to clear the OBD RPC statistics by writing
3489 to the "stats" file.
3494 Description: Client mtime is not the same on different clients after utimes
3495 Details : In some cases, the client was using the utimes() syscall on
3496 a file cached on another node. The clients now validate the
3497 ctime from the MDS + OSTs to determine which one is right.
3502 Description: Inability to activate failout mode
3503 Details : lconf script incorrectly assumed that in pythong string's numeric
3504 value is used in comparisons.
3507 Frequency : always with multiple stripes per file
3509 Description: Inefficient object allocation for mutli-stripe files
3510 Details : When selecting which OSTs to stripe files over, for files with
3511 a stripe count that divides evenly into the number of OSTs,
3512 the MDS is always picking the same starting OST for each file.
3513 Return the OST selection heuristic to the original design.
3518 Description: mount failures may take full timeout to return an error
3519 Details : Under some heavy load conditions it is possible that a
3520 failed mount can wait for the full obd_timeout interval,
3521 possibly several minutes, before reporting an error.
3522 Instead return an error as soon as the status is known.
3524 Frequency : quota enabled and large files being deleted
3526 Description: releasing more than 4GB of quota at once hangs OST
3527 Details : If a user deletes more than 4GB of files on a single OST it
3528 will cause the OST to spin in an infinite loop. Release
3529 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3534 Description: statfs data retrieved from /proc may be stale or zero
3535 Details : When reading per-device statfs data from /proc, in the
3536 {kbytes,files}_{total,free,avail} files, it may appear
3537 as zero or be out of date.
3540 Frequency : systems with MD RAID1 external journal devices
3542 Description: lconf's call to blkid is confused by RAID1 journal devices
3543 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3544 of returning all block devices that match the journal UUID.
3547 Frequency : always, for aggregate stripe size over 4GB
3549 Description: assertion fails when trying to use 4GB stripe size
3550 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3551 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3554 Frequency : always on ppc64
3556 Description: the first write on an ext3 filesystem with mballoc got stuck
3557 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3558 perform endianness conversion.
3560 ------------------------------------------------------------------------------
3562 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3564 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3565 INTEROPERATE with older versions automatically. Please read the
3566 user documentation before upgrading any part of a live system.
3567 * WARNING: Lustre networking configuration changes are required with
3568 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3571 * Support for kernels:
3572 2.6.9-22.0.2.EL (RHEL 4)
3573 2.6.5-7.244 (SLES 9)
3574 2.6.12.6 vanilla (kernel.org)
3577 Severity : enhancement
3578 Bugzilla : 7981/8208
3579 Description: Introduced Lustre Networking (LNET)
3580 Details : LNET is new networking infrastructure for Lustre, it includes
3581 a reorganized network configuration mode (see the user
3582 documentation for full details) as well as support for routing
3583 between different network fabrics. Lustre Networking Devices
3584 (LNDs) for the supported network fabrics have also been
3585 created for this new infrastructure.
3587 Severity : enhancement
3588 Description: Introduced Access control lists
3589 Details : clients can set ACLs on files and directories in order to have
3590 more fine-grained permissions than the standard Unix UGO+RWX.
3591 The MDS must be started with the "-o acl" mount option.
3593 Severity : enhancement
3594 Description: Introduced filesystem quotas
3595 Details : Administrators may now establish per-user quotas on the
3598 Severity : enhancement
3600 Description: Configuration change for the XT3
3601 The PTLLND is now used to run Lustre over Portals on the XT3
3602 The configure option(s) --with-cray-portals are no longer used.
3603 Rather --with-portals=<path-to-portals-includes> is used to
3604 enable building on the XT3. In addition to enable XT3 specific
3605 features the option --enable-cray-xt3 must be used.
3610 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3611 Details : A race between allocating a new llcd and re-getting the llcd_lock
3612 allowed another thread to grab newly-allocated llcd.
3614 Severity : enhancement
3616 Description: 2.6 OST async journal commit and locking fix to improve performance
3617 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3618 2.6 kernels are now async as they already were in 2.4 kernels so
3619 that they can commit concurrently with the network bulk transfer.
3620 For block-allocated files the filter allocation semaphore is held
3621 to avoid filesystem fragmentation during allocation. BKL lock
3622 removed for 2.6 xattr operations where it is no longer needed.
3627 Description: lconf incorrectly determined whether two IP networks could talk
3628 Details : In some more complicated routing and multiple-network
3629 configurations, lconf will avoid trying to make a network
3630 connection to a disjoint part of the IP space. It was doing the
3631 math incorrectly for one set of cases.
3636 Description: Fix for potential infinite loop processing records in an llog.
3637 Details : If an llog record is corrupted/zeroed, it is possible to loop
3638 forever in llog_process(). Validate the llog record length
3639 and skip the remainder of the block on error.
3642 Frequency : occasional (liblustre only)
3644 Description: liblustre could not open files whose last component is a symlink
3645 Details : sysio_path_walk() would incorrectly pass the open intent to
3646 intermediate path components.
3649 Frequency : rare (liblustre only with non-standard tuning)
3650 Bugzilla : 7201 (7350)
3651 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3652 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3653 after it has been released, only do this for referenced locks
3655 Severity : enhancement
3657 Description: specifying an (invalid) directory default stripe_size of -1
3658 would reset the directory default striping
3659 Details : stripe_size -1 was used internally to signal directory stripe
3660 removal, now use "all default" to signal dir stripe removal
3661 as a directory striping of "all default" is not useful
3664 Frequency : common for large clusters running liblustre clients
3666 Description: doing an ls when liblustre clients are running is slow
3667 Details : sending a glimpse AST to a liblustre client waits for every AST
3668 to time out, as liblustre clients will not respond. Since they
3669 cannot cache data we refresh the OST lock LVB from disk instead.
3671 Severity : enhancement
3673 Description: doing an ls at the same time as file IO can be slow
3674 Details : enqueue and other "small" requests can be blocked behind many
3675 large IO requests. Create a new OST IO portal for non-IO
3676 requests so they can be processed faster.
3679 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3681 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3682 created file with mode 000
3683 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3684 allow an ftruncate() as a non-root user to a file with mode 000.
3685 Lustre now respects this flag to disable mode checks when
3686 truncating a file owned by the user
3689 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3692 Description: Revoking locks from clients that went dead or catatonic might take
3694 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3695 cancellation of such locks instant on servers without waiting for
3696 any reply from clients. Clients drops these locks when cancel
3697 notification from server is received without replying.
3700 Frequency : liblustre-only, when liblustre client dies or becomes busy
3702 Description: Doing ls on Linux clients can take a long time with active
3704 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3705 granting such locks to it in the first place if possible. Locks
3706 are taken by proxy on the OST during the read or write and
3707 dropped immediately afterward. Add connect flags handling, do
3708 not grant locks to liblustre clients for glimpse ASTs.
3710 Severity : enhancement
3712 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3713 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3714 sequential reads and behaves badly with applications doing random
3715 reads. Improve it by reading ahead at least read region, and
3716 avoiding excessive large RPC for small reads.
3718 Severity : enhancement
3720 Description: Creating more than 1000 files for a single job may cause a load
3721 imbalance on the OSTs if there are also a large number of OSTs.
3722 Details : qos_prep_create() uses an OST index reseed value that is an
3723 even multiple of the number of available OSTs so that if the
3724 reseed happens in the middle of the object allocation it will
3725 still utilize the OSTs as uniformly as possible.
3730 Description: OST or MDS may oops in ping_evictor_main()
3731 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3732 but doesn't restart at beginning of obd_exports_timed list
3735 Severity : enhancement
3737 Description: improve by-nid export eviction on the MDS and OST
3738 Details : allow multiple exports with the same NID to be evicted at one
3739 time without re-searching the exports list.
3742 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3744 Description: MDS may oops in groups_free()
3745 Details : in rare race conditions a newly allocated group_info struct is
3746 freed again, and this can be NULL. The 2.4 compatibility code
3747 for groups_free() checked for a NULL pointer, but 2.6 did not.
3750 Frequency : common for liblustre clients doing little filesystem IO
3751 Bugzilla : 9352, 7313
3752 Description: server may evict liblustre clients accessing contended locks
3753 Details : if a client is granted a lock or receives a completion AST
3754 with a blocking AST already set it would not reply to the AST
3755 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3758 Frequency : lfs setstripe, only systems with more than 160 OSTs
3760 Description: unable to set striping with a starting offset beyond OST 160
3761 Details : llapi_create_file() incorrectly limited the starting stripe
3762 index to the maximum single-file stripe count.
3765 Frequency : LDAP users only
3767 Description: lconf did not handle in-kernel recovery with LDAP properly
3768 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3770 Severity : enhancement
3772 Description: bind OST threads to NUMA nodes to improve performance
3773 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3774 node and do their allocations there to localize memory access
3776 Severity : enhancement
3778 Description: llmount can determine client NID directly from Myrinet (GM)
3779 Details : the client NID code from gmnalnid was moved directly into
3780 llmount, removing the need to use this or specifying the
3781 client NID explicitly when mounting GM clients with zeroconf
3784 Frequency : if client is started with down MDS
3786 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3787 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3788 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3789 ptlrpc_import_delayed_req() because we didn't check if the
3790 request was interrupted, and we also didn't break out of the
3791 event loop if there was no timeout
3796 Description: data loss during non-page-aligned writes to a single file from
3797 both multiple nodes and multiple threads on one node at same time
3798 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3799 inconsistency led to false short-reads, that were cached and later
3800 used by ->prepare_write() to fill in partially written page,
3801 leading to data loss.
3804 Frequency : always, if lconf --abort_recovery used
3806 Description: lconf --abort_recovery fails with 'Operation not supported'
3807 Details : lconf was attempting to abort recovery on the MDT device and not
3810 Severity : enhancement
3812 Description: remove cleanup logs
3813 Details : replace lconf-generated cleanup logs with lustre internal
3814 cleanup routines. Eliminates the need for client-cleanup and
3817 Severity : enhancement
3819 Description: add support for EAs (user and system) on lustre filesystems
3820 Details : it is now possible to store extended attributes in the Lustre
3821 client filesystem, and with the user_xattr mount option it
3822 is possible to allow users to store EAs on their files also
3824 Severity : enhancement
3826 Description: Add possibility (config option) to show minimal available OST free
3828 Details : When compiled with --enable-mindf configure option, statfs(2)
3829 (and so, df) will return least minimal free space available from
3830 all OSTs as amount of free space on FS, instead of summary of
3831 free spaces of all OSTs.
3833 Severity : enhancement
3835 Description: do not expand extent locks acquired on OST-side
3836 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3837 by server: they are not cached anyway.
3840 Frequency : when mmap is used/binaries executed from Lustre
3842 Description: Unmmap pages before throwing them away from read cache.
3843 Details : llap_shrink cache now attempts to unmap pages before discarding
3844 them (if unmapping failed - do not discard). SLES9 kernel has
3845 extra checks that trigger if this unmapping is not done first.
3850 Description: lconf didn't resolve symlinks before checking to see whether a
3851 given mountpoint was already in use
3854 Frequency : when migrating failover services
3855 Bugzilla : 6395, 9514
3856 Description: When migrating a subset of services from a node (e.g. failback
3857 from a failover service node) the remaining services would
3858 time out and evict clients.
3859 Details : lconf --force (implied by --failover) sets the global obd_timeout
3860 to 5 seconds in order to quickly disconnect, but this caused
3861 other RPCs to time out too quickly. Do not change the global
3862 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3864 Severity : enhancement
3865 Frequency : if MDS is started with down OST
3866 Bugzilla : 9439,5706
3867 Description: Allow startup/shutdown of an MDS without depending on the
3868 availability of the OSTs.
3869 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3870 Add appropriate locking and lov-osc refcounts for safe
3871 cleaning. Add osc abort_inflight calls in case the
3872 synchronize never started.
3875 Frequency : occasional (Cray XT3 only)
3877 Description: root not authorized to access files in CRAY_PORTALS environment
3878 Details : The client process capabilities were not honoured on the MDS in
3879 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3880 been accessed by an authorized user then root was able to access
3881 the file on the local client also. The root user capabilities
3882 are now allowed on the MDS, as this environment has secure UID.
3885 Frequency : occasional
3887 Description: ldiskfs "too long searching" message happens too often
3888 Details : A debugging message (otherwise harmless) prints too often on
3889 the OST console. This has been reduced to only happen when
3890 there are fragmentation problems on the filesystem.
3895 Description: Division by zero in statfs when all OSCs are inactive
3896 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3897 lov_statfs divides by value returned by lov_get_stripecnt().
3901 Bugzilla : 9489, 3273
3902 Description: First write from each client to each OST was only 4kB in size,
3903 to initialize client writeback cache, which caused sub-optimal
3904 RPCs and poor layout on disk for the first writen file.
3905 Details : Clients now request an initial cache grant at (re)connect time
3906 and so that they can start streaming writes to the cache right
3907 away and always do full-sized RPCs if there is enough data.
3908 If the OST is rebooted the client also re-establishes its grant
3909 so that client cached writes will be honoured under the grant.
3914 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3915 Details : Now I/O RPCs go to different portal number and (presumably) fast
3916 lock requests (and glimses) and other RPCs get their own service
3917 threads pool that should be able to service those RPCs
3920 Severity : enhancement
3922 Description: Ability to exchange lustre version between client and servers and
3923 issue warnings at client side if client is too old. Also for
3924 liblustre clients there is ability to refuse connection of too old
3926 Details : New 'version' field is added to connect data structure that is
3927 filled with version info. That info is later checked by server and
3931 Frequency : rare, liblustre only.
3932 Bugzilla : 9296, 9581
3933 Description: Two simultaneous writes from liblustre at offset within same page
3934 might proceed at the same time overwriting eachother with stale
3936 Details : I/O lock withing llu_file_prwv was released too early, before data
3937 actually was hitting the wire. Extended lock-holding time until
3938 server acknowledges receiving data.
3941 Frequency : extremely rare. Never observed in practice.
3943 Description: avoid generating lustre_handle cookie of 0.
3944 Details : class_handle_hash() generates handle cookies by incrementing
3945 global counter, and can hit 0 occasionaly (this is unlikely, but
3946 not impossible, because initial value of cookie counter is
3947 selected randonly). Value of 0 is used as a sentinel meaning
3948 "unassigned handle" --- avoid it. Also coalesce two critical
3949 sections in this function into one.
3951 Severity : enhancement
3953 Description: allow liblustre clients to delegate truncate locking to OST
3954 Details : To avoid overhead of locking, liblustre client instructs OST to
3955 take extent lock in ost_punch() on client's behalf. New connection
3956 flag is added to handle backward compatibility.
3958 Severity : enhancement
3959 Bugzilla : 4928, 7341, 9758
3960 Description: allow number of OST service threads to be specified
3961 Details : a module parameter allows the number of OST service threads
3962 to be specified via "options ost ost_num_threads={N}" in the
3963 OSS's /etc/modules.conf or /etc/modprobe.conf.
3967 Bugzilla : 6146, 9635, 9895
3968 Description: servers crash with bad pointer in target_handle_connect()
3969 Details : In rare cases when a client is reconnecting it was possible that
3970 the connection request was the last reference for that export.
3971 We would temporarily drop the export reference and get a new
3972 one, but this may have been the last reference and the export
3973 was just destroyed. Get new reference before dropping old one.
3975 Severity : enhancement
3976 Frequency : if client is started with failover MDS
3978 Description: Allow multiple MDS hostnames in the mount command
3979 Details : Try to read the configuration from all specified MDS
3980 hostnames during a client mount in case the "primary"
3983 Severity : enhancement
3985 Description: Stop sending data to evicted clients as soon as possible.
3986 Details : Check if the client we are about to send or are sending data to
3987 was evicted already. (Check is done every second of waiting,
3988 for which l_wait_event interface was extended to allow checking
3989 of exit condition at specified intervals).
3992 Frequency : rare, normally only when NFS exporting is done from client
3994 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3996 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3999 Frequency : always, if srand() is called before liblustre initialization
4001 Description: Liblustre uses system PRNG disturbing its usage by user application
4002 Details : Introduce internal to lustre fast and high-quality PRNG for
4003 lustre usage and make liblustre and some other places in generic
4004 lustre code to use it.
4006 Severity : enhancement
4007 Bugzilla : 9477, 9557, 9870
4008 Description: Verify that the MDS configuration logs are updated when xml is
4009 Details : Check if the .xml configuration logs are newer than the config
4010 logs stored on the MDS and report an error if this is the case.
4011 Request --write-conf, or allow starting with --old_conf.
4013 Severity : enhancement
4015 Description: Handle symlinks in the path when checking if Lustre is mounted.
4016 Details : Resolve intermediate symlinks when checking if a client has
4017 mounted a filesystem to avoid duplicate client mounts.
4022 Description: lconf can hit an error exception but still return success.
4023 Details : The lconf command catches the Command error exception at the top
4024 level script context and will exit with the associated exit
4025 status, but doesn't ensure that this exit status is non-zero.
4030 Description: failure of ptlrpc thread startup can cause oops
4031 Details : Starting a ptlrpc service thread can fail if there are a large
4032 number of threads or the server memory is very fragmented.
4033 Handle this without oopsing.
4036 Frequency : always, only if liblustre and non-default acceptor port was used
4038 Description: liblustre cannot connect to servers with non-default acceptor port
4039 Details : tcpnal_set_default_params() was not called and was therefore
4040 ignoring the environment varaible TCPNAL_PORT, as well as other
4041 TCPNAL_ environment variables
4046 Description: two objects could be created on the same OST for a single file
4047 Details : If an OST is down, in some cases it was possible to create two
4048 objects on a single OST for a single file. No problems other
4049 than potential performance impact and spurious error messages.
4053 Bugzilla : 5681, 9562
4054 Description: Client may oops in ll_unhash_aliases
4055 Details : Client dcache may become inconsistent in race condition.
4056 In some cases "getcwd" can fail if the current directory is
4062 Description: Inode refcounting problems in NFS export code
4063 Details : link_raw functions used to call d_instantiate without obtaining
4064 extra inode reference first.
4068 Bugzilla : 9942, 9903
4069 Description: Referencing freed requests leading to crash, memleaks with NFS.
4070 Details : We used to require that call to ll_revalidate_it was always
4071 followed by ll_lookup_it. Also with revalidate_special() it is
4072 possible to call ll_revalidate_it() twice for the same dentry
4073 even if first occurence returned success. This fix changes semantic
4074 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4075 reference on a request referred from the intent.
4076 ll_intent_release() then releases such a request.
4079 Frequency : rare, normally benchmark loads only
4081 Description: unlinked inodes were kept in memory on the client
4082 Details : If a client is repeatedly creating and unlinking files it
4083 can accumulate a lot of stale inodes in the inode slab cache.
4084 If there is no other client load running this can cause the
4085 client node to run out of memory. Instead flush old inodes
4086 from client cache that have the same inode number as a new inode.
4089 Frequency : SLES9 2.6.5 kernel and long filenames only
4090 Bugzilla : 9969, 10379
4091 Description: utime reports stale NFS file handle
4092 Details : SLES9 uses out-of-dentry names in some cases, which confused
4093 the lustre dentry revalidation. Change it to always use the
4097 Frequency : rare, unless heavy write-truncate concurrency is continuous
4098 Bugzilla : 4180, 6984, 7171, 9963, 9331
4099 Description: OST becomes very slow and/or deadlocked during object unlink
4100 Details : filter_destroy() was holding onto the parent directory lock
4101 while truncating+unlinking objects. For very large objects this
4102 may block other threads for a long time and slow overall OST
4103 responsiveness. It may also be possible to get a lock ordering
4104 deadlock in this case, or run out of journal credits because of
4105 the combined truncate+unlink. Solution is to do object truncate
4106 first in one transaction without parent lock, and then do the
4107 final unlink in a new transaction with the parent lock. This
4108 reduces the lock hold time dramatically.
4111 Frequency : rare, 2.4 kernels only
4113 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4114 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4115 before unmounting their filesystem in order to avoid deadlock.
4116 The kernel_locked() function in 2.4 kernels only checks whether
4117 the kernel lock is held, not whether it is this process that is
4118 holding it as 2.6 kernels do.
4123 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4124 Details : The client ptlrpc code may be trying to reconnect to a down
4125 server before a previous connection attempt has timed out.
4126 Increase the reconnect interval to be longer than the connection
4127 timeout interval to avoid sending duplicate connections to
4131 Frequency : echo_client brw_test command
4133 Description: fix echo_client to work with OST preallocated code
4134 Details : OST preallocation code (5137) didn't take echo_client IO path
4135 into account: echo_client calls filter methods outside of any
4136 OST thread and, hence, there is no per-thread preallocated
4137 pages and buffers to use. Solution: hijack pga pages for IO. As
4138 a byproduct, this avoids unnecessary data copying.
4142 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4143 Description: Client can oops in mdc_commit_close() after open replay
4144 Details : It was possible for the MDS to return an open request with no
4145 transaction number in mds_finish_transno() if the client was
4146 evicted, but without actually returning an error. Clients
4147 would later try to replay that open and may trip an assertion
4148 Simplify the client close codepath, and always return an error
4149 from the MDS in case the open is not successful.
4152 Frequency : rare, 2.6 OSTs only
4154 Description: OST may deadlock under high load on fragmented files
4155 Details : If there was a heavy load and highly-fragmented OST filesystems
4156 it was possible to have all the OST threads deadlock waiting on
4157 allocation of biovecs, because the biovecs were not released
4158 until the entire RPC IO was completed. Instead, release biovecs
4159 as soon as they are complete to ensure forward IO progress.
4161 Severity : enhancement
4163 Description: Support for specifying external journal device at mount
4164 Details : If an OST or MDS device is formatted with an external journal
4165 device, this device major/minor is stored in the ext3 superblock
4166 and may not be valid for failover. Allow detecting and
4167 specifying the external journal at mount time.
4172 Description: Mounting an MDS with pending unlinked files may cause oops
4173 Details : target_finish_recovery() calls mds_postrecov() which returned
4174 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4175 considers this an error and immediately begins cleaning up the
4176 lov, just after starting the mds_lov process
4178 Severity : enhancement
4180 Description: Implement 'lfs df' to report actual free space on per-OST basis
4181 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4182 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4183 usage of objects; '-h' to report in human readable format.
4185 ------------------------------------------------------------------------------
4187 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4194 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4195 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4196 at startup during orphan unlinking. Instead, skip buddy update
4197 before setup, as it will be regenerated after recovery is complete.
4202 Description: If an OST is inactive, its locks might reference stale inodes.
4203 Details : lov_change_cbdata() must iterate over all namespaces, even if
4204 they are inactive to clear inode references from the lock.
4206 Severity : enhancement
4207 Frequency : occasional, if non-standard max_dirty_mb used
4209 Description: Client will block write RPCs if not enough grant
4210 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4211 then the client will block writes while waiting for another RPC
4212 to complete instead of consuming its dirty limit. With change
4213 we get improved performance when max_dirty_mb is small.
4215 Severity : enhancement
4216 Bugzilla : 3389, 6253
4217 Description: Add support for supplementary groups on the MDS.
4218 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4219 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4220 lookups for user supplementary groups into a cache.
4224 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4225 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4226 for liblustre O_EXCL is enforced by MDS.
4230 Description: readdir never returns NULL in liblustre.
4231 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4232 dentry in struct dirent.
4236 Frequency : liblustre only, depends on application IO pattern
4237 Description: liblustre clients evicted if not contacting servers
4238 Details : Don't put liblustre clients into the ping_evictor list, so
4239 they will not be evicted by the pinger ever.
4241 Severity : enhancement
4243 Description: Add ability to evict clients by NID from MDS.
4244 Details : By echoing "nid:$NID" string into
4245 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4246 $NID would be instantly evicted from this MDS and from all active
4247 OSTs connected to it.
4251 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4252 Details : lookup_it_finish() used to query file size from OSTs that was not
4257 Description: service threads change working directory to that of init
4258 Details : Starting lustre service threads may pin the working directory
4259 of the parent thread, making that filesystem busy. Threads
4260 now change to the working directory of init to avoid this.
4264 Frequency : during shutdown only
4265 Description: shutdown with a failed MDS or OST can cause unmount to hang
4266 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4271 Frequency : 2.6 only
4272 Description: chmod/chown may include an extra supplementary group
4273 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4274 supplementary group and if none is specified this is used.
4277 Bugzilla : 5479 (6816)
4279 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4280 Details : If lookup is in progress on a file that is unlinked we might try
4281 to revalidate the inode and fail in revalidate after lookup is
4282 complete and ll_file_open() enqueues the open again but
4283 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4286 Frequency : always, if lconf --abort_recovery used
4288 Description: lconf --abort_recovery fails with 'Operation not supported'
4289 Details : lconf was attempting to abort recovery on the MDT device and not
4292 ------------------------------------------------------------------------------
4294 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4299 Frequency : rare (only unsupported configurations with a node running as an
4301 Bugzilla : 6514, 5137
4302 Description: Mounting a Lustre file system on a node running as an OST could
4304 Details : OSTs now preallocates memory needed to write out data at
4305 startup, instead of when needed, to avoid having to
4306 allocate memory in possibly low memory situations.
4307 Specifically, if the file system is mounted on on OST,
4308 memory pressure could force it to try to write out data,
4309 which it needed to allocate memory to do. Due to the low
4310 memory, it would be unable to do so and the node would
4311 become unresponsive.
4313 Severity : enhancement
4315 Description: Addition of lconf --service command line option
4316 Details : lconf now accepts a '--service <arg>' option, which is
4317 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4319 Severity : enhancement
4321 Description: Failover mode is now the default for OSTs.
4322 Details : By default, OSTs will now run in failover mode. To return to
4323 the old behaviour, add '--failout' to the lmc line for OSTs.
4325 Severity : enhancement
4327 Description: Health checks are now provided for MDS and OSTs
4328 Details : Additional detailed health check information on MSD and OSTs
4329 is now provided through the procfs health_check value.
4332 Frequency : occasional, depends on IO load
4334 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4335 after numerous create/delete cycles
4336 Details : The ext3 inode allocation policy would not allocate new inodes
4337 very well on the OSTs because there are no new directories
4338 being created. Instead we look for groups with free space if
4339 the parent directories are nearly full.
4344 Description: Network or server problems during mount may cause partially
4345 mounted clients instead of returning an error.
4346 Details : The config llog parsing code may overwrite the error return
4347 code during mount error handling, returning success instead
4353 Description: MDS can fail to allocate large reply buffers
4354 Details : After long uptimes the MDS can fail to allocate large reply
4355 buffers (e.g. zconf client mount config records) due to memory
4356 fragmentation or consumption by the buffer cache. Preallocate
4357 some large reply buffers so that these replies can be sent even
4358 under memory pressure.
4362 Frequency : rare (liblustre)
4363 Description: fsx running with liblustre complained that using truncate() to
4364 extend the file doesn't work. This patch corrects that issue.
4365 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4366 ATTR_SIZE and lsm use in llu_setattr_raw.
4370 Frequency : rare, only 2.6 kernels
4371 Description: Unusual file access patterns on the MDS may result in inode
4372 data being lost in very rare circumstances.
4373 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4374 code in the 2.6 kernel caused the inode and/or EA data not to
4375 be read from disk, causing single-file corruption.
4379 Frequency : rare, only 2.6 filesystems using extents
4380 Description: Heavy concurrent write and delete load may cause data corruption.
4381 Details : It was possible under high-load situations to have an extent
4382 metadata block in the block device cache from a just-unlinked
4383 file overwrite a newly-allocated data block. We now unmap any
4384 metadata buffers that alias just-allocated data blocks.
4388 Frequency : filesystems with default stripe_count larger than 77
4389 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4390 Details : lconf specifies an inode size of 4096 bytes when the default
4391 stripe_count is larger than 77. This conflicts with the default
4392 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4393 this case to avoid pinning too much memory for large EAs.
4395 ------------------------------------------------------------------------------
4397 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4402 Frequency : rare (extremely heavy IO load with hundreds of clients)
4404 Description: Client is evicted, gets IO error writing to file
4405 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4406 caused clients to be evicted for AST timeouts. The fixes in
4407 bug 5192 mean we no longer need to have such short AST timeouts
4408 so ldlm_timeout has been increased.
4411 Frequency : occasional during --force or --failover shutdown under load
4412 Bugzilla : 5949, 4834
4413 Description: Server oops/LBUG if stopped with --force or --failover under load
4414 Details : a collection of import/export refcount and cleanup ordering
4415 issues fixed for safer force cleanup
4418 Frequency : only filesystems larger than 120 OSTs
4419 Bugzilla : 5990, 6223
4420 Description: lfs getstripe would oops on a very large filesystem
4421 Details : lov_getconfig used kfree on vmalloc'd memory
4424 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4425 Bugzilla : 6242, 6243
4426 Description: reading from files that had been truncated to a non-zero size
4427 but never opened returned no data
4428 Details : ll_file_read() reads zeros from no-object files to EOF
4433 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4434 Details : There's little harm in aborting MDS/OSS recovery and letting it
4435 try again, so I removed the LASSERT and return an error instead.
4437 Severity : enhancement
4439 Description: New debugging infrastructure for tracking down data corruption
4440 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4441 (b) cover more of the client-side code path, and (c) try to narrow
4442 down where problems occurred
4446 Bugzilla : 3819, 4364, 4397, 6313
4447 Description: Racing close and eviction MDS could cause assertion in mds_close
4448 Details : It was possible to get multiple mfd references during close and
4449 client eviction, leading to one thread referencing a freed mfd.
4451 Severity: : enhancement
4452 Bugzilla : 3262, 6359
4453 Description: Attempts to reconnect to servers are now more aggressive.
4454 Details : This builds on the enhanced upcall-less recovery that was added
4455 in 1.4.2. When trying to reconnect to servers, clients will
4456 now try each server in the failover group every 10 seconds. By
4457 default, clients would previously try one server every 25 seconds.
4462 Description: After recovery, certain operations trigger a failed
4463 assertion on a client.
4464 Details : Failing over an mds, using lconf -d --failover, while a
4465 client was doing a readdir() call would cause the client to
4466 LBUG after recovery completed and the readdir() was resent.
4468 Severity : enhancement
4470 Description: Default groups are now added by lconf
4471 Details : You can now run lconf --group <servicename> without having to
4472 manually add groups with lmc.
4475 Frequency : occasional
4477 Description: Nodes with an elan id of 0 trigger a failed assertion
4480 Frequency : always when accessing e.g. tty/console device nodes
4482 Description: tty and some other devices nodes cannot be used on lustre
4483 Details : file's private_data field is used by device data and lustre
4484 values in there got lost. New field was added to struct file to
4485 store fs-specific private data.
4488 Frequency : when exporting Lustre via NFS
4490 Description: NFSD failed occasionally when looking up a path component
4491 Details : NFSD is looking up ".." which was broken in ext3 directories
4492 that had grown large enough to become hashed.
4495 Frequency : Clusters with multiple interfaces not on the same subnet
4497 Description: Nodes will repeatedly try to reconnect to an interface which it
4498 cannot reach and report an error to the log.
4499 Details : Extra peer list entries will be created by lconf with some peers
4500 unreachable. lconf now validates the peer before adding it.
4503 Frequency : Only if a default stripe is set on the filesystem root.
4505 Description: Setting a default stripe on the filesystem root prevented the
4506 filesystem from being remounted.
4507 Details : The client was sending extra request flags in the root getattr
4508 request and did not allocate a reply buffer for the dir EA.
4511 Frequency : occasional, higher if lots of files are accessed by one client
4512 Bugzilla : 6159, 6097
4513 Description: Client trips assertion regarding lsm mismatch/magic
4514 Details : While revalidating inodes the VFS looks up inodes with ifind()
4515 and in rare cases can find an inode that is being freed.
4516 The ll_test_inode() code will free the lsm during ifind()
4517 when it finds an existing inode and then the VFS later attaches
4518 this free lsm to a new inode.
4522 Bugzilla : 6422, 7030
4523 Description: MDS deadlock between mkdir and client eviction
4524 Details : Creating a new file via mkdir or mknod (starting a transaction
4525 and getting the ns lock) can deadlock with client eviction
4526 (gets ns lock and trying to finish a synchronous transaction).
4529 Frequency : occasional
4530 Description: While starting a server, the fsfilt_ext3 module could not be
4532 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4533 kernels. Previously, lconf would still use the ext3 name
4534 when trying to load modules. Now, it will correctly use
4535 ext3 on 2.4 and ldiskfs on 2.6.
4537 Severity : enhancement
4538 Description: The default stripe count has been changed to 1
4539 Details : The interpretation of the default stripe count (0, to lfs
4540 or lmc) has been changed to mean striping across a single
4541 OST, rather than all available. For general usage we have
4542 found a stripe count of 1 or 2 works best.
4544 Severity : enhancement
4545 Description: Add support for compiling against Cray portals.
4546 Details : Conditional compiling for some areas that are different
4550 Frequency : occasional
4551 Bugzilla : 6409, 6834
4552 Description: Creating files with an explicit stripe count may lead to
4553 a failed assertion on the MDS
4554 Details : If some OSTs are full or unavailable, creating files may
4555 trigger a failed assertion on the MDS. Now, Lustre will
4556 try to use other servers or return an error to the
4560 Frequency : occasional
4562 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4563 caused lock timeout during readahead (since 1.4.2).
4564 Details : Processes doing readahead might match a lock that hasn't been
4565 granted yet if there are overlapping and conflicting lock
4566 requests. The readahead process waits on ungranted lock
4567 (original lock is CBPENDING), while OST waits for that process
4568 to cancel CBPENDING read lock and eventually evicts client.
4570 Severity : enhancement
4572 Description: Initial enabling of flock support for clients
4573 Details : Implements fcntl advisory locking and file status functions.
4574 This feature is provided as an optional mount flag (default
4575 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4576 locking are implemented yet, and those that are are not guaranteed
4577 to be completely correct in production environments.
4578 mount -t lustre -o [flock|noflock] ...
4581 Frequency : occasional
4583 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4584 assertion in the ext3 JBD (journaling) layer.
4585 Details : The b_committed_data struct is protected by the big kernel lock
4586 in 2.4 kernels, serializing journal_commit_transaction() and
4587 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4588 access to this struct. In 2.6 kernels there is finer grained
4589 locking to improve SMP performance of the JBD layer.
4593 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4598 Description: merge and cleanup kernel patches.
4601 -----------------------------------------------------------------------------
4603 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4605 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4606 versions of Lustre. Please update all servers and clients to
4607 version 1.4.2 or later at the same time. You must also run
4608 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4609 to update the configuration logs.
4611 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4612 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4613 - swab llog records of type '0' so we get proper header size/idx (5861)
4614 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4615 - fix rename of one directory over another leaking an inode (5953)
4616 - avoid SetPageDirty on 2.6 (5981)
4617 - don't re-add just-being-destroyed locks to the waiting list (5653)
4618 - when creating new directories, inherit the parent's custom
4619 striping settings if present parent (3048)
4620 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4621 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4622 - don't hold client locks on temporary worklist from l_lru (5666)
4623 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4624 - automatically evict dead clients (5921)
4625 - Update file size properly in create+truncate+fstat case (6196)
4626 - Do not unhash mountpoint dentries, do not allow removal of
4628 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4629 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4630 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4631 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4632 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4633 - Correctly handle reads of files with no objects (6243)
4634 - lctl recover will also mark a device active if deactivate used (5933)
4636 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4637 - allow --write-conf on an MDS with different nettype than client (5619)
4638 - don't write config llogs to MDS for mounts not from that MDS (5617)
4639 - lconf should create multiple TCP connections from a client (5201)
4640 - init scripts are now turned off by default; run chkconfig --on
4641 lustre and chkconfig --on lustrefs to use them
4642 - upcalls are no longer needed for clients to recover to failover
4644 - add --abort-recovery option to lconf to abort recovery on device
4646 - add support for an arbitrary number of OSTs (3026)
4647 - Quota support protocol changes.
4648 - forward compatibility changes to wire structs (6007)
4649 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4650 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4651 - improved "lustre status" script
4652 - initialize blocksize for non-regular files (6062)
4653 - added --disable-server and --disable-client configure options (5782)
4654 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4655 - Vanilla 2.4.29 support
4656 - increase maximum number of obd devices to 520 (6242)
4657 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4658 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4659 SLES 9 2.6.5 kernels
4660 - limit stripes per file to 160 (the maximum EA size) (6093)
4662 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4665 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4666 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4667 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4668 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4669 - small ext3 extents cleanups and fixes (5733)
4670 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4671 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4672 - avoid needless client->OST connect, fix handle mismatch (5317)
4673 - fix DLM error path that led to out-of-sync client, long delays (5779)
4674 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4675 - fix several locking issues related to i_size (5492,5624,5654,5672)
4676 - don't move pending lock onto export if it is already evicted (5683)
4677 - fix kernel oops when creating .foo in unlinked directory (5548)
4678 - fix deadlock in obdfilter statistics vs. object create (5811)
4679 - use time_{before,after} to avoid timer jiffies wrap (5882)
4680 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4681 - Do not leak request if server was not able to process it (5154)
4682 - If mds_open unable to find parent dir, make that negative lookup(5154)
4683 - don't create new directories with extent-mapping (5909, 5936)
4685 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4686 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4687 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4688 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4689 - support for RHEL4 kernel on i686 (5773)
4690 - provide error messages when incompatible logs are encountered (5898)
4692 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4693 * version 1.4.0.10 (1.4.1 release candidate 1)
4695 - don't keep a lock reference when lock is not granted (4238)
4696 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4697 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4698 - drop import inflight refcount on signal_completed_replay error (5255)
4699 - unlock page after async write error during send (3677)
4700 - handle missing objects in filter_preprw_read properly (5265)
4701 - no transno return for symlink open, don't save no-trasno open (3440)
4702 - don't try to complete elan receive that already failed (4012)
4703 - free RPC server reply state on error (5406)
4704 - clean up thread from ptlrpc_start_thread() on error (5160)
4705 - readahead could read extra page into cache that wasn't ejected (5388)
4706 - prevent races in class_attach/setup/cleanup/detach (5260)
4707 - don't dereference de->d_inode after l_dput of de (5458)
4708 - use "int" for stripe value returned from lock_to_stripe (5544)
4709 - mballoc allocation and error-checking fixes in 2.6 (5504)
4710 - block device patches to fix I/O request sizes in 2.6 (5482)
4711 - look up hostnames for IB nals (5602)
4712 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4713 - don't start multiple acceptors for the same port (5277)
4714 - fix incorrect LASSERT in mds_getattr_name (5635)
4715 - export a proc file for general "ping" checking (5628)
4716 - fix "lfs check" to not block when the MDS is down (5628)
4718 - service request history (4965)
4719 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4720 - create an "evict_client" /proc entry on OSTs, like the MDS has
4721 - fix mount usage message, return errors per mount(8) (5168)
4722 - change grep [] to grep "[]" in tests so they work in more UMLs
4723 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4724 - remove ancient LOV_MAGIC_V0 EA support (5047)
4725 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4726 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4727 - allow readahead allocations to fail when low on memory (5383)
4728 - mmap locking landed again, after considerable improvement (2828)
4729 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4731 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4734 - send OST transaction number in read/write reply to free req (4966)
4735 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4736 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4737 - the watchdog thread now runs as interruptible (5246)
4738 - flock/lockf fixes (but it's still disabled, pending 5135)
4739 - don't use EXT3 constants in llite code (5094)
4740 - memory shortage at startup could cause assertion (5176)
4742 - reorganization of lov code
4743 - single portals codebase
4745 - add extents/mballoc support (5025)
4746 - direct I/O reads in the obdfilter (4048)
4747 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4749 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4752 - send OST transaction number in read/write reply to free req (4966)
4753 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4754 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4755 - the watchdog thread now runs as interruptible (5246)
4756 - handle missing objects in filter_preprw_read properly (5265)
4757 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4758 - drop import inflight refcount on signal_completed_replay error (5255)
4759 - unlock page after async write error during send (3677)
4760 - return original error code on reconstructed replies (3761)
4761 - no transno return for symlink open, don't save no-trasno open (3440)
4763 - add pid to ldlm debugging output (4922)
4764 - bump the watchdog timeouts -- we can't handle 30sec yet
4765 - extra debugging for orphan dentry/inode bug (5259)
4767 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4770 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4771 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4772 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4773 - fix NULL dereference in /proc/sys/portals/routes (4827)
4774 - allow failed mdc_close() operations to be interrupted (4561)
4775 - stop precreate on OST before MDS would time out on it (4778)
4776 - don't send partial-page writes before EOF from client (4410)
4777 - discard client grant for sub-page writes on large-page clients (4520)
4778 - don't free dentries not owned by NFS code, check generation (4806)
4779 - fix lsm leak if mds_create_objects() fails (4801)
4780 - limit debug_daemon file size, always print CERROR messages (4789)
4781 - use transno after validating reply (3892)
4782 - process timed out requests if import state changes (3754)
4783 - update mtime on OST during writes, return in glimpse (4829)
4784 - add mkfsoptions to LDAP (4679)
4785 - use ->max_readahead method instead of zapping global ra (5039)
4786 - don't interrupt __l_wait_event() during strace
4788 - add software watchdogs to catch hung threads quickly (4941)
4789 - make lustrefs init script start after nfs is mounted
4790 - fix CWARN/ERROR duplication (4930)
4791 - return async write errors to application if possible (2248)
4792 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4793 - print NAL number in %x format (4645)
4794 - update barely-supported suse-2.4.21-171 series (4842)
4795 - support for sles 9 %post scripts
4796 - support for building 2.6 kernel-source packages
4797 - support for sles km_* packages
4799 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4802 - ignore -ENOENT errors in osc_destroy (3639)
4803 - notify osc create thread that OSC is being cleaned up (4600)
4804 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4805 - reconstruct ost_handle() like mds_handle() (4657)
4806 - create a new thread to do import eviction to avoid deadlock (3969)
4807 - let lconf resolve symlinked-to devices (4629)
4808 - don't unlink "objects" from directory with default EA (4554)
4809 - hold socknal file ref over connect in case target is down (4394)
4810 - allow more than 32000 subdirectories in a single directory (3244)
4811 - fix blocks count for O_DIRECT writes (3751)
4812 - OST returns ENOSPC from object create when no space left (4539)
4813 - don't send truncate RPC if file size isn't changing (4410)
4814 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4815 - bind to privileged port in socknal and tcpnal (3689)
4817 - rate limit CERROR/CWARN console message to avoid overload (4519)
4818 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4819 - basic mmap support (3918)
4820 - kernel patch series update from b1_4 (4711)
4822 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4825 - avoid crash during MDS cleanup with OST shut down (2775)
4826 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4827 - don't use bad inodes on the MDS (3744)
4828 - dynamic object preallocation to improve recovery speed (4236)
4829 - don't hold spinlock over lock dumping or change debug flags (4401)
4830 - don't zero obd_dev when it is force cleaned (3651)
4831 - print grants to console if they go negative (4431)
4832 - "lctl deactivate" will stop automatic recovery attempts (3406)
4833 - look for existing locks in ldlm_handle_enqueue() (3764)
4834 - don't resolve lock handle twice in recovery avoiding race (4401)
4835 - revalidate should check working dir is a directory (4134)
4837 - don't always mark "slow" obdfilter messages as errors (4418)
4839 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4842 - don't close LustreDB during write_conf until it is done (3860)
4843 - fix typo in lconf for_each_profile (3821)
4844 - allow dumping logs from multiple threads at one time (3820)
4845 - don't allow multiple threads in OSC recovery (3812)
4846 - fix debug_size parameters (3864)
4847 - fix mds_postrecov to initialize import for llog ctxt (3121)
4848 - replace config semaphore with spinlock (3306)
4849 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4850 - don't allow enqueue to complete on a destroyed export (3822)
4851 - down write_lock before checking llog header bitmap (3825)
4852 - recover from lock replay timeout (3764)
4853 - up llog sem before sending rpc (3652)
4854 - reduce ns lock hold times when setting kms (3267)
4855 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4856 - fix NULL deref and obd_dev leak on setup error (3312)
4857 - replace some LBUG about llog ops with error handling (3841)
4858 - don't match INVALID dentries from d_lookup and spin (3784)
4859 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4860 - fix invalid assertion in ptlrpc_set_wait (3880)
4862 - add libwrap support for the TCP acceptor (3996)
4863 - add /proc/sys/portals/routes for non-root route listing (3994)
4864 - allow setting MDS UUID in .xml (2580)
4865 - print the stack of a process that LBUGs (4228)
4867 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4870 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4871 - make sure to unset replay flag from failed open requests (3440)
4872 - if default stripe count is 0, use OST count for inode size (3636)
4873 - update parent mtime/ctime on client for create/unlink (2611)
4874 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4875 - free recovery state on server during a forced cleanup (3571)
4876 - unregister_reply for resent reqs (3063)
4877 - loop back devices mounting and status check on 2.6 (3563)
4878 - fix resource-creation race that can provoke i_size == 0 (3513)
4879 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4880 - more debugging for page-accounting assertion (3746)
4881 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4882 - avoid deadlock after precreation failure (3758)
4883 - fix race and lock order deadlock in orphan handling (3450, 3750)
4884 - add validity checks when grabbing inodes from l_ast_data (3599)
4886 - add /proc/.../recovery_status to obdfilter (3428)
4887 - lightweight CDEBUG infrastructure, debug daemon (3668)
4888 - change default OSC RPC parameters to be better on small clusters
4889 - turn off OST read cache for files smaller than 32MB
4890 - install man pages and include them in rpms (3100)
4891 - add new init script for (un)mounting lustre filesystems (2593)
4892 - run chkconfig in %post for init scripts (3701)
4893 - drop scimac NAL (unmaintained)
4895 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4898 - clean kiobufs before and after use (3485)
4899 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4900 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4901 - change init script to fail more clearly if not run as root (1528)
4902 - allow clients to reconnect during replay (1742)
4903 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4904 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4905 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4906 - fix cond_resched (3554)
4907 - only evict unfinished clients after recovery (3515)
4908 - allow bulk resend, prevent data loss (3570)
4909 - dynamic ptlrpc request buffer allocation (2102)
4910 - don't allow unlinking open directory if it isn't empty (2904)
4911 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4912 - remove extraneous obd dereference causing LASSERT failure (3334)
4913 - don't use get_cycles() when creating temp. files on the mds (3156)
4914 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4915 - handle EEXIST for set-stripe, set proper directory name (3336)
4917 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4918 - fix ksocknal_fmb_callback() error messages (2918)
4920 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4923 - don't copy lvb into (possibly NULL) reply on error (2983)
4924 - don't deref dentry after dput, don't free lvb on error (2922)
4925 - use the kms to determine writeback rpc length (2947)
4926 - increment oti_logcookies when osc is inactive (2948)
4927 - update client's i_blocks count via lvb messages (2543)
4928 - handle intent open/close of special files properly (1557)
4929 - mount MDS with errors=remount-ro, like obdfilter (2009)
4930 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4931 - don't use cancelling-locks' kms values (2947)
4932 - use highest lock extent for kms, not last one (2925)
4933 - don't dereference ERR_PTR() dentry in error handling path (3107)
4934 - fix thread race in portals_debug_dumplog() (3122)
4935 - create lprocfs device entries at setup instead of at attach (1519)
4936 - common AST error handler, don't evict client on completion race (3145)
4937 - zero nameidata in detach_mnt in 2.6 (3118)
4938 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4939 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4940 - initialize RPC timeout timer earlier for 2.6 (3219)
4941 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4942 - print nal/nid for unknown nid (3258)
4943 - additional checks for oscc recovery before doing precreate (3284)
4944 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4945 - don't crash in mdc_close for bad permissions on open (3285)
4946 - zero i_rdev for non-device files (3147)
4947 - clear page->private before handing to FS, better assertion (3119)
4948 - tune the read pipeline (3236)
4949 - fix incorrect decref of invalidated dentry (2350)
4950 - provide read-ahead stats and refine rpc in flight stats (3328)
4951 - don't hold journal transaction open across create RPC (3313)
4952 - update atime on MDS at close time (3265)
4953 - close LDAP connection when recovering to avoid server load (3315)
4954 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4955 - don't leak open file on MDS after open resend (3325)
4956 - serialize filter_precreate and filter_destroy_precreated (3329)
4957 - loop device shouldn't call sync_dev() for nul device (3092)
4958 - clear page cache after eviction (2766)
4959 - resynchronize MDS->OST in background (2824)
4960 - refuse to mount the same filesystem twice on same mountpoint (3394)
4961 - allow llmount to create routes for mounting behind routers (3320)
4962 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4963 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4964 - fix rare problem with rename on htree directories (3417)
4966 - allow default OST striping configuration per directory (1414)
4967 - fix compilation for qswnal for 2.6 kernels (3125)
4968 - increase maximum number of MDS request buffers for large systems
4969 - change liblustreapi to be useful for external progs like lfsck (3098)
4970 - increase local configuration timeout for slow disks (3353)
4971 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4973 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4976 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4977 - don't overwrite extent policy data in reply if lock was blocked (2901)
4978 - drop filter export grants atomically with removal from device (2663)
4979 - del obd_self_export from work_list in class_disconnect_exports (2908)
4980 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4981 - swab reply message in mdc_close, other PPC fixes (2464)
4982 - fix destroying of named logs (2325)
4983 - overwrite old logs when running lconf --write_conf (2264)
4984 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4985 - fix race in target_handle_connect (2898)
4986 - mds_reint_create() should take same inode create lock (2926)
4987 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4988 - don't close files for self_export to avoid uninitialized obd (2936)
4989 - allow MDS with the same name as client node (2939)
4990 - hold dentry reference for closed log files for unlink (2325)
4991 - reserve space for all logs during transactions (2059)
4992 - don't evict page beyond end of stripe extent (2925)
4993 - don't oops on a deleted current working directory (2399)
4994 - handle hard links to targets without a parent properly (2517)
4995 - don't dereference NULL lock when racing during eviction (2867)
4996 - don't grow lock extents when lots of conflicting locks (2919)
4998 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5001 - account for cache space usage on clients to avoid data loss (974)
5002 - lfsck support in lustre kernel code (2349)
5003 - reduce journal credits needed for BRW writes (2370)
5004 - orphan handling to avoid losing space on client/server crashes
5005 - ptlrpcd can be blocked, stopping ALL progress (2477)
5006 - use lock value blocks to assist in proper KMS, faster stat (1021)
5007 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5008 - recovery for initial connections (2355)
5009 - fixes for mds_cleanup_orphans (1934)
5010 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5011 - block all file creations until orphan recovery completes (1901)
5012 - client remove rq_connection from request struct (2423)
5013 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5014 - recovery timer race (2670)
5015 - mdc_close recovey bug (2532)
5016 - ptlrpc cleanup bug (2710)
5017 - mds timeout on local locks (2588)
5018 - namespace lock held during RPCs (2431)
5019 - handle interrupted sync write properly (2503)
5020 - don't try to handle a message that hasn't been replied to (2699)
5021 - client assert failure during cleanup after abort recovery (2701)
5022 - leak mdc device after failed mount (2712)
5023 - ptlrpc_check_set allows timedout requests to complete (2714)
5024 - wait for inflight reqs when ptlrpcd finishes (2710)
5025 - make sure unregistered services are removed from the srv_list
5026 - reset bulk XID's when resending them (caught by 1138 test)
5027 - unregister_bulk after timeout
5028 - fix lconf error (2694)
5029 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5030 - readahead locks pages, leaves pending causing memory pressure (2673)
5031 - increase OST request buffers to 4096 on large machines (2729)
5032 - fix up permission of existing directories in simple_mkdir (2661)
5033 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5034 - don't assign transno to errored transactions (2742)
5035 - don't delete objects on OST if given a bogus objid from MDS (2751)
5036 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5037 - if rq_no_resend, then timeout request after recovery (2432)
5038 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5039 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5040 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5041 - MDS recovery shouldn't delete objects that it has given out (2730)
5042 - if enqueue arrives after completion, don't clobber LVB (2819)
5043 - don't unlock pages twice when trigger_group_io returns error (2814)
5044 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5045 - don't write pages to disk if there was an error (1450)
5046 - don't ping imports that have recovery disabled (2676)
5047 - take buffered bytes into account when balancing socknal conn (2817)
5048 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5049 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5050 - remove little-endian swabbing of llog records (1987)
5051 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5052 - retry reposting request buffers if they fail (1191)
5053 - grow extent at grant time to avoid granting a revoked lock (2809)
5054 - lock revoke doesn't evict page if covered by a second lock (2765)
5055 - disable VM readahead to avoid reading outside lock extents (2805)
5057 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5058 - updated kernel patches for hp-2.4.20 kernel (2681)
5060 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5063 - fix truncated write corruption (2366)
5064 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5066 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5067 - ensure len(uuid) < 37 in lmc (1171)
5068 - fix ia64 OOPS in llog_test (2255)
5069 - zero end of page at obdfilter for partial page writes (2648)
5070 - don't leave stale dentries around after renames (bug 2428)
5071 - fix timeouts when evicting a client with a single lock held (2642)
5072 - set deadline for the initial HELLO message to drain (2634)
5073 - print out dotted-quad IP addresses in the socknal (2302)
5075 - additional debugging for MDS client eviction problem (2443)
5076 - fix mkfsoptions support for osts (2603, 2604)
5078 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5081 - add series for the vanilla 2.6.0 kernel
5082 - add series for the vanilla 2.4.24 kernel
5083 - add series for a cray x86/64 UL kernel drop
5084 - fix xattr patches for the vanilla 2.4.19 series
5086 - generate true UUIDs in lmc (1171)
5087 - have portals stack dumping break in UML (2466)
5088 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5089 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5090 - wake callback waiting threads on client eviction (2460)
5091 - Add --ptldebug and --subsystem to lmc (1719)
5092 - update assertion to allow safe interrupt allocation
5093 - set rq_no_resend for cancel requests (2432)
5094 - recalculate ptlrpcd timeout after resend (2494)
5095 - call vfs_rmdir when removing pending directories (2368)
5096 - fix renaming a file to itself (2429)
5097 - lmc creates a default one-stripe lov (2454)
5098 - expand procfs space to handle large clusters (2326)
5099 - increase UML stack to avoid overflow
5100 - update lconf's list of debug and subsystem masks
5101 - fix lfs find --obd (2510)
5102 - /proc tunable for disabling filter read caching (2591)
5103 - stop rpm packages from altering slapd.conf (2301)
5104 - disable nagle in the socknal under 0conf (2578)
5105 - choose mds inode size based on stripe count (2572)
5106 - fix kernel-source rpm problems (2516)
5108 - add --disable-doc to avoid pdf generation (2421)
5109 - update documentation, tests, type-os, comments
5110 - avoid format warnings on ia64
5111 - remove the TOE NAL
5112 - tiny code cleanups by removing unused fields
5114 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5117 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5118 - remove the most blatant lies from BUILDING file (2371)
5119 - change default debug level to reasonable production setting
5120 - reduce client side cache size to reduce cache flush time
5121 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5122 - make TCP zerocopy and pinger support enabled by default (2476)
5123 - sync writes completed after process exits caused crashes (2319)
5124 - maintain correct mount count on the MDS (2356)
5125 - backout 1557, because 2316 wasn't really fixed
5126 - better file I/O statistics gathering in /proc
5127 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5128 - another kernel patch to fix zero-copy TCP function export
5129 - don't take duplicate lock when processing re-sent getattr (2420)
5130 - lctl uses obd_self_export instead of creating new conn (2353)
5131 - MDS/OST recovery case which requires object creation asserted (2425)
5132 - move lfs from /usr/sbin to /usr/bin in packages
5133 - fix race between mds_client_add and mds_client_free (2417)
5134 - use kmalloc instead of slabs in portals (2430)
5135 - don't create duplicate records when a failover MDS is present (2442)
5136 - remove unnecessary mount age check (2332)
5137 - don't remove directory inodes from locks prematurely (2451)
5138 - don't break if MDS service name is the same as hostname (2103)
5139 - fix races in client write RPC generation when cache full (2482)
5141 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5144 - remove now-unused request->rq_obd (278)
5145 - if an allocation fails, print out how much memory we've used (1933)
5146 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5147 - add the "configurable stack size" patch to most series files (1256)
5148 - ability to write large log records, for 100+ OST configs (2306)
5149 - fix NULL deref when filter_prep fails (2314)
5150 - fix operator precedence error in filter_sync
5151 - dynamic allocation of socknal TX descriptors (2315)
5152 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5153 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5154 - ensure that CWARN messages in Portals always get to the syslog
5155 - __init/__exit are not for prototype decls (ldlm_init/exit)
5156 - x86-64 compile warning fixes
5157 - fix gateway LMC keyword conflict (2318)
5158 - fix MDS lock inversions in getattr/reint paths (1844)
5159 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5160 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5161 - DLM race condition prevented some lock evictions (2328)
5162 - ENOMEM detection and retry on socknal sends (2230)
5163 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5164 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5165 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5166 - better error messages when a client is rejected during recovery (1505)
5167 - avoid cancelling locks which were never granted, after failure (2330)
5168 - fix i_sem/journal inversion in mds_client_add (2333)
5169 - fix truncate/getattr lock cycle deadlock (2334)
5170 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5171 - fix two rare exit paths which could leak an l_lock() ref (2321)
5172 - fencepost error in MDS/OST orphan recovery (2226)
5173 - make log record alignment 8 bytes (1988)
5174 - lstripe now fails when requested offset > ost_count (2237)
5175 - ensure that all kernel series have a complete list.h (1607)
5176 - fix crashes in special-file operations (2316)
5177 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5179 - allow configurable automake binary, for testing new versions
5180 - small update to the lfs documentation
5182 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5184 * fix negative export reference count in fsfilt_sync (2312)
5186 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5187 * release candidate 0.9.1
5189 - orphans are moved into the PENDING directory for possible recovery
5190 - replayed opens now open by fid for orphan/rename safety (1042)
5191 - last close of an orphan inode generates a transno (683)
5192 - chdir() and mount() now pin the directory entry (1020)
5193 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5194 - discard very old requests without processing them (1502)
5195 - remove some common, well-understood CERRORs (1505)
5196 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5197 - clear "grant" flags in OST replies until OST grant code lands (1644)
5198 - fix read performance by not clobbering i_blksize on client (1598)
5199 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5200 - make LRU size a /proc tunable, clears locks when reduced (707)
5201 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5202 - server locks take references on exports now (1558)
5203 - build fixes for 2.4.20-rh trees (1663)
5204 - return an error from lov_create if all OSCs are inactive (1751)
5205 - fix import levels when a reconnect happens without a timeout (1597)
5206 - exit early from mds_open if we get a lookup error (1749)
5207 - partial page read at EOF wouldn't wait for disk before sending (1642)
5208 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5209 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5210 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5211 - write/truncate lock inversion (1639)
5212 - Don't auto-load obdclass, portals modules during cleanup (1495)
5213 - fix timestamps from jumping to "now" (1763)
5214 - extra journal assertions (1648)
5215 - add an extra multiunlink test (1771)
5216 - fix read_record/write_record API (1776)
5217 - fix leak of offset_extent, possible incorrect i_size later (1772)
5218 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5219 - Debugging for the kqswnal_get_idle_tx problems (1820)
5220 - Allow recovery to be attempted multiple times (1536)
5221 - Write out MDS last_rcvd file after it is first created (1600)
5222 - Fix tx_descriptor leak in failed transmit situations (1827)
5223 - ext3 journaling fixes for assertion failure after IO error (1871)
5224 - class_export_put() on freed export after completion AST error (1896)
5225 - Fix revalidate looping in VFS (1322)
5226 - Don't access a freed export during MDS_REINT timeout (1521)
5227 - Add open-unlink recovery support on the MDS (1673,1764)
5228 - Return an error if no MDS data was read from last_rcvd (1946)
5229 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5230 - Don't setup a disk device that doesn't match exported UUID (317)
5231 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5232 - avoid committing NULL handle in force close
5233 - local.sh is now a one-stripe LOV configuration
5234 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5235 - fix ext3 htree duplicate directory entry corruption (1516)
5236 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5237 - update to new LOV EA format (2097)
5238 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5240 2003-06-15 Phil Schwan <phil@clusterfs.com>
5243 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5244 - per-import/export recovery handling (958, 931, 959)
5245 - multiple last-rcvd slots, for serving multiple FSes (949)
5246 - connections are again shared between multiple imp/exports (963, 964)
5247 - "umount -f" would hang if any requests needed to be sent (393, 978)
5248 - avoid pinning large req buffer by copying for queued messages (989)
5249 - add "uuid" to "lctl device" command to help upcalls (991)
5250 - "open" RPCs with transnos would confuse recovery counters (1037)
5251 - do proper endian conversion of all wire messages (288, 340, 891)
5252 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5253 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5254 - fix ptlrpc_connection leak in target_handle_connect (1174)
5255 - fix import refcounting bug in OST and MDS cleanup (1134)
5256 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5257 - fix very unlikely obd_types race condition (501)
5258 - remove osc_open hack for echo_client (1187)
5259 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5260 - a failure in read_inode2 leads to deadlock (1139)
5261 - cancel ack-locks as soon as transaction is committed (1072)
5262 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5263 - make sure to commitrw after any preprw to avoid deadlock (1162)
5264 - failing to execute a file in a lustre FS would lock inode (1203)
5265 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5266 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5267 - obd_uuid2tgt would walk off the end of the list (1255)
5268 - on IA64 the osc would give portals incorrect bulk size (1258)
5269 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5270 - fix lock inversion caused by new llite matching code (1282)
5271 - limit the number of dirty pages on a client to 10MB (1286)
5272 - timed out locks were not being corrected cancelled (1289)
5273 - fix O_DIRECT above 4GB on IA-32 (1292)
5274 * major user-visible changes
5275 - fail out/fail over policy now controlled by the upcall (993)
5277 - add OBD_PING to check server availability and failure (954)
5278 - lustre messages are now sent in sending host order (288, 340, 891)
5279 - add eadatalen to MDS getattr reply (340)
5280 - OST read replies may contain second buffer, with per-page status (593)
5282 2003-03-11 Phil Schwan <phil@clusterfs.com>
5285 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5286 - failed open()s could cause deadlock; fixed (867, 869)
5287 - stop cancelling OST locks when files are closed (481)
5288 - overlapping XID spaces caused network corruption (851, 853)
5289 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5290 - setattr_raw added, to do single-RPC, server-side setattrs
5291 - lmc/lconf syntax change for OST UUIDs
5292 - fix crashy race condition between ptlrpc_free_req and osc_close
5293 - don't use request in mdc_enqueue if we hit a timeout (889)
5294 - don't set the inode i_size for regular files from the MDS (896)
5295 - handle out of order completion AST (842)
5296 - don't LBUG if a lock request times out after receiving AST (913)
5297 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5298 - if a bad lock AST arrives, send an error instead of dropping entirely
5299 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5300 - fix leak in bulk IO when only partially completed (899, 900, 926)
5301 - fix O_DIRECT for ia64 (55)
5302 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5303 - C-z after timeout could hang a process forever; fixed (977)
5305 - client-side I/O cache (678, 924, 929, 941, 970)
5307 - READPAGE and SETATTRs which don't take server-side locks get
5310 2003-02-11 Phil Schwan <phil@clusterfs.com>
5313 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5314 - Fix fsfilt_extN_readpage() to read a full page of directory
5315 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5316 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5317 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5318 - Most intent operations were removed, in favour of a new RPC mode
5319 that does a single RPC to the server and bypasses most of the VFS
5320 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5321 - Aggressively cancel local locks on DLM servers
5322 - mds_reint_unlink sends EA to the client if it's the last nlink.
5323 client uses that EA to unlink OST objects.
5324 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5325 - recursive symlinks were fixed (439)
5326 - fixed NULL deref in DEBUG_REQ
5327 - filter_update_lastobjid no longer calls sync, which annoyed extN
5328 - fixed multi-client small-writes to a single file problem (445)
5329 - fixed mtime updates during file writes (607)
5330 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5331 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5332 - fixed Linux OST size reporting problem (444, 656)
5333 - OST now updates object mtime with writes or setattr (607, 619)
5334 - client verifies file size before zeroing page past EOF (445)
5335 - OST now writes last allocated objid to disk with allocation (108)
5336 - LOV on echo now works (409)
5338 - mds_reint_unlink sends a new buffer, with the EA included. this
5339 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5340 set if a regular file was being unlinked, and it was the last link
5341 - use PtlGet from the target for bulk writes (315)
5342 - OST now updates object mtime with writes or setattr (607, 619)
5343 - LDLM now has a grant-time callback to revalidate locked items, if
5345 - Many MDS operations were reorganized to combat race conditions
5347 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5348 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5350 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5353 - Fully reactivate OST imports after reconnection (512, others)
5354 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5355 - More graceful error handling for truncating on dead OST (515)
5356 - Don't error out unless we're actually accessing dead stripes (474)
5357 - Fix garbage sizes when stripes are missing (410)
5358 - LRU counters were broken, causing constant lock purge (433, 432)
5359 - garbage on read from stripes with failed OSTs (441)
5360 - mark OSCs as active before reconnecting during recovery (438)
5361 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5362 - lfind did not preserve OST order in output (443)
5363 - symlinks cause hung clients, incorrect data (439)
5364 - stop dereferencing request after dropping refcount (457)
5365 - don't LASSERT(spin_is_locked) on non-SMP (455)
5366 - fixes for many rename() bugs
5367 - fstat didn't correctly synchronize attributes (399)
5368 - server must handle lock cancellation during blocking AST prep (487)
5369 - bulk descriptors were free()d too soon (511)
5370 - fix paths in lconf, which would load incorrect modules (451, 507)
5371 - fix confusing lconf 'host not found' error message (386)
5372 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5373 - fix race condition in mdc_blocking_ast() for inode access (526)
5374 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5375 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5376 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5377 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5378 - fix lstripe in several ways (kernel side) (527)
5379 - fix request leak in ldlm_cli_enqueue (262)
5380 - incorrect OSC was marked inactive after OST failure
5381 - call mds_fs_cleanup before unmounting filesystem (524)
5382 - fix races between taking ns_lock and ldlm_lock_change_resource
5383 - fix races updating LOV export open file list
5384 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5385 - fix recovery NULL deref in ldlm_cli_cancel_unused
5386 - fix some DLM races by using new hash table for lock handles (419)
5387 - permit the client to specify desired inodes, at replay
5388 - duplicate requests when we queue them for replay reintegration
5389 - fix last_rcvd offset calculation
5390 - sync after each recovered transaction, so we always make progress
5391 - never, not always, ERESTART requests without transnos
5392 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5394 - skip replay if the MDS says that the client is already connected
5395 - don't check for a recovery-enabled export to match lctl's UUID
5396 - don't INC_USE_COUNT for phantom exports
5397 - don't crash when cleaning up phantom exports (567)
5398 - don't double-finish or set replay data for errored mdc_open requests
5399 - abort requests when they time out, so we don't get old replies
5400 - send/receive replies for AST messages again
5401 - if the client says that it doesn't have the lock, cancel it on the
5403 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5404 mark it as destroyed, it will all work out when decref is called
5405 - fix module use counts (22, 581)
5407 - ASTs now expect a reply (server cancels lock on error reply)
5409 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5412 - fix many simultaneous client startup (392)
5413 - fix dentry->d_it clobbering
5414 - credentials weren't being shipped for readdir/getattr operations
5415 - remove invalid assertions triggered during some concurrent MD
5417 - proper Lustre versions added (336, 389)
5418 - fix memory leak for create error case (398)
5419 - fix LOV locking bug that would get cli/srv out of sync
5420 - fix echo client over LOV (409)
5421 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5422 - fix double-O_EXCL intent crash (424)
5423 - avoid sending multiple lock CANCELs (352)
5425 - MDS can do multi-client recovery (modulo bugs in new code)
5427 - many updates, edits, cleanups
5429 2002-11-18 Phil Schwan <phil@clusterfs.com>
5432 - fix null d_it dereference (346)
5433 - fix full OST/dbench hang (333)
5434 - fix permission problem with file removal (286)
5435 - fix removal of OSCs from LOV when they fail
5436 - fix NULL deref during bulk timeout (214)
5437 - fix problems related to multiple filesystems on one MDS (241)
5438 - fixed serious subtle metadata locking bugs
5439 - free locks on clients when inodes are removed due to memory
5441 - fix inode pointer in lock data (285)
5442 - partial support for multiple MDS on a single host (241)
5443 - data locks weren't cancelled at clear_inode time (290, 311)
5444 - intent locks could lead to unbounded lock growth (205)
5445 - added a maximum lock count, an LRU list, and a flusher
5446 - fix multiple rename (365)
5447 - properly abstracted the echo client
5448 - OSC locked 1 byte too many; fixed
5449 - rewrote brw callback code:
5450 - fixed recovery bugs related to LOVs (306)
5451 - fixed too-many-pages-in-one-write crash (191)
5452 - fixed (again) crash in sync_io_timeout (214)
5453 - probably fixed callback-related race (385)
5455 - Add capability to MDS protocol
5456 - LDLM cancellations and callbacks on different portals
5458 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5461 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5462 - defer open object destruction to close time (601981, 138)
5463 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5464 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5465 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5466 - don't call obd_setattr in ll_file_release for destroyed objects
5467 * protocol change to lustre_msg: move |version| and add |flags|
5468 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5469 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5470 * added replay of create, unlink, link and rename operations during
5471 MDS failover; recovery should be much more robust now
5472 * remove failed OSCs from LOVs (only lov_create uses this so far)
5473 * the lustre-HOWTO was brought (more) up to date (582544)
5475 2002-10-23 Phil Schwan <phil@clusterfs.com>
5478 - in-use dentries weren't being reused properly (617851)
5479 - prevent multiple LDLM setup (599178)
5480 - fix LOV size calculations for truncate (617853)
5481 - fix client handling of MDS intent errors (POSIX)
5482 - fix permission bug in lovstripe.c test (624321)
5483 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5484 - truncate past end of file could corrupt data
5485 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5486 - a race in recovery could return ETIMEDOUT to apps (623947)
5487 - building outside the source directory was fixed
5488 * the lustre-HOWTO was brought (more) up to date (582544)
5489 * major progress was made on recovery functionality
5491 2002-10-10 Phil Schwan <phil@clusterfs.com>
5494 - recovery deadlock fix
5495 - rm -rf causes LBUG fix (617817)
5496 - file open by multiple tasks fix (618962)
5497 - directory permissions bugs (602707 and 620007)
5498 - journal_stop fixed with locking (611313)
5499 - O_APPEND failures resolved (618273, perhaps 614459)
5500 - lconf PATH fix (619770)
5501 - IA64 build fix (621450)
5502 - RPC buffer sizes scale with amount of memory
5504 2002-10-01 Phil Schwan <phil@clusterfs.com>
5507 - locks would be cancelled without throwing away data pages,
5508 resulting in inconsistent data (605627)
5509 - inode attributes were not always being refreshed (605627, 612449)
5510 - lconf now continues to cleanup after lctl reports an error
5511 - MDS now enforces user permissions (602707)
5512 - lprocfs cleanup fixed, but not yet enabled (614157)
5513 - fixed infinite server hang, should a client not respond to an AST
5514 - avoid going into recovery if user calls readlink() with a buffer
5515 that's too small (613941)
5516 - AST RPCs no longer require replies (614867) -- this may be changed
5517 - don't crash server if client sends an IOV that's too big (611336)
5518 - fixed lock conversion deadlock (611892)
5519 - fixed the following of symlinks (614622)
5520 * recovery: the server can remove locks from a client that dies, other
5521 clients can make progress
5522 * more extN patch fixes
5523 * compile-time configurable ptlrpc buffer allocations
5525 - collaborative read cache document
5526 - Lustre Lite Performance CDR document-in-progress
5528 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5531 - fix typo in patch-2.4.18
5533 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5536 - clear ptlrpc request each time in handle_incoming_request()
5537 - unlink of files now destroys the object on the OST
5539 2002-09-19 Peter Braam <braam@clusterfs.com>
5541 * add hard link support
5542 * change obdfile creation method
5543 * kernel patch changed
5545 2002-09-19 Peter Braam <braam@clusterfs.com>
5548 - stack overflow bug in extN fixed
5550 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5552 * documentation updates
5553 - add man pages for config tools
5554 - update tests/README to describe testing with new config tools
5555 - finish metadata API descriptions
5556 * bug fixes and cleanups
5557 - statfs workaround for 16TB limit
5558 - LOV stripe allocation improved, can stripe on subset of OSTs
5559 - LOV file size/IO offset was wrong for files > 4GB in size
5560 - object EA data was being dropped, caused files to be unreadable
5561 - memory overflow with non-LOV OST caused memory corruption
5562 - fixed regression tests to work with new config tools, obdfilter
5563 - fixed bug when directory size became larger than 1 block
5564 - fixed bug (for single client case) when PWD was deleted
5565 - invalidate local directory pages when doing intent-based ops
5566 - avoid LDLM oops when lock callback contained bad data
5568 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5570 * documentation updates
5571 * bug fixes and cleanups
5572 - configuration tools
5575 - 64-bit compile warnings
5576 - 64-bit internal statfs data
5578 * test_brw on persistent OST devices
5580 * lprocfs (disabled)
5582 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5584 * documentation updates
5585 * bug fixes and cleanups
5586 * configuration tools
5588 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5591 * many small fixes to 0_5_4
5592 * io/network handling
5593 * thinkos in MDS operations
5595 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5598 * crucial basic fixes to 0.5.3
5599 * IOR, Iozone work over Elan
5602 2002-08-07 Phil Schwan <phil@clusterfs.com>
5603 * version 0_5_3, our first alpha
5604 * we use the new Portals iovs
5605 * documentation updates
5606 * bug fixes and cleanups
5607 * small changes in the DLM wire protocol
5609 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5610 * version 0_5_1 with some initial stability,
5611 * locking on MD and file I/O.
5612 * documentation updates
5613 * several bug fixes since 0.5.0
5614 * small changes in wire protocol
5616 2002-07-18 Phil Schwan <phil@clusterfs.com>
5618 * delivered as Lustre Light Alpha
5619 * fixed a crash after handling invalid MDS requests
5620 * fixed directory pages for architectures with non-4k pages sizes
5622 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5623 * release version v0_4_4
5624 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5626 * Fixes a number of interruption problems with OST operations.
5627 * Update documentation for portals header changes
5628 * Move all wire protocol structs/defines to lustre_idl.h
5629 * Fixes symlink length bug.
5630 * Add tcpdump to repository.
5632 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5633 * release version v0_4_3
5634 * Fixes statfs for inodes on extN.
5635 * Fixes bug in runtests which would delete /etc/hosts.
5636 * Use 64-bit object IDs wherever possible (not into VFS though)
5637 Remove ost_get_info, which is unused by lustre, and out of date.
5639 2002-07-03 Peter Braam <braam@clusterfs.com>
5640 * release version v0_4_2 Fixes a lookup error (type not passed)
5641 * move forward to head of Portals
5642 * move forward to latest Lustre kernel
5644 2002-06-25 Peter Braam <braam@clusterfs.com>
5645 * release version v0_4_1. Hopefully stable on single node use.