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.
19 Description: don't skip ost target if they assigned to file
20 Details : Drop slow OSCs if we can, but not for requested start idx.
21 This means "if OSC is slow and it is not the requested
22 start OST, then it can be skipped, otherwise skip it only
23 if it is inactive/recovering/out-of-space.
27 Description: more cleanup in mds_lov
28 Details : not send LOV EA under replay, we can't know about they size at this
29 time. Don't allow client connect to mds before any ost connected,
30 for avoid problems with LOV EA size and returning EIO to client.
32 Severity : enhancement
34 Description: Interoperability at server side (Disk interoperability)
36 Severity : enhancement
38 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
40 Severity : enhancement
42 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
44 Severity : enhancement
46 Description: New client IO stack (CLIO).
48 Severity : enhancement
50 Description: Commit on sharing. Eliminate inter-client dependencies between
51 uncommitted transactions by doing transaction commits.
52 Thereby clients may recovery independently.
55 Frequency : Create a symlink file with a very long name
57 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
58 Details : If there is no extra space in the request for early cancels,
59 ldlm_req_handles_avail() returns 0 instead of a negative value.
61 Severity : enhancement
63 Description: Add /proc entry for import status
64 Details : The mdc, osc, and mgc import directories now have
65 an import directory that contains useful import data for debugging
68 Severity : enhancement
70 Description: Re-disable certain /proc logging
71 Details : Enable and disable client's offset_stats, extents_stats and
72 extents_stats_per_process stats logging on the fly.
75 Frequency : Only on FC kernels 2.6.22+
77 Description: oops in statahead
78 Details : Do not drop reference count for the dentry from VFS when lookup,
79 VFS will do that by itself.
81 Severity : enhancement
83 Description: Generic /proc file permissions
84 Details : Set /Proc file permissions in a more generic way to enable non-
85 root users operate on some /proc files.
89 Description: Hitting mdc_commit_close() ASSERTION
90 Details : Properly handle request reference release in
91 ll_release_openhandle().
95 Description: quota recovery deadlock during mds failover
96 Details : This patch includes att18982, att18236, att18237 in bz14840.
98 1. fix osts hang when mds does failover with quotaon
99 2. prevent watchdog storm when osts threads wait for the
104 Frequency : only patchless client
105 Description: add workaround for race between add/remove dentry from hash
107 Severity : enhancement
109 Description: Allow OST glimpses to return PW locks
113 Description: LBUG when llog conf file is full
114 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
118 Description: Prevent import from entering FULL state when server in recovery
122 Description: service mount cannot take device name with ":"
123 Details : Only when device name contains ":/" will mount treat it as
129 Description: replace ptlrpcd with the statahead thread to interpret the async
130 statahead RPC callback
134 Frequency : on recovery
135 Description: I/O failures after umount during fail back
136 Details : if client reconnected to restarted server we need join to recovery
137 instead of find server handler is changed and process self
138 eviction with cancel all locks.
140 Severity : enhancement
142 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
144 Severity : enhancement
146 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
148 Severity : enhancement
150 Description: Upcall on Lustre log has been dumped
151 Details : Allow for a user mode script to be called once a Lustre log has
152 been dumped. It passes the filename of the dumped log to the
153 script, the location of the script can be specified via
154 /proc/sys/lnet/debug_log_upcall.
159 Description: avoid idr_remove called for id which is not allocated.
160 Details : Move assigment s_dev for clustered nfs to end of initialization,
161 to avoid problem with error handling.
166 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
167 Details : When connection is reused this not moved from CONN_UNUSED_HASH
168 into CONN_USED_HASH and this prodice warning when put connection
169 again in unused hash.
172 Severity : enhancement
174 Description: File striping can now be set to use an arbitrary pool of OSTs.
176 Severity : enhancement
178 Description: Export bytes_read/bytes_write count on OSC/OST.
182 Description: Early reply size mismatch, MGC loses connection
183 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
184 the connect flags are properly negotiated.
189 Description: quotas are not honored with O_DIRECT
190 Details : all writes with the flag O_DIRECT will use grants which leads to
191 this problem. Now using OBD_BRW_SYNC to guard this.
195 Description: add quota statistics
196 Details : 1. sort out quota proc entries and proc code.
197 2. add quota statistics
199 Severity : enhancement
201 Description: enable quota support for HEAD.
205 Description: Properly propagate oinfo flags from lov to osc for statfs
206 Details : restore missing copy oi_flags to lov requests.
208 Severity : enhancement
210 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
211 : l_getgroups(8), lst(8), routerstat(8)
212 Details : included man pages for llobdstat(8), llstat(8),
213 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
215 Severity : enhancement
217 Description: configure's --enable-quota should check the
218 : kernel .config for CONFIG_QUOTA
219 Details : configure is terminated if --enable-quota is passed but
220 : no quota support is in kernel
224 Description: Remove portals compatibility
225 Details : Remove portals compatibility, not interoperable with releases
230 Description: Resolve device initialization race
231 Details : Prevent proc handler from accessing devices added to the
232 obd_devs array but yet be intialized.
234 Severity : enhancement
236 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
238 Severity : enhancement
240 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
245 Description: Using wrong pointer in osc_brw_prep_request
246 Details : Access to array[-1] can produce panic if kernel compiled with
247 CONFIG_PAGE_ALLOC enabled
251 Description: Client runs out of low memory
252 Details : Consider only lowmem when counting initial number of llap pages
256 Description: *optional* service tags registration
257 Details : if the "service tags" package is installed on a Lustre node
258 When the filesystem is mounted, a local-node service tag will
259 be created. See http://inventory.sun.com/ for more information
260 about the Service Tags asset management system.
264 Description: Kernel BUG tries to release flock
265 Details : Lustre does not destroy flock lock before last reference goes
266 away. So always drop flock locks when client is evicted and
267 perform unlock regardless of successfulness of speaking to MDS.
271 Description: add refcount protection for osc callbacks, avoid panic on shutdown
275 Description: sanity test 65a fails if stripecount of -1 is set
276 Details : handle -1 striping on filesystem in ll_dirstripe_verify
281 Description: ASSERTION(CheckWriteback(page,cmd)) failed
282 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
285 Severity : enhancement
287 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
291 Description: do not process already freed flock
292 Details : flock can possibly be freed by another thread before it reaches
293 to ldlm_flock_completion_ast.
297 Description: LBUG during stress test
298 Details : Need properly lock accesses the flock deadlock detection list.
302 Description: oops in page fault handler
303 Details : kernel page fault handler can return two special 'pages' in error
304 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
308 Description: timeout with invalidate import.
309 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
310 should be handled by ptlrpcd. This produce long age waiting and
311 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
313 Severity : enhancement
315 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
319 Description: enable MGS and MDT services start separately
320 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
321 a co-located MGS without starting the MGS, which is a complement
322 to 'nosvc' mount option.
326 Frequency : after recovery
327 Description: precreate to many object's after del orphan.
328 Details : del orphan st in oscc last_id == next_id and this triger growing
329 count of precreated objects. Set flag LOW to skip increase count
330 of precreated objects.
334 Frequency : rare, on clear nid stats
335 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
336 Details : when clean nid stats sometimes try destroy live entry,
337 and this produce panic in free.
341 Description: Stack overflow during MDS log replay
342 ease stack pressure by using a thread dealing llog_process.
346 Description: wait until IO finished before start new when do lock cancel.
347 Details : VM protocol want old IO finished before start new, in this case
348 need wait until PG_writeback is cleared until check dirty flag and
349 call writepages in lock cancel callback.
351 Severity : enhancement
353 Description: using special macro for print time and cleanup in includes.
357 Description: mds_mfd_close() ASSERTION(rc == 0)
358 Details : In mds_mfd_close(), we need protect inode's writecount change
359 within its orphan write semaphore to prevent possible races.
363 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
367 Frequency : rare, on shutdown ost
368 Description: don't hit live lock with umount ost.
369 Details : shrink_dcache_parent can be in long loop with destroy dentries,
370 use shrink_dcache_sb instead.
374 Description: don't panic with use echo client
375 Details : echo client pass NULL as client nid pointer and this produce null
380 Description: fix build on ppc32
381 Details : compile code with -m64 flag produce wrong object file for ppc32.
385 Description: add message levels for liblustreapi
389 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
390 Details : Move the 'good_osts' check before the 'total_bavail' check. This
391 will result in an -EAGAIN and in the exit call path we call
392 alloc_rr() which will with increasing aggressiveness attempt to
393 aquire precreated objects on the minimum number of required OSCs.
397 Description: Use old size assignment to avoid deadlock
398 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
399 the scheduling while holding a spinlock. We do not need locking
400 for size in ll_update_inode() because size is only updated from
401 the MDS for directories or files without objects, so there is no
402 other place to do the update, and concurrent access to such inodes
403 are protected by the inode lock.
407 Description: resolve "_IOWR redefined" build error on SLES10
411 Description: dump the memory debugging after all modules are unloaded to
412 suppress false negative in conf_sanity test 39
414 Severity : enhancement
416 Description: build kernel-ib packages for OFED 1.3 in our release cycle
421 Description: fix SLES kernel versioning
422 Details : the kernel version for our SLES 10 kernel did not include a "-"
423 before the "smp" at the end. while this was not a problem in
424 general, it did mean that software trying to use the kernel
425 version to try to detect a vendor specific kernel would fail.
426 this was most evident by the OFED build scripts.
430 Description: Don't update lov_desc members until making sure they are valid
431 Details : When updating lov_desc members via proc fs, need fix their
432 validities before doing the real update.
436 Description: don't put request into delay list while invalidate in flight.
437 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
438 import in flight. this produce timeout for invalidate and sometimes
439 can cause stale data.
443 Frequency : on ppc only
444 Description: not convert ost objects for directory because it's not exist.
445 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
449 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
451 Severity : enhancement
453 Description: Update to support 2.6.22.14 vanilla kernel.
457 Frequency : rare, on recovery
458 Description: read procfs can produce deadlock in some situation
459 Details : Holding lprocfs lock which send rpc can produce block for destroy
460 obd objects and this also block reconnect with -EALREADY. This
461 isn't fix all lprocfs bugs - but make it rare.
463 Severity : enhancement
465 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
468 Frequency : frequent on X2 node
470 Description: mdc_set_open_replay_data LBUG
471 Details : Set replay data for requests that are eligible for replay.
475 Description: lustre_mgs: operation 101 on unconnected MGS
476 Details : When MGC is disconnected from MGS long enough, MGS will evict the
477 MGC, and late on MGC cannot successfully connect to MGS and a lot
478 of the error messages complaining that MGS is not connected.
482 Frequency : on network error
483 Description: panic with double free request if network error
484 Details : mdc_finish_enqueue is finish request if any network error ocuring,
485 but it's true only for synchronus enqueue, for async enqueue
486 (via ptlrpcd) this incorrect and ptlrpcd want finish request
489 Severity : enhancement
491 Description: client-side metadata stat-ahead during readdir(directory readahead)
492 Details : perform client-side metadata stat-ahead when the client detects
493 readdir and sequential stat of dir entries therein
496 Frequency : on start mds
498 Description: Implement get_info(last_id) in obdfilter.
501 Frequency : occasional
503 Description: Correctly check stale fid, not start epoch if ost not support SOM
504 Details : open with flag O_CREATE need set old fid in op_fid3 because
505 op_fid2 was overwritten with new generated fid, but mds can answer
506 with one of these two fids and both is not stale. Setattr
507 incorrectly started an epoch and assume will be called
508 done_writing, but without SOM done_writing ever being called.
511 Frequency : rare, depends on device drivers and load
513 Description: MDS or OSS nodes crash due to stack overflow
514 Details : Code changes in 1.8.0 increased the stack usage of some functions.
515 In some cases, in conjunction with device drivers that use a lot
516 of stack the MDS (or possibly OSS) service threads could overflow
517 the stack. One change which was identified to consume additional
518 stack has been reworked to avoid the extra stack usage.
521 Frequency : occasional
523 Description: Do not fail import if osc_interpret_create gets -EAGAIN
524 Details : If osc_interpret_create got -EAGAIN it immediately exits and
525 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
526 oscc_has_objects and see OSC has no objests and call
527 oscc_internal_create to resend create request.
529 Severity : enhancement
531 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
533 Severity : enhancement
535 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
538 Frequency : very rare
540 Description: Fix replay if there is an un-replied request and open
541 Details : In some cases, older replay request will revert the
542 mcd->mcd_last_xid on MDS which is used to record the client's
545 Severity : enhancement
547 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
549 Severity : enhancement
551 Description: Add rhel5 support to HEAD.
553 Serverity : enhancement
555 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
560 Description: Don't allow skipping OSTs if index has been specified.
561 Details : Don't allow skipping OSTs if index has been specified, make
562 locking in internal create lots better.
566 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
567 Details : access to bitfield in structure is always rounded to long
568 and this produce problem with not atomic change any bit.
572 Description: Lustre make rpms failed.
573 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
574 builds Lustre rpms from tarball.
577 Frequency : rare on shutdown ost
579 Description: If llog cancel was not send before clean_exports phase, this can
580 produce deadlock in llog code.
581 Details : If llog thread has last reference to obd and call class_import_put
582 this produce deadlock because llog_cleanup_commit_master wait when
583 last llog_commit_thread exited, but this never success because was
584 called from llog_commit_thread.
588 Description: allow userland application know is lost one of stripes.
589 Details : fill lvb_blocks with error code on ost and return it to
590 application if error flag found.
594 Description: NULL lov_tgts causing MDS oops
595 Details : more safe checks for NULL lov_tgts for avoid oops.
597 Severity : enhancement
599 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
603 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
604 Details : If an OST is down the MDS will hang indefinitely in
605 obd_statfs_rqset() waiting for the statfs data. While for
606 MDS QOS usage of statfs, it should not stuck in waiting.
608 Severity : enhancement
610 Description: remote_acl support
611 Details : Support ACL-based permission check for remote user.
612 Support setfacl/getfacl for remote user with the utils
613 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
614 the system "{s,g}etfacl" utils.
616 Severity : enhancement
618 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
620 Severity : enhancement
622 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
626 Description: make mgs_setparam() handle fsname containing dash
627 Details : fsname containing a dash does not work with lctl conf_param
629 Severity : enhancement
631 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
633 Severity : enhancement
635 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
639 Frequency : rare, at shutdown
640 Description: access already free / zero obd_namespace.
641 Details : if client_disconnect_export was called without force flag set,
642 and exist connect request in flight, this can produce access to
643 NULL pointer (or already free pointer) when connect_interpret
644 store ocd flags in obd_namespace.
648 Frequency : only at startup
649 Description: not alloc memory with spinlock held.
650 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
651 if any spinlock held.
653 Severity : enhancement
655 Description: make lustre randomly fail allocating memory
656 Details : Make lustre randomly failed allocating memory for testing purpose.
658 Severity : enhancement
660 Description: lost problems with lov objid file
661 Details : Fixes some scability and access to not inited memory problems
662 in work with lov objdid file.
667 Description: lfs find does not continue on file error
668 Details : Continue other files processing when a file/dir is absent.
672 Description: Inconsistent usage of lustre_pack_reply()
673 Details : Standardize the usage of lustre_pack_reply() such that it
674 always generate a CERROR on failure.
677 Frequency : occasional
679 Description: MDS hang or stay in waiting lock
680 Details : If client receive lock with CBPENDING flag ldlm need send lock
681 cancel as separate rpc, to avoid situation when cancel request
682 can't processed due all i/o threads stay in wait lock.
686 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
690 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
694 Description: Update to SLES9 kernel-2.6.5-7.287.3.
698 Description: Update to RHEL4 latest kernel.
700 Severity : enhancement
702 Description: Build SLES10 patchless client fails
703 Details : The configure was broken by run ./configure with
704 --with-linux-obj=.... argument for patchless client. When the
705 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
706 can't search header adequately. Use absolute path such as
707 -I($LINUX)/include instead.
711 Description: interrupt oig_wait produce painc on resend.
712 Details : brw_redo_request can be used for resend requests from ptlrpcd and
713 private set, and this produce situation when rq_ptlrpcd_data not
714 copyed to new allocated request and triggered LBUG on assert
715 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
716 ptlrpcd set if request is changed and can be safety changed to use
721 Description: LASSERT_{REQ,REP}SWAB macros are buggy
722 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
723 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
724 these macros and replace them with their difinitions instead.
728 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
729 Details : Update which_patch & target file for SLES10 latest kernel.
732 Description: Remove client patches from SLES 10 kernel.
733 Details : This causes SLES 10 clients to behave as patchless clients
734 even on a Lustre-patched (server) kernel.
736 Severity : enhancement
738 Description: self-adjustable client's lru lists
739 Details : use adaptive algorithm for managing client cached locks lru
740 lists according to current server load, other client's work
741 pattern, memory activities, etc. Both, server and client
742 side namespaces provide number of proc tunables for controlling
745 Severity : enhancement
747 Description: light-weight GSS support
748 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
749 and reduce performance overhead.
751 Severity : enhancement
753 Description: Linux keyring support
754 Details : Support using service of Linux keyring for Lustre GSS internal
755 context refresh/cache mechanism.
759 Description: Fix errors in lfs documentation
760 Details : Fixes man pages
764 Description: don't use GFP_* in generic Lustre code.
765 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
769 Description: obdclass is limited by single OBD_ALLOC(idarray)
770 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
774 Description: warnings with build patchless client with vanila 2.6.19 and up
775 Details : change old ctl_table style and replace ctl_table/ctl_table_header
776 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
780 Frequency : Only for SLES
781 Description: sanity_quota fail test_1
782 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
783 file, which makes TSTID[2] inunique.
787 Frequency : Only for SLES9
788 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
789 Details : Update target/ChangeLog/which_patch .
793 Frequency : Only for test_10 in sanity-quota.sh
794 Description: a bug in quota test code
795 Details : checking a test flag in a wrong place
799 Frequency : Only for test_13 in sanity-quota.sh
800 Description: a bug in quota test script
801 Details : shouldn't check overfull allocated quota
805 Description: jbd statistics
806 Details : Port older jbd statistics patch for sles10
808 Severity : enhancement
810 Description: Add printing inode info into message about error in writepage.
815 Description: reply_lock_interpret crash due to race with it and lock cancel.
816 Details : Do not replay locks that are being cancelled. Do not reference
817 locks by their address during replay, just by their handle.
821 Frequency : When flocks are used.
822 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
823 Details : Flock locks might destroy just granted lock if it could be merged
824 with another existing flock, this is done in completion handler,
825 so teach ldlm_cli_enquque_fini that this is a valid case for
831 Description: Oops in read and write path when failing to allocate lock.
832 Details : Check if lock allocation failed and return error back.
836 Description: lstripe command fails for valid OST index
837 Details : The stripe offset is compared to lov->desc.ld_tgt_count
838 instead of lov->desc.ld_active_tgt_count.
842 Description: sanity.sh failed test 103
843 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
848 Description: df doesn't work properly if diskfs blocksize != 4K
849 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
853 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
854 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
856 Severity : enhancement
858 Description: Adaptive timeouts
859 Details : RPC timeouts adapt to changing server load and network
860 conditions to reduce resend attempts and improve recovery time.
864 Description: llapi_file_create() does not allow some changes
865 Details : add llapi_file_open() that allows specifying the mode and
866 open flags, and also returns an open file handle.
870 Description: merge and cleanup kernel patches.
871 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
875 Description: Add journal checksum support.(Kernel part)
876 Details : The journal checksum feature adds two new flags i.e
877 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
878 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
879 indicates that the commit block contains the checksum for
880 the blocks described by the descriptor blocks. Now commit
881 record can be sent to disk without waiting for descriptor
882 blocks to be written to disk. This behavior is controlled
883 using JBD2_FEATURE_ASYNC_COMMIT flag.
887 Description: OSS needs mutliple precreate threads
888 Details : Add ability to start more than one create thread per OSS.
892 Description: Sanity.sh test_65k failed: setstripe should have succeeded
893 Details : execute lfs setstripe on client
897 Description: mds_obd_create error creating tmp object
898 Details : When the user sets quota on root, llog will be affected and can't
899 create files and write files.
902 Frequency : Always on ia64 patchless client, and possibly others.
904 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
905 Details : This allows the patchless client to be loaded on architectures
910 Description: RedHat Update kernel for RHEL5
911 Details : Add the kernel config file for RHEL5.
915 Description: RedHat Update kernel for RHEL5
916 Details : Modify the kernel config file more closer RHEL5.
920 Description: Build failure against Centos5 (RHEL5)
921 Details : Define PAGE_SIZE when it isn't present.
925 Description: test_7 on sanity-quota.sh can't be used on separate machines
926 Details : add facet to handle it
930 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
931 Details : with NFS, the anon dentry's parent was set to itself in
932 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
933 corresponding dentry other than use rec->ur_name.
935 Severity : enhancement
937 Description: lfs setstripe enhancement
938 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
942 Description: enable data checksumming by default
943 Details : enable checksum by default, allow --disable-checksum
944 configure option and "-o nochecksum" mount option. Checksums
945 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
949 Description: lustre support for RHEL5
950 Details : Add support for RHEL5.
954 Description: Client eviction due to failover config
955 Details : after a connection loss, the lustre client should attempt to
956 reconnect to the last active server first before trying the
957 other potential connections.
961 Description: when mds and osts use different quota unit(32bit and 64bit),
962 quota will be released repeatly.
963 Details : void sending multiple quota reqs to mds, which will keep the
964 status between the reqs.
968 Description: osts not allocated evenly to files
969 Details : change the condition to increase offset_idx
972 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
974 Description: Data corruption for OSTs that are formatted larger than 2TB
976 Details : When generating the bio request for lustre file writes the
977 sector number would overflow a temporary variable before being
978 used for the IO. The data reads correctly from Lustre (which
979 will overflow in a similar manner) but other file data or
980 filesystem metadata may be corrupted in some cases.
984 Description: Tune the kernel for good SCSI performance.
985 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
986 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
991 Description: rewrite ext2-derived code in obdclass/uuid.c
992 Details : rewrite inherited code (uuid parsing code from ext2 utils)
993 from scratch preserving functionality.
997 Description: "lfs find -obd UUID" prints directories
998 Details : "lfs find -obd UUID" will return all directory names instead
999 of just file names. It is incorrect because the directories
1000 do not reside on the OSTs.
1004 Description: use i_size_read and i_size_write in 2.6 port
1005 Details : replace inode->i_size access with i_size_read/write()
1008 Frequency : when removing large files
1010 Description: scheduling issue during removal of large Lustre files
1011 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1012 It causes scheduling issues when removing large files (17TB in the
1016 Frequency : only with liblustre clients on XT3
1018 Description: evictions taking too long
1019 Details : allow llrd to evict clients directly on OSTs
1022 Frequency : only on ppc
1024 Description: /proc/fs/lustre/devices broken on ppc
1025 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1026 We should make sure the type of 'index' is loff_t and then cast
1027 explicitly as needed below. Do not assign an explicitly cast
1031 Frequency : only for rhel5
1033 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1034 Details : Modify the target file & which_kernel.
1036 Severity : enhancement
1038 Description: omit set fsid for export NFS
1039 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1040 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1044 Frequency : Always, for kernels after 2.6.16
1045 Description: Fix warning idr_remove called for id=.. which is not allocated.
1046 Details : Last kernels save old s_dev before kill super and not allow
1047 to restore from callback - restore it before call kill_anon_super.
1051 Description: buffer overruns could theoretically occur
1052 Details : llapi_semantic_traverse() modifies the "path" argument by
1053 appending values to the end of the origin string, and a
1054 overrun may occur. Adding buffer overrun check in liblustreapi.
1058 Description: Fix error on 'ls .' at the top of the Lustre mount.
1059 Details : Don't revalidate dentry if it is a root dentry.
1063 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1064 Details : Modify vm-tunables-rhel4.patch.
1068 Description: Kernel config for 2.6.18-vanilla.
1069 Details : Modify targets/2.6-vanilla.target.in.
1070 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1071 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1072 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1073 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1077 Description: improve handling recoverable errors
1078 Details : if request processig with error which can be recoverable on server
1079 request should be resend, otherwise page released from cache and
1084 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1085 Details : Modify target file & which_patch.
1086 A flaw was found in the IA32 system call emulation provided
1087 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1088 value could be stored in the %RAX register, which could trigger an
1089 out-of-bounds system call table access. An untrusted local user
1090 could exploit this flaw to run code in the kernel
1091 (ie a root privilege escalation). (CVE-2007-4573).
1095 Description: O_DIRECT bypasses client statistics.
1096 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1097 read_ahead_stats were updated. copied the stats section from
1098 osc_send_oap_rpc() into async_internal().
1102 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1106 Description: change order of libsysio includes
1107 Details : '#include sysio.h' should always come before '#include xtio.h'
1111 Description: handle "serious error: objid * already exists" more gracefully
1112 Details : If LAST_ID value on disk is smaller than the objects existing in
1113 the O/0/d* directories, it indicates disk corruption and causes an
1114 LBUG(). If the object is 0-length, then we should use the existing
1115 object. This will help to avoid a full fsck in most cases.
1117 Severity : enhancement
1119 Description: adapt the lustre_config script to support the upgrade case
1120 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1121 server targets to 1.6 in parallel.
1125 Description: To avoid grant space > avaible space when the disk is almost
1126 full. Without this patch you might see the error "grant XXXX >
1127 available" or some LBUG about grant, when the disk is almost
1129 Details : In filter_check_grant, for non_grant cache write, we should
1130 check the left space by if (*left > ungranted + bytes), instead
1131 of (*left > ungranted), because only we are sure the left space
1132 is enough for another "bytes", then the ungrant space should be
1133 increase. In client, we should update cl_avail_grant only there
1134 is OBD_MD_FLGRANT in the reply.
1138 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1139 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1140 described in CVE-2007-4573. This problem could allow untrusted
1141 local users to gain root access.
1144 Frequency : when using O_DIRECT and quotas
1146 Description: Incorrect file ownership on O_DIRECT output files
1147 Details : block usage reported by 'lfs quota' does not take into account
1148 files that have been written with O_DIRECT.
1153 Description: touch file failed when fs is not full
1154 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1155 otherwise we can get ENOSP while fs is not full.
1159 Description: parallel lock callbacks
1160 Details : Instead of sending blocking and completion callbacks as separated
1161 requests, adding them to a set and sending in parallel.
1164 Frequency : only for Cray XT3
1165 Bugzilla : 12829/13455
1166 Description: Changing primary group doesn't change the group lustre assigns to
1168 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1169 overridden with the primary group provided by the group upcall,
1170 whereas the supplied fsgid can be trusted if it is in the list of
1171 supplementary groups returned by the group upcall.
1173 Severity : enhancement
1175 Description: Allow masking D_WARNING, D_ERROR messages from console
1176 Details : Console messages can now be disabled via lnet.printk.
1180 Description: User code with malformed file open parameter crashes client node
1181 Details : Before packing join_file req, all the related reference should be
1182 checked carefully in case some malformed flags cause fake
1183 join_file req on client.
1187 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1188 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1189 mode is equal to the granted mode. However, it should also check
1190 that LDLM_FL_CP_REQD is not set.
1194 Description: Heavy nfs access might result in deadlocks
1195 Details : After ELC code landed, it is now improper to enqueue any mds
1196 locks under och_sem, because enqueue might want to decide to
1197 cancel open locks for same inode we are holding och_sem for.
1201 Description: Client eviction while running blogbench
1202 Details : A lot of unlink operations with concurrent I/O can lead to a
1203 deadlock causing evictions. To address the problem, the number of
1204 oustanding OST_DESTROY requests is now throttled to
1205 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1206 callbacks are processed in priority.
1210 Description: enable ACLs on MDS by default
1211 Details : ACLs must be enabled on MDS by default.
1214 Frequency : PPC/PPC64 only
1216 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1217 Details : fix duplicated definitions between asm-ppc64/types.h and
1218 lustre_types.h on PPC.
1221 Frequency : PPC/PPC64 only
1223 Description: asm-ppc/segment.h does not exist
1224 Details : fix compile issue on PPC.
1228 Description: better handle error messages in extents code
1231 Frequency : RHEL4 only
1233 Description: mkfs is very slow on IA64/RHEL4
1234 Details : A performance regression has been discovered in the MPT Fusion
1235 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1236 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1237 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1239 Severity : enhancement
1241 Description: SNMP support enhancement
1242 Details : Adding total number of sampled request for an MDS node in snmp
1245 Severity : enhancement
1247 Description: Optimize ldlm waiting list processing for PR extent locks
1248 Details : When processing waiting list for read extent lock and meeting read
1249 lock that is same or wider to it that is not contended, skip
1250 processing rest of the list and immediatelly return current
1251 status of conflictness, since we are guaranteed there are no
1252 conflicting locks in the rest of the list.
1256 Description: Time out and refuse to reconnect
1257 Details : When the failover node is the primary node, it is possible
1258 to have two identical connections in imp_conn_list. We must
1259 compare not conn's pointers but NIDs, otherwise we can defeat
1260 connection throttling.
1264 Description: port llog fixes from b1_6 into HEAD
1265 Details : Port llog reference couting and some llog cleanups from b1_6
1266 (bug 10800) into HEAD, for protect from panic and access to already
1267 free llog structures.
1271 Description: Detect stride IO mode in read-ahead
1272 Details : When a client does stride read, read-ahead should detect that and
1273 read-ahead pages according to the detected stride pattern.
1277 Description: data checksumming impacts single node performance
1278 Details : add support for several checksum algorithm. Currently, only CRC32
1279 and Adler-32 are supported. The checksum type can be changed on
1280 the fly via /proc/fs/lustre/osc/*/checksum_type.
1284 Description: use adler32 for page checksums
1285 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1290 Description: build for x2 fails
1291 Details : fix compile issue on Cray systems.
1295 Description: Properly match for duplicate locks
1296 Details : Due to different lock order from skiplists code, we need to
1297 traverse entire list for now
1300 Frequency : only on PPC/SLES10
1302 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1303 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1304 be defined. Add a hack in mkfs_lustre.c to work around this compile
1309 Description: LASSERT on MDS when client holding flock lock dies
1310 Details : ldlm pool logic depends on number of granted locks equal to
1311 number of released locks which is not true for flock locks, so
1312 just exclude such locks from consideration.
1316 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1317 Details : Use fsfilt_sync() for both the whole filesystem sync and
1318 individual file sync to eliminate dangerous inode locking
1319 with I_LOCK that can lead to a deadlock.
1323 Description: performance in 1.6.3
1324 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1325 sure that 1MB requests are not fragmented by the block layer.
1327 Severity : enhancement
1329 Description: organize the server-side client stats on per-nid basis
1330 Details : Change the structure of stats under obdfilter and mds to
1338 The "uuid"s file would list the uuids of _active_ exports.
1339 And the clear entry is to clear all stats and stale nids.
1341 Severity : enhancement
1343 Description: eliminate client locks in face of contention
1344 Details : file contention detection and lockless i/o implementation
1345 for contended files.
1349 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1352 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1354 Description: MDS may lose file striping (and hence file data) in some cases
1355 Details : If there are additional extended attributes stored on the MDS,
1356 in particular ACLs, SELinux, or user attributes (if user_xattr
1357 is specified for the client mount options) then there is a risk
1358 of attribute loss. Additionally, the Lustre file striping
1359 needs to be larger than default (e.g. striped over all OSTs),
1360 and an additional attribute must be stored initially in the
1361 inode and then increase in size enough to be moved to the
1362 external attribute block (e.g. ACL growing in size) for file
1367 Description: skiplist implementation simplification
1368 Details : skiplists are used to group compatible locks on granted list
1369 that was implemented as tracking first and last lock of each
1370 lock group the patch changes that to using doubly linked lists
1374 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1375 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1376 because LASSERT on both the data supplied by a client, and the
1377 data on disk is dangerous and incorrect.
1379 Severity : enhancement
1381 Description: Slow truncate/writes to huge files at high offsets.
1382 Details : Directly associate cached pages to lock that protect those pages,
1383 this allows us to quickly find what pages to write and remove
1384 once lock callback is received.
1388 Description: more ldlm soft lockups
1389 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1390 starve other threads from the resource lock for a long time in
1391 case of long waiting queue, so change the debug level from
1392 D_OTHER to the less frequently used D_INFO.
1394 Severity : enhancement
1396 Description: add -gid, -group, -uid, -user options to lfs find
1400 Description: Hung threads in invalidate_inode_pages2_range
1401 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1402 once one is completed. As a result, some RPCs are stuck in the
1403 queue and are never sent.
1407 Description: filter threads hungs on waiting journal commit
1408 Details : Cleanup filter group llog code, then only filter group llog will
1409 be only created in the MDS/OST syncing process.
1413 Description: Procfs and llog threads access destoryed import sometimes.
1414 Details : Sync the import destoryed process with procfs and llog threads by
1415 the import refcount and semaphore.
1417 Severity : enhancement
1419 Description: openlock cache of b1_6 port to HEAD
1424 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1425 Details : If insertion of an extent fails, then discard the inode
1426 preallocation and free data blocks else it can lead to duplicate
1431 Description: don't always update ctime in ext3_xattr_set_handle()
1432 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1433 In some cases the ctime should not be updated, for example for
1434 2.0->1.8 compatibility it is necessary to delete an xattr and it
1435 should not update the ctime.
1439 Bugzilla : 15713/16362
1440 Description: Assertion in iopen_connect_dentry in 1.6.3
1441 Details : looking up an inode via iopen with the wrong generation number can
1442 populate the dcache with a disconneced dentry while the inode
1443 number is in the process of being reallocated. This causes an
1444 assertion failure in iopen since the inode's dentry list contains
1445 both a connected and disconnected dentry.
1449 Description: assertion failure in ldlm_handle2lock()
1450 Details : fix a race between class_handle_unhash() and class_handle2object()
1451 introduced in lustre 1.6.5 by bug 13622.
1456 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1457 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1462 Description: oops in sd_iostats_seq_show()
1463 Details : unloading/reloading the scsi low level driver triggers a kernel
1464 bug when trying to access the sd iostat file.
1469 Description: Kernel panics during QLogic driver reload
1470 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1471 patch, causing memory corruption.
1476 Description: journal_dev option does not work in b1_6
1477 Details : pass mount option during pre-mount.
1479 Severity : enhancement
1481 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1482 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1483 extent information of a file. It can be used to map logical blocks
1484 in a file to physical blocks in the block device.
1488 Description: LDLM soft lockups - improvement
1489 Details : It is be possible to send the lock handle along with each read
1490 or write request because the client is already doing a lock match
1491 itself so there isn't any reason the OST should have to re-do that
1495 Frequency : only on Cray X2
1497 Description: X2 build failures
1498 Details : fix build failures on Cray X2.
1502 Description: xid & resent requests
1503 Details : Initialize RPC XID from clock at startup (randomly if clock is
1506 Severity : enhancement
1508 Description: Add lustre_start utility to start or stop multiple Lustre servers
1513 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1514 Details : In case of memory pressure, list_del() can be called twice on
1515 req->rq_history_list, causing a kernel oops.
1519 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1520 Details : fix stack overflow in the distributed lock manager by defering
1521 export eviction after a failed AST to the elt thread instead of
1522 handling it in the dlm interpret routine.
1526 Description: Convert some comments to new format.
1527 Details : Update documenting comments to match doxygen conventions.
1531 Description: Grammar fixes.
1532 Details : A couple of trivial spelling fixes.
1536 Description: OSD_COUNTERS-mandatory
1537 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1538 to simplify the code.
1542 Description: simplify cmm_device freeing logic.
1543 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1547 Description: Add lockdep support to dt_object_operations locking interface.
1548 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1549 parameter indicating lock ordering. Update mdd code to use new
1554 Description: Introduce failloc constants for lockless IO tests.
1555 Details : Add two new failloc constants to test lockless IO. Only one of
1556 them in implemented---another is checked in yet to be landed
1561 Description: Add lockdep support for inode mutex.
1562 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1563 in the situations where Lustre has to lock more than one inode
1568 Description: Add optional invariants checking support.
1569 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1570 configure switch. This macro is to be used for consistency and
1571 sanity checks that are too expensive to be left in `production'
1576 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1577 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1582 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1583 Details : Use spin_lock_nested() in (the only) situation where more than
1584 one ldlm_lock is locked simultaneously. Also, fix possible
1585 dead-lock in ldlm_lock_change_resource() by enforcing particular
1590 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1591 Details : Instead of specifying each ldlm_lock call-back through separate
1592 parameter, wrap them into struct ldlm_callback_suite.
1596 Description: Kill join_lru obd method and its callers.
1597 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1598 in memory---a requirement implemented through ->o_join_lru() obd
1599 method in HEAD. Remove this method and its users.
1603 Description: Add asynchronous ldlm ENQUEUE completion handler.
1604 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1605 case that stock ldlm_completion_ast() cannot handle as it waits
1606 until lock is granted. Introduce new ldlm_completion_ast_async()
1607 for this. Also comment ldlm_completion_ast().
1611 Description: ldlm_error <-> errno conversion.
1612 Details : Add functions to map (rather arbitrary) between LDLM error codes
1613 and standard errno values. CLIO needs this to prevent LDLM specific
1614 constants from escaping ldlm and osc.
1618 Description: Kill unused ldlm_handle2lock_ns() function.
1619 Details : Kill unused ldlm_handle2lock_ns() function.
1623 Description: Add lu_ref support to ldlm_lock
1624 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1625 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1626 and ldlm_resource. LDLM interface has to be changed, because code that
1627 releases a reference on a lock, has to "know" what reference this is.
1628 In the most frequent case
1630 lock = ldlm_handle2lock(handle);
1632 LDLM_LOCK_PUT(lock);
1634 no changes are required. When any other reference (received _not_ from
1635 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1636 instead of LDLM_LOCK_PUT().
1638 Arguably, changes are pervasive, and interface requires some discipline
1639 for proper use. On the other hand, it was very instrumental in finding
1640 a few leaked lock references.
1644 Description: Add ldlm_lock_addref_try().
1645 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1646 attempts to addref a lock that might be being canceled
1651 Description: Add ldlm_weigh_callback().
1652 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1653 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1654 hard-coded `number of pages' logic.
1658 Description: Add lockdep annotations to llog code.
1659 Details : Use appropriately tagged _nested() locking calls in the places
1660 where llog takes more than one ->lgh_lock lock.
1664 Description: Add loi_kms_set().
1665 Details : Wrap kms updates into a helper function.
1669 Description: Constify instances of struct lsm_operations.
1670 Details : Constify instances of struct lsm_operations.
1674 Description: lu_conf support.
1675 Details : On a server, a file system object is uniquely identified
1676 by a fid, which is sufficient to locate and load all object
1677 state (inode). On a client, on the other hand, more data are
1678 necessary instantiate an object. Change lu_object_find() and
1679 friends to take additional `lu_conf' argument describing object.
1680 Typically this includes layout information.
1684 Description: lu_context fixes.
1685 Details : Introduce new lu_context functions that are needed on the client
1686 side, where some system threads (ptlrpcd) are shared by multiple
1687 modules, and so cannot be stopped during module shutdown.
1691 Description: Add start and stop methods to lu_device_type_operations.
1692 Details : Introduce two new methods in lu_device_type_operations, that are
1693 invoked when first instance of a given type is created and last one
1694 is destroyed respectively. This is need by CLIO.
1698 Description: Add lu_ref support to struct lu_device.
1699 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1700 track leaked references.
1704 Description: Introduce lu_kmem_descr.
1705 Details : lu_kmem_descr and its companion interface allow to create
1706 and destroy a number of kmem caches at once.
1710 Description: Fix lu_object finalization race.
1711 Details : Fix a race between lu_object_find() finding an object and its
1712 concurrent finalization. This race is (most likely) not possible
1713 on the server, but might happen on the client.
1717 Description: Introduce lu_ref interface.
1718 Details : lu_ref is a debugging module allowing to track references to
1719 a given object. It is quite cpu expensive, and has to be
1720 explicitly enabled with --enable-lu_ref. See usage description
1725 Description: Factor lu_site procfs stats into a separate function.
1726 Details : Separate lu_site stats printing code into a separate function
1727 that can be reused on a client.
1731 Description: Constify instances of struct {lu,dt,md}_device_operations.
1732 Details : Constify instances of struct {lu,dt,md}_device_operations.
1736 Description: Introduce struct md_site and move meta-data specific parts of
1737 struct lu_site here.
1738 Details : Move md-specific fields out of struct lu_site into special struct
1739 md_site, so that lu_site can be used on a client.
1743 Description: Kill mdd_lov_destroy().
1744 Details : Remove unused mdd code.
1748 Description: Add st_block checking to multistat.c.
1749 Details : Add st_block checking to multistat.c.
1753 Description: Add lu_ref support to struct obd_device.
1754 Details : Add lu_ref tracking to obd_device.
1758 Description: Kill obd_set_fail_loc().
1759 Details : Remove unused code.
1763 Description: Add special type for ptlrpc_request interpret functions.
1764 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1765 there. Actual usage will be in CLIO.
1769 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1770 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1771 doesn't work with lockdep.
1775 Description: Add rwv.c test program.
1776 Details : New testing program exercising readv(2) and writev(2) (Qian).
1780 Description: Add sendfile.c test program.
1781 Details : New testing program exercising sendfile(2) (Jay).
1785 Description: Ratelimit a message that can be very frequent.
1786 Details : Ratelimit a memory allocation failure message that can
1791 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1792 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1796 Description: Liblustre build fixes.
1797 Details : Liblustre build fixes.
1801 Description: libcfs: add cfs_{need,cond}_resched() interface.
1802 Details : libcfs: add cfs_{need,cond}_resched() definition and
1803 implementations for Linux, NT, and liblustre.
1805 Severity : enhancement
1807 Description: More exported tunables for mballoc
1808 Details : Add support for tunable preallocation window and new tunables for
1809 large/small requests
1813 Description: Detect corruption of block bitmap and checking for preallocations
1814 Details : Checks validity of on-disk block bitmap. Also it does better
1815 checking of number of applied preallocations. When corruption is
1816 found, it turns filesystem readonly to prevent further corruptions.
1820 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1821 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1822 window to 0. If it does hit cache miss, and read-pattern is still
1823 stride-io mode, does not reset the stride window, but also does not
1824 increase the stride window length in this case.
1828 Frequency : only for big-endian servers
1829 Description: Check if system is big-endian while mounting fs with extents feature
1830 Details : Mounting a filesystem with extents feature will fail on big-endian
1831 systems since ext3-based ldiskfs is not supported on big-endian
1832 systems. This can be over-riden with "bigendian_extents" mount option.
1834 Severity : enhancement
1836 Description: The root squash functionality
1837 Details : A security feature, which is to prevent users from being able
1838 to mount lustre on their desktop, run as root, and delete
1839 all of the files in the filesystem. The goal is accomplished by
1840 remapping user id (UID) and group id (GID) of the root user to
1841 a UID and GID specified by the system administartor via Lustre
1842 configuration management server (MGS). The functionality also
1843 allows to specify sets of clients for which the remapping does
1848 Description: Excessive recovery window
1849 Details : With AT enabled, the recovery window can be excessively long (6000+
1850 seconds). To address this problem, we no longer use
1851 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1852 timeout no longer depends on the service time, it is set to
1853 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1854 time via pb_service_time.
1856 --------------------------------------------------------------------------------
1858 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1860 * Support for kernels:
1861 2.6.5-7.283 (SLES 9),
1862 2.6.9-55.EL (RHEL 4),
1863 2.6.16.46-0.14 (SLES 10),
1864 2.6.18.8 vanilla (kernel.org)
1865 * Client support for unpatched kernels:
1866 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1867 2.6.16 - 2.6.22 vanilla (kernel.org)
1868 * Due to recently discovered recovery problems, we do not recommend
1869 using patchless RHEL 4 clients with this or any earlier release.
1870 * Recommended e2fsprogs version: 1.39.cfs8
1871 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1872 * Starting with this release, the ldiskfs backing filesystem required
1873 by Lustre is now in its own package, lustre-ldiskfs. This package
1874 should be installed. It is versioned separately from Lustre and
1875 may be released separately in future.
1879 Description: block reactivating mgc import until all deactivates complete
1880 Details : Fix race when failing back MDT/MGS to itself (testing)
1883 Frequency : at statup only
1885 Description: mds_lov_synchronize race leads to various problems
1886 Details : simultaneous MDT->OST connections at startup can cause the
1887 sync to abort, leaving the OSC in a bad state.
1889 Severity : enhancement
1891 Description: add optional extra BUILD_VERSION info
1892 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1893 to override the lustre version.
1896 Frequency : 2.6.18 servers only
1898 Description: ll_kern_mount() doesn't release the module reference
1899 Details : The ldiskfs module reference count never drops down to 0
1900 because ll_kern_mount() doesn't release the module reference.
1905 Description: server LBUG when using old ost_num_threads parameter
1906 Details : Accept the old ost_num_threads parameter but warn that it
1907 is deprecated, and fix an off-by-one error that caused an LBUG.
1912 Description: Transient SCSI error results in persistent IO issue
1913 Details : iobuf->dr_error is not reinitialized to 0 between two
1917 Frequency : sometimes when underlying device returns I/O errors
1919 Description: OSTs not going read-only during write failures
1920 Details : OSTs are not remounted read-only when the journal commit threads
1921 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1922 instead of the ext3 wrappers.
1926 Description: poor connect scaling with increasing client count
1927 Details : Don't run filter_grant_sanity_check for more than 100 exports
1928 to improve scaling for large numbers of clients.
1931 Frequency : SLES10 only
1933 Description: sanity-quota.sh quotacheck failed: rc = -22
1934 Details : Quotas cannot be enabled on SLES10.
1937 Frequency : liblustre clients only
1939 Description: getdirentries does not give error when run on compute nodes
1940 Details : getdirentries does not fail when the size specified as an argument
1941 is too small to contain at least one entry
1943 Severity : enhancement
1945 Description: Add LNET router traceability for debug purposes
1946 Details : If a checksum failure occurs with a router as part of the
1947 IO path, the NID of the last router that forwarded the bulk data
1948 is printed so it can be identified.
1953 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1954 Details : Due to a race condition, liblustre clients were occasionally
1955 evicted incorrectly.
1957 Severity : enhancement
1959 Description: lfs setstripe use optional parameters instead of postional
1962 Severity : enhancement
1964 Description: Nanosecond timestamp support for ldiskfs
1965 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1966 resolution timestamps. There is not yet support for this at
1967 the Lustre filesystem level.
1970 Frequency : during server recovery
1972 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1973 Details : request with rq_no_resend flag not awake l_wait_event if they get
1977 Frequency : nfs export on patchless client
1979 Description: connectathon hang when test nfs export over patchless client
1980 Details : Disconnected dentry cannot be found with lookup, so we do not need
1981 to unhash it or make it invalid
1984 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1985 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1986 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1991 Description: open req refcounting wrong on reconnect
1992 Details : If reconnect happened between getting open reply from server and
1993 call to mdc_set_replay_data in ll_file_open, we will schedule
1994 replay for unreferenced request that we are about to free.
1995 Subsequent close will crash in variety of ways.
1996 Check that request is still eligible for replay in
1997 mdc_set_replay_data().
2002 Description: disable writes to filesystem when reading health_check file
2003 Details : the default for reading the health_check proc file has changed
2004 to NOT do a journal transaction and write to disk, because this
2005 can cause reads of the /proc file to hang and block HA state
2006 checking on a healthy but otherwise heavily loaded system. It
2007 is possible to return to the previous behaviour during configure
2008 with --enable-health-write.
2010 Severity : enhancement
2012 Description: 64-bit inode version
2013 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2014 to the inode. This will be required for version-based recovery.
2019 Description: MDS fails to start if a duplicate client export is detected
2020 Details : in some rare cases it was possible for a client to connect to
2021 an MDS multiple times. Upon recovery the MDS would detect this
2022 and fail during startup. Handle this more gracefully.
2024 Severity : enhancement
2026 Description: Add -o localflock option to simulate old noflock
2028 Details : This will achieve local-only flock/fcntl locks
2034 Description: log_commit_thread vs filter_destroy race leads to crash
2035 Details : Take import reference before releasing llog record semaphore
2040 Description: Wrong request locking in request set processing
2041 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2042 imp_delayed_list, in this place should be used imp_lock.
2045 Frequency : when reconnection
2047 Description: Grant Leak when osc reconnect to OST
2048 Details : When osc reconnect ost, OST(filter) should check whether it
2049 should grant more space to client by comparing fed_grant and
2050 cl_avail_grant, and return the granted space to client instead
2051 of "new granted" space, because client will call osc_init_grant
2052 to update the client grant space info.
2055 Frequency : when client reconnect to OST
2057 Description: Grant Leak when osc do resend and replay bulk write
2058 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2059 bulk write request, because the grant info while be sync between
2060 OSC and OST when reconnect, and we should ignore the grant info
2061 these of resend/replay write req.
2066 Description: Grant space more than avaiable left space sometimes.
2067 Details : When then OST is about to be full, if two bulk writing from
2068 different clients came to OST. Accord the avaliable space of the
2069 OST, the first req should be permitted, and the second one
2070 should be denied by ENOSPC. But if the seconde arrived before
2071 the first one is commited. The OST might wrongly permit second
2072 writing, which will cause grant space > avaiable space.
2075 Frequency : when client is evicted
2077 Description: Grant might be wrongly erased when osc is evicted by OST
2078 Details : when the import is evicted by server, it will fork another
2079 thread ptlrpc_invalidate_import_thread to invalidate the
2080 import, where the grant will be set to 0. While the original
2081 thread will update the grant it got when connecting. So if
2082 the former happened latter, the grant will be wrongly errased
2083 because of this race.
2088 Description: Checking Stale with correct fid
2089 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2090 to check whether it is stale, because sometimes, we want the
2091 enqueue happened anyway, and op_data.fid2 will not be initialized.
2093 Severity : enhancement
2095 Description: update patchless client
2096 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2099 Frequency : only with 2.4 kernel
2101 Description: random memory corruption
2102 Details : size of struct ll_inode_info is to big for union inode.u and this
2103 can be cause of random memory corruption.
2108 Description: Memory leak in recovery
2109 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2110 It should also check obd_fail before fsfilt_start, otherwise if
2111 fsfilt_start return -EROFS,(failover mds during mds recovery).
2112 then the req will return with repmsg->transno = 0 and rc = EROFS.
2113 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2114 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2115 be freed no matter whether fsfilt_commit success or not.
2118 Frequency : only with huge count clients
2120 Description: Prevents from taking the superblock lock in llap_from_page for
2122 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2123 for llap_from_page call in ll_removepage prevents from taking the
2124 superblock lock for a soon died page.
2129 Description: Not check open intent error before release open handle
2130 Details : in some rare cases, the open intent error is not checked before
2131 release open handle, which may cause
2132 ASSERTION(open_req->rq_transno != 0), because it tries to release
2133 the failed open handle.
2138 Description: Set cat log bitmap only after create log success.
2139 Details : in some rare cases, the cat log bitmap is set too early. and it
2140 should be set only after create log success.
2144 Description: Accessing a block bevice can re-enable I/O when Lustre is
2145 tearing down a device.
2146 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2151 Description: service threads may hog cpus when there are a lot of requests
2152 Details : Insert cond_resched to give other threads a chance to use some CPU
2157 Description: the cat log was not initialized in recovery
2158 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2159 unlink log on mds will not be initialized until mds post
2160 recovery. And also in mds post recovery, the unlink log will
2161 initialization will be done asynchronausly, so there will be race
2162 between add unlink log and unlink log initialization.
2166 Description: brw_stats were being printed incorrectly
2167 Details : brw_stats were being printed as log2 but all of them were not
2168 recorded as log2. Also remove some code duplication arising from
2169 filter_tally_{read,write}.
2173 Frequency : rare, only in recovery.
2174 Description: ASSERTION(req->rq_type != LI_POISON) failed
2175 Details : imp_lock should be held while iterating over imp_sending_list for
2176 prevent destroy request after get timeout in ptlrpc_queue_wait.
2180 Description: replay-single.sh test 52 fails
2181 Details : A lock's skiplist need to be cleanup when it being unlinked
2182 from its resource list.
2186 Description: Short directio read returns full requested size rather than
2188 Details : Direct I/O operations should return actual amount of bytes
2189 transferred rather than requested size.
2191 Severity : enhancement
2193 Description: metadata RPC reduction (e.g. for rm performance)
2194 Details : decrease the amount of synchronous RPC between clients and servers
2195 by canceling conflicing lock before the operation on the client
2196 and packing thier handles into the main operation RPC to server.
2198 Severity : enhancement
2200 Description: Async OSC create to avoid the blocking unnecessarily.
2201 Details : If a OST has no remain object, system will block on the creating
2202 when need to create a new object on this OST. Now, ways use
2203 pre-created objects when available, instead of blocking on an
2204 empty osc while others are not empty. If we must block, we block
2205 for the shortest possible period of time.
2209 Description: improve handling recoverable errors
2210 Details : if request processig with error which can be recoverable on server
2211 request should be resend, otherwise page released from cache and
2214 Severity : enhancement
2216 Description: refine locking for avoid write wrong info into lov_objid file
2217 Details : fix possible races with add new target and write/update data in
2220 --------------------------------------------------------------------------------
2222 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2227 Frequency : on some architectures
2229 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2230 Details : Uninitialized flags sometimes cause configuration commands to
2234 Frequency : patchless clients only
2236 Description: missing __iget() symbol export
2237 Details : The __iget() symbol export is missing. To avoid the need for
2238 this on patchless clients the deathrow inode reaper is turned
2239 off, and we depend on the VM to clean up old inodes. This
2240 dependency was during via the fix for bug 12181.
2242 --------------------------------------------------------------------------------
2244 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2246 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2247 INTEROPERATE with older versions automatically. In many cases a
2248 special upgrade step is needed. Please read the
2249 user documentation before upgrading any part of a live system.
2250 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2251 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2252 * WARNING: Lustre configuration and startup changes are required with
2253 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2257 Severity : enhancement
2259 Description: MountConf
2260 Details : Lustre configuration is now managed via mkfs and mount
2261 commands instead of lmc and lconf. New obd types (MGS, MGC)
2262 are added for dynamic configuration management. See
2263 https://mail.clusterfs.com/wikis/lustre/MountConf for
2266 Severity : enhancement
2268 Description: dynamic OST addition
2269 Details : OSTs can now be added to a live filesystem
2271 Severity : enhancement
2273 Description: startup order invariance
2274 Details : MDTs and OSTs can be started in any order. Clients only
2275 require the MDT to complete startup.
2277 Severity : enhancement
2279 Description: parallel, asynchronous orphan cleanup
2280 Details : orphan cleanup is now performed in separate threads for each
2281 OST, allowing parallel non-blocking operation.
2283 Severity : enhancement
2285 Description: optimized stripe assignment
2286 Details : stripe assignments are now made based on ost space available,
2287 ost previous usage, and OSS previous usage, in order to try
2288 to optimize storage space and networking resources.
2290 Severity : enhancement
2292 Description: Permanently set tunables
2293 Details : All writable /proc/fs/lustre tunables can now be permanently
2294 set on a per-server basis, at mkfs time or on a live system.
2296 Severity : enhancement
2298 Description: Lustre message v2
2299 Details : Add lustre message format v2.
2301 Severity : enhancement
2303 Description: client OST exclusion list
2304 Details : Clients can be started with a list of OSTs that should be
2305 declared "inactive" for known non-responsive OSTs.
2309 Description: SPEC SFS validation failure on NFS v2 over lustre.
2310 Details : Changes the blocksize for regular files to be 2x RPC size,
2311 and not depend on stripe size.
2313 Severity : enhancement
2315 Description: Multiple MD RPCs in flight.
2316 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2317 To avoid overly-overloading MDS, introduce a limit on number of
2318 MDS RPCs in flight for a single client and add /proc controls
2319 to adjust this limit.
2321 Severity : enhancement
2323 Description: client read/write statistics
2324 Details : Add client read/write call usage stats for performance
2325 analysis of user processes.
2326 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2327 file access. extents_stats shows chunk size distribution.
2328 extents_stats_per_process show chunk size distribution per
2331 Severity : enhancement
2333 Description: mds statistics
2334 Details : Add detailed mds operations statistics in
2335 /proc/fs/lustre/mds/*/stats.
2339 Description: Failure of copying files with lustre special EAs.
2340 Details : Client side always return success for setxattr call for lustre
2341 special xattr (currently only "trusted.lov").
2346 Description: Refcount LNET uuids
2347 Details : The global LNET uuid list grew linearly with every startup;
2348 refcount repeated list entries instead of always adding to
2351 Severity : enhancement
2353 Description: Dynamic service threads
2354 Details : Within a small range, start extra service threads
2355 automatically when the request queue builds up.
2358 Frequency : mixed-endian client/server environments
2360 Description: mixed-endian crashes
2361 Details : The new msg_v2 system had some failures in mixed-endian
2364 Severity : enhancement
2366 Description: Easy OST removal
2367 Details : OSTs can be permanently deactivated with e.g. 'lctl
2368 conf_param lustre-OST0001.osc.active=0'
2370 Severity : enhancement
2372 Description: MGS proc entries
2373 Details : Added basic proc entries for the MGS showing what filesystems
2376 Severity : enhancement
2378 Description: provide MGS failover
2379 Details : Added config lock reacquisition after MGS server failover.
2381 Severity : enhancement
2383 Description: add Linux 2.4 support
2384 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2388 Description: a large application tries to do I/O to the same resource and dies
2389 in the middle of it.
2390 Details : Check the req->rq_arrival time after the call to
2391 ost_brw_lock_get(), but before we do anything about
2392 processing it & sending the BULK transfer request. This
2393 should help move old stale pending locks off the queue as
2394 quickly as obd_timeout.
2397 Frequency : when an incorrect nid is specified during startup
2399 Description: ptlrpc connect to non-existant node causes kernel crash
2400 Details : LNET can't be re-entered from an event callback, which
2401 happened when we expire a message after the export has been
2402 cleaned up. Instead, hand the zombie cleanup off to another
2405 Severity : enhancement
2407 Description: plain/inodebits lock performance improvement
2408 Details : Grouping plain/inodebits in granted list by their request modes
2409 and bits policy, thus improving the performance of search through
2413 Frequency : only if OST filesystem is corrupted
2415 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2416 Details : for a short time RPCs with bulk IO are in the replay list,
2417 but replay of bulk IOs is unimplemented. If the OST filesystem
2418 is corrupted due to disk cache incoherency and then replay is
2419 started it is possible to trip an assertion. Avoid putting
2420 committed RPCs into the replay list at all to avoid this issue.
2423 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2425 Description: System hang on startup
2426 Details : This bug allowed the liblustre (e.g. catamount) client to
2427 return to the app before handling all startup RPCs. This
2428 could leave the node unresponsive to lustre network traffic
2429 and manifested as a server ptllnd timeout.
2431 Severity : enhancement
2433 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2434 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2435 boolean to control whether to print peer debug info when a
2436 client's RPC times out.
2439 Frequency : only for kernels with patches from Lustre below 1.4.3
2441 Description: Remove old rdonly API
2442 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2445 Frequency : only for devices with external journals
2447 Description: Set external device read-only also
2448 Details : During a commanded failover stop, we set the disk device
2449 read-only while the server shuts down. We now also set any
2450 external journal device read-only at the same time.
2453 Frequency : when upgrading from 1.4 while trying to change parameters
2455 Description: The wrong (new) MDC name was used when setting parameters for
2456 upgraded MDT's. Also allows changing of OSC (and MDC)
2457 parameters if --writeconf is specified at tunefs upgrade time.
2460 Frequency : when setting specific ost indicies
2462 Description: QOS code breaks on skipped indicies
2463 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2464 created with --index need not be sequential.
2466 Severity : enhancement
2468 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2469 Details : The uninit_groups feature works in conjunction with the kernel
2470 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2471 up the pass1 processing of e2fsck. This is a read-only feature
2472 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2473 mount filesystems that have had this feature enabled.
2475 Severity : enhancement
2477 Description: Improve multi-block allocation algorithm to avoid fragmentation
2478 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2479 allocation locality and avoid filesystem fragmentation.
2481 ------------------------------------------------------------------------------
2483 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2485 * Support for kernels:
2486 2.4.21-47.0.1.EL (RHEL 3)
2487 2.6.5-7.283 (SLES 9)
2488 2.6.9-42.0.10.EL (RHEL 4)
2489 2.6.12.6 vanilla (kernel.org)
2490 2.6.16.27-0.9 (SLES 10)
2491 * Recommended e2fsprogs version: 1.39.cfs5
2493 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2497 Frequency : occasional, depends on client load and configuration
2498 Bugzilla : 12181, 12203
2499 Description: data loss for recently-modified files
2501 Details : In some cases it is possible that recently written or created
2502 files may not be written to disk in a timely manner (this should
2503 normally be within 30s unless client IO load is very high).
2504 The problem appears as zero-length files or files that are a
2505 multiple of 1MB in size after a client crash or client eviction
2506 that are missing data at the end of the file.
2508 This problem is more likely to be hit on clients where files are
2509 repeatedly created and unlinked in the same directory, clients
2510 have a large amount of RAM, have many CPUs, the filesystem has
2511 many OSTs, the clients are rebooted frequently, and/or the files
2512 are not accessed by other nodes after being written.
2514 The presence of the problem can be detected by looking at
2515 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2516 smaller than the second (nr_unused) then dirty files will not
2517 be flushed automatically to disk. "sync; sleep 10" should be
2518 run several times on the node before unmounting it to update
2519 Lustre (this is also safe to run on nodes without this problem).
2521 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2522 that can cause this same problem, so customers using that kernel
2523 also need to update the kernel in addition to Lustre. In order
2524 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2526 It is normal that files written just before a client crash (less
2527 than 30s) may not yet have been flushed to disk, even for local
2531 Frequency : frequent on thin XT3 nodes
2533 Description: UUID collision on thin XT3 Linux nodes
2534 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2535 randomly, since we relied on an insufficiently-seeded PRNG.
2540 Description: OSS hangs after "All ost request buffers busy"
2541 Details : A deadlock between quota and journal operations caused OSS
2542 hangs after printing "All ost request buffers busy."
2545 Frequency : always on liblustre builds
2547 Description: Cleanup compiler warnings on liblustre
2550 Frequency : always on liblustre builds on XT3
2552 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2556 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2557 > 32000 subdirectories
2558 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2559 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2560 subdirectory count crosses 32000. This will aid e2fsck to
2561 correctly handle more than 32000 subdirectories.
2564 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2566 Description: System hang on startup
2567 Details : This bug allowed the liblustre (e.g. catamount) client to
2568 return to the app before handling all startup RPCs. This
2569 could leave the node unresponsive to lustre network traffic
2570 and manifested as a server ptllnd timeout.
2572 Severity : enhancement
2574 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2575 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2576 boolean to control whether to print peer debug info when a
2577 client's RPC times out.
2582 Description: make O_SYNC working on 2.6 kernels
2583 Details : 2.6 kernels use different method for mark pages for write,
2584 so need add a code to lustre for O_SYNC work.
2589 Description: Failure to close file and release space on NFS
2590 Details : Put inode details into lock acquired in ll_intent_file_open.
2591 Use mdc_intent_lock in ll_intent_open to properly
2592 detect all kind of errors unhandled by mdc_enqueue.
2597 Description: proc file read during shutdown sometimes raced obd removal,
2599 Details : Add lock to prevent obd access after proc file removal.
2602 Frequency : Only for files larger than 4GB on 32-bit clients.
2604 Description: improperly doing page alignment of locks
2605 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2606 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2611 Description: RPCs being resent when they shouldn't be
2612 Details : Some RPCs that should not be resent are being resent. This
2613 can cause inconsistencies in the RPC state machine. Do not
2614 resend such requests.
2617 Frequency : rare, only with NFS export
2619 Description: Crash on NFS re-export node
2620 Details : under very unusual load conditions an assertion is hit in
2621 ll_intent_file_open()
2624 Frequency : only if OST filesystem is corrupted
2626 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2627 Details : for a short time RPCs with bulk IO are in the replay list,
2628 but replay of bulk IOs is unimplemented. If the OST filesystem
2629 is corrupted due to disk cache incoherency and then replay is
2630 started it is possible to trip an assertion. Avoid putting
2631 committed RPCs into the replay list at all to avoid this issue.
2636 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2637 Details : Large single O_DIRECT read and write calls can fail to allocate
2638 a sufficiently large buffer to process the request. In case of
2639 allocation failure the allocation is retried with a smaller
2640 buffer and broken into smaller requests.
2642 Severity : enhancement
2644 Description: Add -o localflock option to simulate old noflock behaviour.
2645 Details : This will achieve local-only flock/fcntl locks coherentness.
2650 Description: versioning check is incomplete
2651 Details : Checking the version difference of client vs. server, report
2652 error if the gap is too big.
2657 Description: add support PG_writeback bit
2658 Details : add support for PG_writeback bit for Lustre, for more carefull
2659 work with page cache in 2.6 kernel. This also fix some deadlocks
2660 and remove hack for work O_SYNC with 2.6 kernel.
2662 Severity : enhancement
2664 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2665 Details : The uninit_groups feature works in conjunction with the kernel
2666 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2667 up the pass1 processing of e2fsck. This is a read-only feature
2668 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2669 mount filesystems that have had this feature enabled.
2671 Severity : enhancement
2673 Description: Improve multi-block allocation algorithm to avoid fragmentation
2674 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2675 allocation locality and avoid filesystem fragmentation.
2677 ------------------------------------------------------------------------------
2679 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2681 * Support for kernels:
2682 2.6.9-42.0.3.EL (RHEL 4)
2683 2.6.5-7.276 (SLES 9)
2684 2.4.21-47.0.1.EL (RHEL 3)
2685 2.6.12.6 vanilla (kernel.org)
2686 2.6.16.21-0.8 (SLES10)
2687 * Recommended e2fsprogs version: 1.39.cfs2-0
2689 * The backwards-compatible /proc/sys/portals symlink has been removed
2690 in this release. Before upgrading, please ensure that you change
2691 any configuration scripts or /etc/sysctl.conf files that access
2692 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2693 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2694 in advance of the upgrade on any system running Lustre 1.4.6 or
2695 newer, since /proc/sys/lnet was added in that version.
2696 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2700 Frequency : only when quota is used
2702 Description: avoid scanning export list for quota master
2703 Details : Change the algorithms to avoid scanning export list in order
2704 to improve the efficiency.
2707 Frequency : MDS failover only, very rarely
2709 Description: "went back in time" messages on mds failover
2710 Details : The greatest transno may be lost when the current operation
2711 finishes with an error (transno==0) and the client's last_rcvd
2712 record is over-written. Save the greatest transno in the
2713 mds_last_transno for this case.
2716 Frequency : always for specific kernels and striping counts
2718 Description: client may get "Matching packet too big" without ACL support
2719 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2720 when trying to access files in certain configurations. The
2721 clients should in fact be denied when mounting because they do
2722 not understand ACLs.
2725 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2727 Description: many clients connecting with IO in progress causes connect timeouts
2728 Details : Avoid synchronous journal commits to avoid delays caused by many
2729 clients connecting/disconnecting when bulk IO is in progress.
2730 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2731 OST_IO_PORTAL to avoid delays behind potentially many pending
2735 Frequency : occasionally with multiple writers to a single file
2737 Description: shared writes to file may result in wrong size reported by stat()
2738 Details : Allow growing of kms when extent lock is cancelled
2741 Frequency : always with random mmap IO to multi-striped file
2743 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2744 Details : Only if the hole is at the end of OST object so that kms is too
2745 small. Fix is to increase kms accordingly in ll_nopage.
2748 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2750 Description: writes to a missing object would leak memory on the OST
2751 Details : If there is an inconsistency between the MDS and OST filesystems,
2752 such that the MDS references an object that doesn't exist, writes
2753 to that object will leak memory due to incorrect cleanup in the
2754 error handling path, eventually running out of memory on the OST.
2759 Description: Creating too long symlink causes lustre errors
2760 Details : Check symlink and name lengths before sending requests to MDS.
2763 Frequency : only if flock is enabled (not on by default)
2765 Description: posix locks not released on fd closure on 2.6.9+
2766 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2767 kernels, this caused such locks not to be released on fd close and
2768 then assertions on fs unmount about still used locks.
2771 Frequency : MDS failover only, very rarely
2773 Description: clients may get ASSERTION(granted_lock != NULL)
2774 Details : When request was taking a long time, and a client was resending
2775 a getattr by name lock request. The were multiple lock requests
2776 with the same client lock handle and
2777 mds_getattr_name->fixup_handle_for_resent_request found one of the
2778 lock handles but later failed with ASSERTION(granted_lock != NULL).
2783 Description: handle->h_buffer_credits > 0, assertion failure
2784 Details : h_buffer_credits is zero after truncate, causing assertion
2785 failure. This patch extends the transaction or creates a new
2789 Frequency : NFS re-export or patchless client
2790 Bugzilla : 11179, 10796
2791 Description: Crash on NFS re-export node (__d_move)
2792 Details : We do not want to hash the dentry if we don't have a lock.
2793 But if this dentry is later used in d_move, we'd hit uninitialised
2794 list head d_hash, so we just do this to init d_hash field but
2795 leave dentry unhashed.
2798 Frequency : NFS re-export or patchless client
2800 Description: NFS exports has problem with symbolic link
2801 Details : lustre client didn't properly install dentry when re-exported
2802 to NFS or running patchless client.
2805 Frequency : NFS re-export or patchless client
2807 Description: Various nfs/patchless fixes.
2808 Details : fixes reuse disconected alias for lookup process - this fixes
2809 warning "find_exported_dentry: npd != pd",
2810 fix permission error with open files at nfs.
2811 fix apply umask when do revalidate.
2814 Frequency : occasional
2816 Description: Crash on NFS re-export node
2817 Details : calling clear_page() on the wrong pointer triggered oops in
2818 generic_mapping_read().
2821 Frequency : rarely, using O_DIRECT IO
2823 Description: unaligned directio crashes client with LASSERT
2824 Details : check for unaligned buffers before trying any requests.
2827 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2829 Description: stale data returned from RAID cache
2830 Details : If only a small amount of IO is done to the RAID device before
2831 reading it again it is possible to get stale data from the RAID
2832 cache instead of reading it from disk.
2835 Frequency : always for sles10 kernel
2837 Description: sles10 support
2838 Details : ll_follow_link: compile fixes and using of nd_set_link
2839 under newer kernels.
2842 Frequency : depends on arch, kernel and compiler version, always on sles10
2845 Description: recursive or deep enough symlinks cause stack overflow
2846 Details : getting rid of large stack-allocated variable in
2850 Frequency : depends on hardware
2852 Description: lustre write performance loss in the SLES10 kernel
2853 Details : the performance loss is caused by using of write barriers in the
2854 ext3 code. The SLES10 kernel turns barrier support on by
2855 default. The fix is to undo that change for ldiskfs.
2857 ------------------------------------------------------------------------------
2859 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2861 * Support for kernels:
2862 2.6.9-42.0.3EL (RHEL 4)
2863 2.6.5-7.276 (SLES 9)
2864 2.4.21-47.0.1.EL (RHEL 3)
2865 2.6.12.6 vanilla (kernel.org)
2869 Frequency : quota enabled and large files being deleted
2871 Description: releasing more than 4GB of quota at once hangs OST
2872 Details : If a user deletes more than 4GB of files on a single OST it
2873 will cause the OST to spin in an infinite loop. Release
2874 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2879 Description: statfs data retrieved from /proc may be stale or zero
2880 Details : When reading per-device statfs data from /proc, in the
2881 {kbytes,files}_{total,free,avail} files, it may appear
2882 as zero or be out of date.
2885 Frequency : systems with MD RAID1 external journal devices
2887 Description: lconf's call to blkid is confused by RAID1 journal devices
2888 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2889 of returning all block devices that match the journal UUID.
2892 Frequency : always, for aggregate stripe size over 4GB
2894 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2895 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2896 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2899 Frequency : always if "lfs find" used on a local file/directory
2901 Description: "lfs find" segfaults if used on a local file/directory
2902 Details : The case where a directory component was not specified wasn't
2903 handled correctly. Handle this properly.
2906 Frequency : always on ppc64
2908 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2909 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2910 perform endianness conversion.
2913 Frequency : rarely (truncate to non-zero file size after write under load)
2914 Bugzilla : 10730, 10687
2915 Description: Files padded with zeros to next 4K multiple
2916 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2917 it is possible that files that are truncated to a non-zero size
2918 immediately after being written are filled with zero bytes beyond
2919 the truncated size. No file data is lost.
2921 Severity : enhancement
2923 Description: Allow recovery/failover for liblustre clients.
2924 Details : liblustre clients were unaware of failover configurations until
2927 Severity : enhancement
2929 Description: user file locks should fail when not mounting with flock option
2930 Details : Set up an error-returning stub in ll_file_operations.lock field
2931 to prevent incorrect behaviour when client is mounted without
2932 flock option. Also, set up properly f_op->flock field for
2936 Frequency : always on ia64
2938 Description: "lfs df" loops on printing out MDS statfs information
2939 Details : The obd_ioctl_data was not initialized and in some systems
2940 this caused a failure during the ioctl that did not return
2941 an error. Initialize the struct and return an error on failure.
2944 Frequency : SLES 9 only
2946 Description: Error of copying files with lustre special EAs as root
2947 Details : Client side always return success for setxattr call for lustre
2948 special xattr (currently only "trusted.lov").
2951 Frequency : rarely on clusters with both ia64+i386 clients
2953 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2954 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2955 clients are doing concurrent writes to the same file it is
2956 possible that the ia64 clients may LASSERT because the OST
2957 extent locks are not PAGE_SIZE aligned. Ensure that grown
2958 locks are always aligned on the request boundary.
2961 Frequency : specific use, occasional
2963 Description: Overwriting in use executable truncates on-disk binary image
2964 Details : If one node attempts to overwrite an executable in use by
2965 another node, we now correctly return ETXTBSY instead of
2966 truncating the file.
2968 Severity : enhancement
2970 Description: Async OSC create to avoid the blocking unnecessarily.
2971 Details : If a OST has no remain object, system will block on the creating
2972 when need to create a new object on this OST. Now, ways use
2973 pre-created objects when available, instead of blocking on an
2974 empty osc while others are not empty. If we must block, we block
2975 for the shortest possible period of time.
2980 Description: chmod on Lustre root is propagated to other clients
2981 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2982 invalid by the follow_mount time.
2985 Frequency : liblustre clients only
2987 Description: Race in 'instant cancel' lock handling could lead to such locks
2988 never to be granted in case of SMP MDS
2989 Details : Do not destroy not yet granted but cbpending locks in
2993 Frequency : replay/resend of open
2995 Description: non null lock assetion failure in mds_intent_policy
2996 Details : Trying to replay/resend lockless open requests resulted in
2997 mds_open() returning 0 with no lock. Now it sets a flag if
2998 a lock is going to be returned.
3000 Severity : enhancement
3002 Description: Checksum enhancements
3003 Details : New checksum enhancements allow for resending RPCs that failed
3006 Severity : enhancement
3008 Description: Tunables on number of dirty pages in cacche
3009 Details : Allow to set limit on number of dirty pages cached.
3014 Description: client crash on unmount - lock still has references
3015 Details : In some error handling cases it was possible to leak a lock
3016 reference on a client while accessing a file. This was not
3017 harmful to the client during operation, but would cause the
3018 client to crash when the filesystem is unmounted.
3021 Frequency : specific case, rare
3023 Description: ETXTBSY on mds though file not in use by client
3024 Details : ETXTBSY is no longer incorrectly returned when attempting to
3025 chmod or chown a directory that the user previously tried to
3026 execute or a currently-executing binary.
3029 Frequency : extremely rare except on liblustre-based clients
3031 Description: Lustre space not freed when files are deleted
3032 Details : Clean up open-unlinked files after client eviction. Previously
3033 the unlink was skipped and the files remained as orphans.
3038 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3039 Details : In some cases it was possible to send a blocking callback to
3040 a client doing a glimpse, even though that client didn't get
3041 a lock granted. When the glimpse lock is cancelled on the OST
3042 the freed lock is left on the waiting list and corrupted the list.
3045 Frequency : all core dumps
3047 Description: Broke core dumps to lustre
3048 Details : Negative dentry may be unhashed if parent does not have UPDATE
3049 lock, but some callers, e.g. do_coredump, expect dentry to be
3050 hashed after successful create, hash it in ll_create_it.
3052 ------------------------------------------------------------------------------
3054 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3056 * Support for kernels:
3057 2.6.9-42.0.2.EL (RHEL 4)
3058 2.6.5-7.276 (SLES 9)
3059 2.4.21-40.EL (RHEL 3)
3060 2.6.12.6 vanilla (kernel.org)
3064 Frequency : always on RHEL 3
3066 Description: Number of open files grows over time
3067 Details : The number of open files grows over time, whether or not
3068 Lustre is started. This was due to a filp leak introduced
3069 by one of our kernel patches.
3071 ------------------------------------------------------------------------------
3073 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3075 * Support for kernels:
3076 2.6.9-42.EL (RHEL 4)
3077 2.6.5-7.276 (SLES 9)
3078 2.4.21-40.EL (RHEL 3)
3079 2.6.12.6 vanilla (kernel.org)
3084 Bugzilla : 5719, 9635, 9792, 9684
3085 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3086 Details : If a server is under heavy load and cannot reply to new
3087 connection requests before the client resends the (re)connect,
3088 the connection handling code can behave badly if two service
3089 threads are concurrently handing separate (re)connections from
3090 the same client. Add better locking to the connection handling
3091 code, and ensure that only a single connection will be processed
3092 for a given client UUID, even if the lock is dropped.
3094 Severity : enhancement
3096 Description: add TCP zero-copy support to kernel
3097 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3098 sends if the hardware supports scatter-gather and checksumming.
3099 This allows socklnd to do client-write and server-read more
3100 efficiently and reduce CPU utilization from skbuf copying.
3103 Frequency : only if NFS exporting from client
3105 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3106 Details : If setting attributes on a file created under NFS that had
3107 never been opened it would be possible to oops the client
3108 if the file had no objects.
3111 Frequency : always for liblustre
3113 Description: liblustre client does MDS+OSTs setattr RPC for each write
3114 Details : When doing a write from a liblustre client, the client
3115 incorrectly issued an RPC to the MDS and each OST the file was
3116 striped over in order to update the timestamps. When writing
3117 with small chunks and many clients this could overwhelm the MDS
3118 with RPCs. In all cases it would slow down the write because
3119 these RPCs are unnecessary.
3121 Severity : enhancement
3123 Description: allow number of MDS service threads to be changed at module load
3124 Details : It is now possible to change the number of MDS service threads
3125 running. Adding "options mds mds_num_threads={N}" to the MDS's
3126 /etc/modprobe.conf will set the number of threads for the next
3127 time Lustre is restarted (assuming the "mds" module is also
3128 reloaded at that time). The default number of threads will
3129 stay the same, 32 for most systems.
3134 Description: OST crash if filesystem is unformatted or corrupt
3135 Details : If an OST is started on a device that has never been formatted
3136 or if the filesystem is corrupt and cannot even mount then the
3137 error handling cleanup routines would dereference a NULL pointer.
3142 Description: NULL pointer deref in llap_from_page.
3143 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3144 incorrect) mapping if the page was truncated/reclaimed while it was
3145 searched for. Check for this condition and skip such pages when
3146 doing readahead. Introduce extra check to llap_from_page() to
3147 verify page->mapping->host is non-NULL (so page is not anonymous).
3150 Frequency : Sometimes when using sys_sendfile
3152 Description: "page not covered by a lock" warnings from ll_readpage
3153 Details : sendfile called ll_readpage without right page locks present.
3154 Now we introduced ll_file_sendfile that does necessary locking
3155 around call to generic_file_sendfile() much like we do in
3159 Frequency : with certain MDS communication failures at client mount time
3161 Description: NULL pointer deref after failed client mount
3162 Details : a client connection request may delayed by the network layer
3163 and not be sent until after the PTLRPC layer has timed out the
3164 request. If the client fails the mount immediately it will try
3165 to clean up before the network times out the request. Add a
3166 reference from the request import to the obd device and delay
3167 the cleanup until the network drops the request.
3170 Frequency : occasionally during client (re)connect
3172 Description: assertion failure during client (re)connect
3173 Details : processing a client connection request may be delayed by the
3174 client or server longer than the client connect timeout. This
3175 causes the client to resend the connection request. If the
3176 original connection request is replied in this interval, the
3177 client may trip an assertion failure in ptlrpc_connect_interpret()
3178 which thought it would be the only running connect process.
3181 Frequency : only with obd_echo servers and clients that are rebooted
3183 Description: kernel BUG accessing uninitialized data structure
3184 Details : When running an obd_echo server it did not start the ping_evictor
3185 thread, and when a client was evicted an uninitialized data
3186 structure was accessed. Start the ping_evictor in the RPC
3187 service startup instead of the OBD startup.
3189 Severity : enhancement
3190 Bugzilla : 10193 (patchless)
3191 Description: Remove dependency on various unexported kernel interfaces.
3192 Details : No longer need reparent_to_init, exit_mm, exit_files,
3193 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3196 Frequency : rare (only users of deprecated and unsupported LDAP config)
3198 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3199 Details : LDAP apparently contains 'lustreName' attributes instead of
3200 'name'. A simple remapping of the name is sufficient.
3203 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3205 Description: waiting_locks_callback trips kernel BUG if client is evicted
3206 Details : Running with the dump_on_timeout debug flag turned on makes
3207 it possible that the waiting_locks_callback() can try to dump
3208 the Lustre kernel debug logs from an interrupt handler. Defer
3209 this log dumping to the expired_lock_main() thread.
3211 Severity : enhancement
3213 Description: Support NFS exporting on 2.6 kernels.
3214 Details : Implement non-rawops metadata methods for NFS server to use without
3215 changing NFS server code.
3218 Frequency : very rare (synthetic metadata workload only)
3220 Description: two racing renames might cause an MDS thread to deadlock
3221 Details : Running the "racer" program may cause one MDS thread to rename
3222 a file from being the source of a rename to being the target of
3223 a rename at exactly the same time that another thread is doing
3224 so, and the second thread has already enqueued these locks after
3225 doing a lookup of the target and is trying to relock them in
3226 order. Ensure that we don't try to re-lock the same resource.
3229 Frequency : only very large systems with liblustre clients
3231 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3232 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3233 all OSTs in parallel. This allows the eviction of stale liblustre
3234 clients to proceed much faster than if they were done in series,
3235 and also offers similar improvements for other set_info RPCs.
3240 Description: excessive CPU usage during initial read phase on client
3241 Details : During the initial read phase on a client, it would agressively
3242 retry readahead on the file, consuming too much CPU and impacting
3243 performance (since 1.4.5.8). Improve the readahead algorithm
3244 to avoid this, and also improve some other common cases (read
3245 of small files in particular, where "small" is files smaller than
3246 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3251 Description: MDS crash when receiving packet with unknown intent.
3252 Details : Do not LBUG in unknown intent case, just return -EFAULT
3254 Severity : enhancement
3255 Bugzilla : 9293, 9385
3256 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3257 Details : Do not serialize getattr (non-intent version) and statfs.
3260 Frequency : occasional, when OST network is overloaded/intermittent
3262 Description: client evicted by OST after bulk IO timeout
3263 Details : If a client sends a bulk IO request (read or write) the OST
3264 may evict the client if it is unresposive to its data GET/PUT
3265 request. This is incorrect if the network is overloaded (takes
3266 too long to transfer the RPC data) or dropped the OST GET/PUT
3267 request. There is no need to evict the client at all, since
3268 the pinger and/or lock callbacks will handle this, and the
3269 client can restart the bulk request.
3272 Frequency : Always when mmapping file with no objects
3274 Description: client crashes when mmapping file with no objects
3275 Details : Check that we actually have objects in a file before doing any
3276 operations on objects in ll_vm_open, ll_vm_close and
3282 Description: Request leak when working with deleted CWD
3283 Details : Introduce advanced request refcount tracking for requests
3284 referenced from lustre intent.
3286 Severity : Enhancement
3288 Description: Cache open file handles on client.
3289 Details : MDS now will return special lock along with openhandle, if
3290 requested and client is allowed to hold openhandle, even if unused,
3291 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3292 closing files for every read/write openration.
3294 Severity : Enhancement
3296 Description: Cache open negative dentries on client when possible.
3297 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3298 negative dentries on lock revocation.
3303 Description: Remounting a client read-only wasn't possible with a zconf mount
3304 Details : It wasn't possible to remount a client read-only with llmount.
3306 Severity : enhancement
3307 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3308 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3309 MPICH is included to add Lustre-specific ADIO interfaces.
3310 This is based closely on the UFS ADIO layer and only differs
3311 in file creation, in order to allow the OST striping to be set.
3312 This is user-contributed code and not supported by CFS.
3317 Description: extended inode attributes (immutable, append-only) work improperly
3318 when 2.4 and 2.6 kernels are used on client/server or vice versa
3319 Details : Introduce kernel-independent values for these flags.
3321 Severity : enhancement
3324 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3325 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3326 granularity. Now, Fractional MB granularity is be supported,
3327 this is very useful for low memory system.
3329 Severity : enhancement
3331 Description: Getattr by fid
3332 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3333 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3334 have locks and avoids getattr rpc requests for it on every stat.
3337 Frequency : Always, for filesystems larger than 2TB
3339 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3340 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3341 filesystems larger than 2TB to be mounted with mballoc enabled.
3344 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3346 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3347 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3348 filesystems larger than 2TB is mounted then the kernel will
3349 silently corrupt the start of the filesystem. CONFIG_LBD is
3350 enabled for all CFS-supported kernels, but the possibility of
3351 this happening with a modified kernel config exists.
3353 Severity : enhancement
3355 Description: add client O_DIRECT support for 2.6 kernels
3356 Details : It is now possible to do O_DIRECT reads and writes to files
3357 in the Lustre client mountpoint on 2.6 kernel clients.
3359 Severity : enhancement
3361 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3362 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3363 parallel, not waiting for response from every OST before sending
3364 a rpc to the next OST.
3369 Description: setattr vs write race when updating file timestamps
3370 Details : Client processes that update a file timestamp into the past
3371 right after writing to the file (e.g. tar) it is possible that
3372 the updated file modification time can be reset to the current
3373 time due to a race between processing the setattr and write RPC.
3375 Severity : enhancement
3377 Description: Bring 'lfs find' closer in line with regular Linux find.
3378 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3379 -print0 options and obtains all the needed info through the lustre
3382 Severity : enhancement
3384 Description: support up to 1024 configured devices on one node
3385 Details : change obd_dev array from statically allocated to dynamically
3386 allocated structs as they are first used to reduce memory usage
3391 Description: Flush dirty partially truncated pages during truncate
3392 Details : Immediatelly flush partially truncated pages in filter_setattr,
3393 this way we completely avoid having any pages in page cache on OST
3394 and can retire ugly workarounds during writes to flush such pages.
3399 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3400 Details : protect inode from truncation within vfs_unlink() context
3401 just take a reference before calling vfs_unlink() and release it
3402 when parent's i_sem is free.
3407 Description: last_id value checked outside lock on OST caused LASSERT failure
3408 Details : If there were multiple MDS->OST object precreate requests in
3409 flight, it was possible that the OST's last object id was checked
3410 outside a lock and incorrectly tripped an assertion. Move checks
3411 inside locks, and discard old precreate requests.
3414 Frequency : always, if extents are used on OSTs
3416 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3417 Details : OSTs using the extents format would not zero the high 16 bits of
3418 the index physical block number. This is not a problem for any
3419 OST filesystems smaller than 16TB, and no kernels support ext3
3420 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3421 new/modified files) and can be fixed for existing filesystems
3422 with e2fsprogs-1.39-cfs1.
3427 Description: import connection selection may be incorrect if timer wraps
3428 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3429 import connections to be ignored if the 32-bit jiffies counter
3430 wraps. Use a 64-bit jiffies counter.
3433 Frequency : very large clusters immediately after boot
3435 Description: LNET request buffers exhausted under heavy short-term load
3436 Details : If a large number of client requests are generated on a service
3437 that has previously never seen so many requests it is possible
3438 that the request buffer growth cannot keep up with the spike in
3439 demand. Instead of dropping incoming requests, they are held in
3440 the LND until the RPC service can accept more requests.
3443 Frequency : Sometimes during replay
3445 Description: Assertion failure in ll_local_open after replay.
3446 Details : If replay happened on an open request reply before we were able
3447 to set replay handler, reply will become not swabbed tripping the
3448 assertion in ll_local_open. Now we set the handler right after
3449 recognising of open request
3452 Frequency : very rare
3454 Description: kernel reports "badness in vsnprintf"
3455 Details : Reading from the "recovery_status" /proc file in small chunks
3456 may cause a negative length in lprocfs_obd_rd_recovery_status()
3457 call to vsnprintf() (which is otherwise harmless). Exit early
3458 if there is no more space in the output buffer.
3460 Severity : enhancement
3462 Description: clear OBD RPC statistics by writing to them
3463 Details : It is now possible to clear the OBD RPC statistics by writing
3464 to the "stats" file.
3469 Description: Client mtime is not the same on different clients after utimes
3470 Details : In some cases, the client was using the utimes() syscall on
3471 a file cached on another node. The clients now validate the
3472 ctime from the MDS + OSTs to determine which one is right.
3477 Description: Inability to activate failout mode
3478 Details : lconf script incorrectly assumed that in pythong string's numeric
3479 value is used in comparisons.
3482 Frequency : always with multiple stripes per file
3484 Description: Inefficient object allocation for mutli-stripe files
3485 Details : When selecting which OSTs to stripe files over, for files with
3486 a stripe count that divides evenly into the number of OSTs,
3487 the MDS is always picking the same starting OST for each file.
3488 Return the OST selection heuristic to the original design.
3493 Description: mount failures may take full timeout to return an error
3494 Details : Under some heavy load conditions it is possible that a
3495 failed mount can wait for the full obd_timeout interval,
3496 possibly several minutes, before reporting an error.
3497 Instead return an error as soon as the status is known.
3499 Frequency : quota enabled and large files being deleted
3501 Description: releasing more than 4GB of quota at once hangs OST
3502 Details : If a user deletes more than 4GB of files on a single OST it
3503 will cause the OST to spin in an infinite loop. Release
3504 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3509 Description: statfs data retrieved from /proc may be stale or zero
3510 Details : When reading per-device statfs data from /proc, in the
3511 {kbytes,files}_{total,free,avail} files, it may appear
3512 as zero or be out of date.
3515 Frequency : systems with MD RAID1 external journal devices
3517 Description: lconf's call to blkid is confused by RAID1 journal devices
3518 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3519 of returning all block devices that match the journal UUID.
3522 Frequency : always, for aggregate stripe size over 4GB
3524 Description: assertion fails when trying to use 4GB stripe size
3525 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3526 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3529 Frequency : always on ppc64
3531 Description: the first write on an ext3 filesystem with mballoc got stuck
3532 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3533 perform endianness conversion.
3535 ------------------------------------------------------------------------------
3537 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3539 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3540 INTEROPERATE with older versions automatically. Please read the
3541 user documentation before upgrading any part of a live system.
3542 * WARNING: Lustre networking configuration changes are required with
3543 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3546 * Support for kernels:
3547 2.6.9-22.0.2.EL (RHEL 4)
3548 2.6.5-7.244 (SLES 9)
3549 2.6.12.6 vanilla (kernel.org)
3552 Severity : enhancement
3553 Bugzilla : 7981/8208
3554 Description: Introduced Lustre Networking (LNET)
3555 Details : LNET is new networking infrastructure for Lustre, it includes
3556 a reorganized network configuration mode (see the user
3557 documentation for full details) as well as support for routing
3558 between different network fabrics. Lustre Networking Devices
3559 (LNDs) for the supported network fabrics have also been
3560 created for this new infrastructure.
3562 Severity : enhancement
3563 Description: Introduced Access control lists
3564 Details : clients can set ACLs on files and directories in order to have
3565 more fine-grained permissions than the standard Unix UGO+RWX.
3566 The MDS must be started with the "-o acl" mount option.
3568 Severity : enhancement
3569 Description: Introduced filesystem quotas
3570 Details : Administrators may now establish per-user quotas on the
3573 Severity : enhancement
3575 Description: Configuration change for the XT3
3576 The PTLLND is now used to run Lustre over Portals on the XT3
3577 The configure option(s) --with-cray-portals are no longer used.
3578 Rather --with-portals=<path-to-portals-includes> is used to
3579 enable building on the XT3. In addition to enable XT3 specific
3580 features the option --enable-cray-xt3 must be used.
3585 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3586 Details : A race between allocating a new llcd and re-getting the llcd_lock
3587 allowed another thread to grab newly-allocated llcd.
3589 Severity : enhancement
3591 Description: 2.6 OST async journal commit and locking fix to improve performance
3592 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3593 2.6 kernels are now async as they already were in 2.4 kernels so
3594 that they can commit concurrently with the network bulk transfer.
3595 For block-allocated files the filter allocation semaphore is held
3596 to avoid filesystem fragmentation during allocation. BKL lock
3597 removed for 2.6 xattr operations where it is no longer needed.
3602 Description: lconf incorrectly determined whether two IP networks could talk
3603 Details : In some more complicated routing and multiple-network
3604 configurations, lconf will avoid trying to make a network
3605 connection to a disjoint part of the IP space. It was doing the
3606 math incorrectly for one set of cases.
3611 Description: Fix for potential infinite loop processing records in an llog.
3612 Details : If an llog record is corrupted/zeroed, it is possible to loop
3613 forever in llog_process(). Validate the llog record length
3614 and skip the remainder of the block on error.
3617 Frequency : occasional (liblustre only)
3619 Description: liblustre could not open files whose last component is a symlink
3620 Details : sysio_path_walk() would incorrectly pass the open intent to
3621 intermediate path components.
3624 Frequency : rare (liblustre only with non-standard tuning)
3625 Bugzilla : 7201 (7350)
3626 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3627 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3628 after it has been released, only do this for referenced locks
3630 Severity : enhancement
3632 Description: specifying an (invalid) directory default stripe_size of -1
3633 would reset the directory default striping
3634 Details : stripe_size -1 was used internally to signal directory stripe
3635 removal, now use "all default" to signal dir stripe removal
3636 as a directory striping of "all default" is not useful
3639 Frequency : common for large clusters running liblustre clients
3641 Description: doing an ls when liblustre clients are running is slow
3642 Details : sending a glimpse AST to a liblustre client waits for every AST
3643 to time out, as liblustre clients will not respond. Since they
3644 cannot cache data we refresh the OST lock LVB from disk instead.
3646 Severity : enhancement
3648 Description: doing an ls at the same time as file IO can be slow
3649 Details : enqueue and other "small" requests can be blocked behind many
3650 large IO requests. Create a new OST IO portal for non-IO
3651 requests so they can be processed faster.
3654 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3656 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3657 created file with mode 000
3658 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3659 allow an ftruncate() as a non-root user to a file with mode 000.
3660 Lustre now respects this flag to disable mode checks when
3661 truncating a file owned by the user
3664 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3667 Description: Revoking locks from clients that went dead or catatonic might take
3669 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3670 cancellation of such locks instant on servers without waiting for
3671 any reply from clients. Clients drops these locks when cancel
3672 notification from server is received without replying.
3675 Frequency : liblustre-only, when liblustre client dies or becomes busy
3677 Description: Doing ls on Linux clients can take a long time with active
3679 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3680 granting such locks to it in the first place if possible. Locks
3681 are taken by proxy on the OST during the read or write and
3682 dropped immediately afterward. Add connect flags handling, do
3683 not grant locks to liblustre clients for glimpse ASTs.
3685 Severity : enhancement
3687 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3688 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3689 sequential reads and behaves badly with applications doing random
3690 reads. Improve it by reading ahead at least read region, and
3691 avoiding excessive large RPC for small reads.
3693 Severity : enhancement
3695 Description: Creating more than 1000 files for a single job may cause a load
3696 imbalance on the OSTs if there are also a large number of OSTs.
3697 Details : qos_prep_create() uses an OST index reseed value that is an
3698 even multiple of the number of available OSTs so that if the
3699 reseed happens in the middle of the object allocation it will
3700 still utilize the OSTs as uniformly as possible.
3705 Description: OST or MDS may oops in ping_evictor_main()
3706 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3707 but doesn't restart at beginning of obd_exports_timed list
3710 Severity : enhancement
3712 Description: improve by-nid export eviction on the MDS and OST
3713 Details : allow multiple exports with the same NID to be evicted at one
3714 time without re-searching the exports list.
3717 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3719 Description: MDS may oops in groups_free()
3720 Details : in rare race conditions a newly allocated group_info struct is
3721 freed again, and this can be NULL. The 2.4 compatibility code
3722 for groups_free() checked for a NULL pointer, but 2.6 did not.
3725 Frequency : common for liblustre clients doing little filesystem IO
3726 Bugzilla : 9352, 7313
3727 Description: server may evict liblustre clients accessing contended locks
3728 Details : if a client is granted a lock or receives a completion AST
3729 with a blocking AST already set it would not reply to the AST
3730 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3733 Frequency : lfs setstripe, only systems with more than 160 OSTs
3735 Description: unable to set striping with a starting offset beyond OST 160
3736 Details : llapi_create_file() incorrectly limited the starting stripe
3737 index to the maximum single-file stripe count.
3740 Frequency : LDAP users only
3742 Description: lconf did not handle in-kernel recovery with LDAP properly
3743 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3745 Severity : enhancement
3747 Description: bind OST threads to NUMA nodes to improve performance
3748 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3749 node and do their allocations there to localize memory access
3751 Severity : enhancement
3753 Description: llmount can determine client NID directly from Myrinet (GM)
3754 Details : the client NID code from gmnalnid was moved directly into
3755 llmount, removing the need to use this or specifying the
3756 client NID explicitly when mounting GM clients with zeroconf
3759 Frequency : if client is started with down MDS
3761 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3762 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3763 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3764 ptlrpc_import_delayed_req() because we didn't check if the
3765 request was interrupted, and we also didn't break out of the
3766 event loop if there was no timeout
3771 Description: data loss during non-page-aligned writes to a single file from
3772 both multiple nodes and multiple threads on one node at same time
3773 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3774 inconsistency led to false short-reads, that were cached and later
3775 used by ->prepare_write() to fill in partially written page,
3776 leading to data loss.
3779 Frequency : always, if lconf --abort_recovery used
3781 Description: lconf --abort_recovery fails with 'Operation not supported'
3782 Details : lconf was attempting to abort recovery on the MDT device and not
3785 Severity : enhancement
3787 Description: remove cleanup logs
3788 Details : replace lconf-generated cleanup logs with lustre internal
3789 cleanup routines. Eliminates the need for client-cleanup and
3792 Severity : enhancement
3794 Description: add support for EAs (user and system) on lustre filesystems
3795 Details : it is now possible to store extended attributes in the Lustre
3796 client filesystem, and with the user_xattr mount option it
3797 is possible to allow users to store EAs on their files also
3799 Severity : enhancement
3801 Description: Add possibility (config option) to show minimal available OST free
3803 Details : When compiled with --enable-mindf configure option, statfs(2)
3804 (and so, df) will return least minimal free space available from
3805 all OSTs as amount of free space on FS, instead of summary of
3806 free spaces of all OSTs.
3808 Severity : enhancement
3810 Description: do not expand extent locks acquired on OST-side
3811 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3812 by server: they are not cached anyway.
3815 Frequency : when mmap is used/binaries executed from Lustre
3817 Description: Unmmap pages before throwing them away from read cache.
3818 Details : llap_shrink cache now attempts to unmap pages before discarding
3819 them (if unmapping failed - do not discard). SLES9 kernel has
3820 extra checks that trigger if this unmapping is not done first.
3825 Description: lconf didn't resolve symlinks before checking to see whether a
3826 given mountpoint was already in use
3829 Frequency : when migrating failover services
3830 Bugzilla : 6395, 9514
3831 Description: When migrating a subset of services from a node (e.g. failback
3832 from a failover service node) the remaining services would
3833 time out and evict clients.
3834 Details : lconf --force (implied by --failover) sets the global obd_timeout
3835 to 5 seconds in order to quickly disconnect, but this caused
3836 other RPCs to time out too quickly. Do not change the global
3837 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3839 Severity : enhancement
3840 Frequency : if MDS is started with down OST
3841 Bugzilla : 9439,5706
3842 Description: Allow startup/shutdown of an MDS without depending on the
3843 availability of the OSTs.
3844 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3845 Add appropriate locking and lov-osc refcounts for safe
3846 cleaning. Add osc abort_inflight calls in case the
3847 synchronize never started.
3850 Frequency : occasional (Cray XT3 only)
3852 Description: root not authorized to access files in CRAY_PORTALS environment
3853 Details : The client process capabilities were not honoured on the MDS in
3854 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3855 been accessed by an authorized user then root was able to access
3856 the file on the local client also. The root user capabilities
3857 are now allowed on the MDS, as this environment has secure UID.
3860 Frequency : occasional
3862 Description: ldiskfs "too long searching" message happens too often
3863 Details : A debugging message (otherwise harmless) prints too often on
3864 the OST console. This has been reduced to only happen when
3865 there are fragmentation problems on the filesystem.
3870 Description: Division by zero in statfs when all OSCs are inactive
3871 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3872 lov_statfs divides by value returned by lov_get_stripecnt().
3876 Bugzilla : 9489, 3273
3877 Description: First write from each client to each OST was only 4kB in size,
3878 to initialize client writeback cache, which caused sub-optimal
3879 RPCs and poor layout on disk for the first writen file.
3880 Details : Clients now request an initial cache grant at (re)connect time
3881 and so that they can start streaming writes to the cache right
3882 away and always do full-sized RPCs if there is enough data.
3883 If the OST is rebooted the client also re-establishes its grant
3884 so that client cached writes will be honoured under the grant.
3889 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3890 Details : Now I/O RPCs go to different portal number and (presumably) fast
3891 lock requests (and glimses) and other RPCs get their own service
3892 threads pool that should be able to service those RPCs
3895 Severity : enhancement
3897 Description: Ability to exchange lustre version between client and servers and
3898 issue warnings at client side if client is too old. Also for
3899 liblustre clients there is ability to refuse connection of too old
3901 Details : New 'version' field is added to connect data structure that is
3902 filled with version info. That info is later checked by server and
3906 Frequency : rare, liblustre only.
3907 Bugzilla : 9296, 9581
3908 Description: Two simultaneous writes from liblustre at offset within same page
3909 might proceed at the same time overwriting eachother with stale
3911 Details : I/O lock withing llu_file_prwv was released too early, before data
3912 actually was hitting the wire. Extended lock-holding time until
3913 server acknowledges receiving data.
3916 Frequency : extremely rare. Never observed in practice.
3918 Description: avoid generating lustre_handle cookie of 0.
3919 Details : class_handle_hash() generates handle cookies by incrementing
3920 global counter, and can hit 0 occasionaly (this is unlikely, but
3921 not impossible, because initial value of cookie counter is
3922 selected randonly). Value of 0 is used as a sentinel meaning
3923 "unassigned handle" --- avoid it. Also coalesce two critical
3924 sections in this function into one.
3926 Severity : enhancement
3928 Description: allow liblustre clients to delegate truncate locking to OST
3929 Details : To avoid overhead of locking, liblustre client instructs OST to
3930 take extent lock in ost_punch() on client's behalf. New connection
3931 flag is added to handle backward compatibility.
3933 Severity : enhancement
3934 Bugzilla : 4928, 7341, 9758
3935 Description: allow number of OST service threads to be specified
3936 Details : a module parameter allows the number of OST service threads
3937 to be specified via "options ost ost_num_threads={N}" in the
3938 OSS's /etc/modules.conf or /etc/modprobe.conf.
3942 Bugzilla : 6146, 9635, 9895
3943 Description: servers crash with bad pointer in target_handle_connect()
3944 Details : In rare cases when a client is reconnecting it was possible that
3945 the connection request was the last reference for that export.
3946 We would temporarily drop the export reference and get a new
3947 one, but this may have been the last reference and the export
3948 was just destroyed. Get new reference before dropping old one.
3950 Severity : enhancement
3951 Frequency : if client is started with failover MDS
3953 Description: Allow multiple MDS hostnames in the mount command
3954 Details : Try to read the configuration from all specified MDS
3955 hostnames during a client mount in case the "primary"
3958 Severity : enhancement
3960 Description: Stop sending data to evicted clients as soon as possible.
3961 Details : Check if the client we are about to send or are sending data to
3962 was evicted already. (Check is done every second of waiting,
3963 for which l_wait_event interface was extended to allow checking
3964 of exit condition at specified intervals).
3967 Frequency : rare, normally only when NFS exporting is done from client
3969 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3971 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3974 Frequency : always, if srand() is called before liblustre initialization
3976 Description: Liblustre uses system PRNG disturbing its usage by user application
3977 Details : Introduce internal to lustre fast and high-quality PRNG for
3978 lustre usage and make liblustre and some other places in generic
3979 lustre code to use it.
3981 Severity : enhancement
3982 Bugzilla : 9477, 9557, 9870
3983 Description: Verify that the MDS configuration logs are updated when xml is
3984 Details : Check if the .xml configuration logs are newer than the config
3985 logs stored on the MDS and report an error if this is the case.
3986 Request --write-conf, or allow starting with --old_conf.
3988 Severity : enhancement
3990 Description: Handle symlinks in the path when checking if Lustre is mounted.
3991 Details : Resolve intermediate symlinks when checking if a client has
3992 mounted a filesystem to avoid duplicate client mounts.
3997 Description: lconf can hit an error exception but still return success.
3998 Details : The lconf command catches the Command error exception at the top
3999 level script context and will exit with the associated exit
4000 status, but doesn't ensure that this exit status is non-zero.
4005 Description: failure of ptlrpc thread startup can cause oops
4006 Details : Starting a ptlrpc service thread can fail if there are a large
4007 number of threads or the server memory is very fragmented.
4008 Handle this without oopsing.
4011 Frequency : always, only if liblustre and non-default acceptor port was used
4013 Description: liblustre cannot connect to servers with non-default acceptor port
4014 Details : tcpnal_set_default_params() was not called and was therefore
4015 ignoring the environment varaible TCPNAL_PORT, as well as other
4016 TCPNAL_ environment variables
4021 Description: two objects could be created on the same OST for a single file
4022 Details : If an OST is down, in some cases it was possible to create two
4023 objects on a single OST for a single file. No problems other
4024 than potential performance impact and spurious error messages.
4028 Bugzilla : 5681, 9562
4029 Description: Client may oops in ll_unhash_aliases
4030 Details : Client dcache may become inconsistent in race condition.
4031 In some cases "getcwd" can fail if the current directory is
4037 Description: Inode refcounting problems in NFS export code
4038 Details : link_raw functions used to call d_instantiate without obtaining
4039 extra inode reference first.
4043 Bugzilla : 9942, 9903
4044 Description: Referencing freed requests leading to crash, memleaks with NFS.
4045 Details : We used to require that call to ll_revalidate_it was always
4046 followed by ll_lookup_it. Also with revalidate_special() it is
4047 possible to call ll_revalidate_it() twice for the same dentry
4048 even if first occurence returned success. This fix changes semantic
4049 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4050 reference on a request referred from the intent.
4051 ll_intent_release() then releases such a request.
4054 Frequency : rare, normally benchmark loads only
4056 Description: unlinked inodes were kept in memory on the client
4057 Details : If a client is repeatedly creating and unlinking files it
4058 can accumulate a lot of stale inodes in the inode slab cache.
4059 If there is no other client load running this can cause the
4060 client node to run out of memory. Instead flush old inodes
4061 from client cache that have the same inode number as a new inode.
4064 Frequency : SLES9 2.6.5 kernel and long filenames only
4065 Bugzilla : 9969, 10379
4066 Description: utime reports stale NFS file handle
4067 Details : SLES9 uses out-of-dentry names in some cases, which confused
4068 the lustre dentry revalidation. Change it to always use the
4072 Frequency : rare, unless heavy write-truncate concurrency is continuous
4073 Bugzilla : 4180, 6984, 7171, 9963, 9331
4074 Description: OST becomes very slow and/or deadlocked during object unlink
4075 Details : filter_destroy() was holding onto the parent directory lock
4076 while truncating+unlinking objects. For very large objects this
4077 may block other threads for a long time and slow overall OST
4078 responsiveness. It may also be possible to get a lock ordering
4079 deadlock in this case, or run out of journal credits because of
4080 the combined truncate+unlink. Solution is to do object truncate
4081 first in one transaction without parent lock, and then do the
4082 final unlink in a new transaction with the parent lock. This
4083 reduces the lock hold time dramatically.
4086 Frequency : rare, 2.4 kernels only
4088 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4089 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4090 before unmounting their filesystem in order to avoid deadlock.
4091 The kernel_locked() function in 2.4 kernels only checks whether
4092 the kernel lock is held, not whether it is this process that is
4093 holding it as 2.6 kernels do.
4098 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4099 Details : The client ptlrpc code may be trying to reconnect to a down
4100 server before a previous connection attempt has timed out.
4101 Increase the reconnect interval to be longer than the connection
4102 timeout interval to avoid sending duplicate connections to
4106 Frequency : echo_client brw_test command
4108 Description: fix echo_client to work with OST preallocated code
4109 Details : OST preallocation code (5137) didn't take echo_client IO path
4110 into account: echo_client calls filter methods outside of any
4111 OST thread and, hence, there is no per-thread preallocated
4112 pages and buffers to use. Solution: hijack pga pages for IO. As
4113 a byproduct, this avoids unnecessary data copying.
4117 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4118 Description: Client can oops in mdc_commit_close() after open replay
4119 Details : It was possible for the MDS to return an open request with no
4120 transaction number in mds_finish_transno() if the client was
4121 evicted, but without actually returning an error. Clients
4122 would later try to replay that open and may trip an assertion
4123 Simplify the client close codepath, and always return an error
4124 from the MDS in case the open is not successful.
4127 Frequency : rare, 2.6 OSTs only
4129 Description: OST may deadlock under high load on fragmented files
4130 Details : If there was a heavy load and highly-fragmented OST filesystems
4131 it was possible to have all the OST threads deadlock waiting on
4132 allocation of biovecs, because the biovecs were not released
4133 until the entire RPC IO was completed. Instead, release biovecs
4134 as soon as they are complete to ensure forward IO progress.
4136 Severity : enhancement
4138 Description: Support for specifying external journal device at mount
4139 Details : If an OST or MDS device is formatted with an external journal
4140 device, this device major/minor is stored in the ext3 superblock
4141 and may not be valid for failover. Allow detecting and
4142 specifying the external journal at mount time.
4147 Description: Mounting an MDS with pending unlinked files may cause oops
4148 Details : target_finish_recovery() calls mds_postrecov() which returned
4149 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4150 considers this an error and immediately begins cleaning up the
4151 lov, just after starting the mds_lov process
4153 Severity : enhancement
4155 Description: Implement 'lfs df' to report actual free space on per-OST basis
4156 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4157 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4158 usage of objects; '-h' to report in human readable format.
4160 ------------------------------------------------------------------------------
4162 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4169 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4170 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4171 at startup during orphan unlinking. Instead, skip buddy update
4172 before setup, as it will be regenerated after recovery is complete.
4177 Description: If an OST is inactive, its locks might reference stale inodes.
4178 Details : lov_change_cbdata() must iterate over all namespaces, even if
4179 they are inactive to clear inode references from the lock.
4181 Severity : enhancement
4182 Frequency : occasional, if non-standard max_dirty_mb used
4184 Description: Client will block write RPCs if not enough grant
4185 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4186 then the client will block writes while waiting for another RPC
4187 to complete instead of consuming its dirty limit. With change
4188 we get improved performance when max_dirty_mb is small.
4190 Severity : enhancement
4191 Bugzilla : 3389, 6253
4192 Description: Add support for supplementary groups on the MDS.
4193 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4194 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4195 lookups for user supplementary groups into a cache.
4199 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4200 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4201 for liblustre O_EXCL is enforced by MDS.
4205 Description: readdir never returns NULL in liblustre.
4206 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4207 dentry in struct dirent.
4211 Frequency : liblustre only, depends on application IO pattern
4212 Description: liblustre clients evicted if not contacting servers
4213 Details : Don't put liblustre clients into the ping_evictor list, so
4214 they will not be evicted by the pinger ever.
4216 Severity : enhancement
4218 Description: Add ability to evict clients by NID from MDS.
4219 Details : By echoing "nid:$NID" string into
4220 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4221 $NID would be instantly evicted from this MDS and from all active
4222 OSTs connected to it.
4226 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4227 Details : lookup_it_finish() used to query file size from OSTs that was not
4232 Description: service threads change working directory to that of init
4233 Details : Starting lustre service threads may pin the working directory
4234 of the parent thread, making that filesystem busy. Threads
4235 now change to the working directory of init to avoid this.
4239 Frequency : during shutdown only
4240 Description: shutdown with a failed MDS or OST can cause unmount to hang
4241 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4246 Frequency : 2.6 only
4247 Description: chmod/chown may include an extra supplementary group
4248 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4249 supplementary group and if none is specified this is used.
4252 Bugzilla : 5479 (6816)
4254 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4255 Details : If lookup is in progress on a file that is unlinked we might try
4256 to revalidate the inode and fail in revalidate after lookup is
4257 complete and ll_file_open() enqueues the open again but
4258 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4261 Frequency : always, if lconf --abort_recovery used
4263 Description: lconf --abort_recovery fails with 'Operation not supported'
4264 Details : lconf was attempting to abort recovery on the MDT device and not
4267 ------------------------------------------------------------------------------
4269 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4274 Frequency : rare (only unsupported configurations with a node running as an
4276 Bugzilla : 6514, 5137
4277 Description: Mounting a Lustre file system on a node running as an OST could
4279 Details : OSTs now preallocates memory needed to write out data at
4280 startup, instead of when needed, to avoid having to
4281 allocate memory in possibly low memory situations.
4282 Specifically, if the file system is mounted on on OST,
4283 memory pressure could force it to try to write out data,
4284 which it needed to allocate memory to do. Due to the low
4285 memory, it would be unable to do so and the node would
4286 become unresponsive.
4288 Severity : enhancement
4290 Description: Addition of lconf --service command line option
4291 Details : lconf now accepts a '--service <arg>' option, which is
4292 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4294 Severity : enhancement
4296 Description: Failover mode is now the default for OSTs.
4297 Details : By default, OSTs will now run in failover mode. To return to
4298 the old behaviour, add '--failout' to the lmc line for OSTs.
4300 Severity : enhancement
4302 Description: Health checks are now provided for MDS and OSTs
4303 Details : Additional detailed health check information on MSD and OSTs
4304 is now provided through the procfs health_check value.
4307 Frequency : occasional, depends on IO load
4309 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4310 after numerous create/delete cycles
4311 Details : The ext3 inode allocation policy would not allocate new inodes
4312 very well on the OSTs because there are no new directories
4313 being created. Instead we look for groups with free space if
4314 the parent directories are nearly full.
4319 Description: Network or server problems during mount may cause partially
4320 mounted clients instead of returning an error.
4321 Details : The config llog parsing code may overwrite the error return
4322 code during mount error handling, returning success instead
4328 Description: MDS can fail to allocate large reply buffers
4329 Details : After long uptimes the MDS can fail to allocate large reply
4330 buffers (e.g. zconf client mount config records) due to memory
4331 fragmentation or consumption by the buffer cache. Preallocate
4332 some large reply buffers so that these replies can be sent even
4333 under memory pressure.
4337 Frequency : rare (liblustre)
4338 Description: fsx running with liblustre complained that using truncate() to
4339 extend the file doesn't work. This patch corrects that issue.
4340 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4341 ATTR_SIZE and lsm use in llu_setattr_raw.
4345 Frequency : rare, only 2.6 kernels
4346 Description: Unusual file access patterns on the MDS may result in inode
4347 data being lost in very rare circumstances.
4348 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4349 code in the 2.6 kernel caused the inode and/or EA data not to
4350 be read from disk, causing single-file corruption.
4354 Frequency : rare, only 2.6 filesystems using extents
4355 Description: Heavy concurrent write and delete load may cause data corruption.
4356 Details : It was possible under high-load situations to have an extent
4357 metadata block in the block device cache from a just-unlinked
4358 file overwrite a newly-allocated data block. We now unmap any
4359 metadata buffers that alias just-allocated data blocks.
4363 Frequency : filesystems with default stripe_count larger than 77
4364 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4365 Details : lconf specifies an inode size of 4096 bytes when the default
4366 stripe_count is larger than 77. This conflicts with the default
4367 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4368 this case to avoid pinning too much memory for large EAs.
4370 ------------------------------------------------------------------------------
4372 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4377 Frequency : rare (extremely heavy IO load with hundreds of clients)
4379 Description: Client is evicted, gets IO error writing to file
4380 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4381 caused clients to be evicted for AST timeouts. The fixes in
4382 bug 5192 mean we no longer need to have such short AST timeouts
4383 so ldlm_timeout has been increased.
4386 Frequency : occasional during --force or --failover shutdown under load
4387 Bugzilla : 5949, 4834
4388 Description: Server oops/LBUG if stopped with --force or --failover under load
4389 Details : a collection of import/export refcount and cleanup ordering
4390 issues fixed for safer force cleanup
4393 Frequency : only filesystems larger than 120 OSTs
4394 Bugzilla : 5990, 6223
4395 Description: lfs getstripe would oops on a very large filesystem
4396 Details : lov_getconfig used kfree on vmalloc'd memory
4399 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4400 Bugzilla : 6242, 6243
4401 Description: reading from files that had been truncated to a non-zero size
4402 but never opened returned no data
4403 Details : ll_file_read() reads zeros from no-object files to EOF
4408 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4409 Details : There's little harm in aborting MDS/OSS recovery and letting it
4410 try again, so I removed the LASSERT and return an error instead.
4412 Severity : enhancement
4414 Description: New debugging infrastructure for tracking down data corruption
4415 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4416 (b) cover more of the client-side code path, and (c) try to narrow
4417 down where problems occurred
4421 Bugzilla : 3819, 4364, 4397, 6313
4422 Description: Racing close and eviction MDS could cause assertion in mds_close
4423 Details : It was possible to get multiple mfd references during close and
4424 client eviction, leading to one thread referencing a freed mfd.
4426 Severity: : enhancement
4427 Bugzilla : 3262, 6359
4428 Description: Attempts to reconnect to servers are now more aggressive.
4429 Details : This builds on the enhanced upcall-less recovery that was added
4430 in 1.4.2. When trying to reconnect to servers, clients will
4431 now try each server in the failover group every 10 seconds. By
4432 default, clients would previously try one server every 25 seconds.
4437 Description: After recovery, certain operations trigger a failed
4438 assertion on a client.
4439 Details : Failing over an mds, using lconf -d --failover, while a
4440 client was doing a readdir() call would cause the client to
4441 LBUG after recovery completed and the readdir() was resent.
4443 Severity : enhancement
4445 Description: Default groups are now added by lconf
4446 Details : You can now run lconf --group <servicename> without having to
4447 manually add groups with lmc.
4450 Frequency : occasional
4452 Description: Nodes with an elan id of 0 trigger a failed assertion
4455 Frequency : always when accessing e.g. tty/console device nodes
4457 Description: tty and some other devices nodes cannot be used on lustre
4458 Details : file's private_data field is used by device data and lustre
4459 values in there got lost. New field was added to struct file to
4460 store fs-specific private data.
4463 Frequency : when exporting Lustre via NFS
4465 Description: NFSD failed occasionally when looking up a path component
4466 Details : NFSD is looking up ".." which was broken in ext3 directories
4467 that had grown large enough to become hashed.
4470 Frequency : Clusters with multiple interfaces not on the same subnet
4472 Description: Nodes will repeatedly try to reconnect to an interface which it
4473 cannot reach and report an error to the log.
4474 Details : Extra peer list entries will be created by lconf with some peers
4475 unreachable. lconf now validates the peer before adding it.
4478 Frequency : Only if a default stripe is set on the filesystem root.
4480 Description: Setting a default stripe on the filesystem root prevented the
4481 filesystem from being remounted.
4482 Details : The client was sending extra request flags in the root getattr
4483 request and did not allocate a reply buffer for the dir EA.
4486 Frequency : occasional, higher if lots of files are accessed by one client
4487 Bugzilla : 6159, 6097
4488 Description: Client trips assertion regarding lsm mismatch/magic
4489 Details : While revalidating inodes the VFS looks up inodes with ifind()
4490 and in rare cases can find an inode that is being freed.
4491 The ll_test_inode() code will free the lsm during ifind()
4492 when it finds an existing inode and then the VFS later attaches
4493 this free lsm to a new inode.
4497 Bugzilla : 6422, 7030
4498 Description: MDS deadlock between mkdir and client eviction
4499 Details : Creating a new file via mkdir or mknod (starting a transaction
4500 and getting the ns lock) can deadlock with client eviction
4501 (gets ns lock and trying to finish a synchronous transaction).
4504 Frequency : occasional
4505 Description: While starting a server, the fsfilt_ext3 module could not be
4507 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4508 kernels. Previously, lconf would still use the ext3 name
4509 when trying to load modules. Now, it will correctly use
4510 ext3 on 2.4 and ldiskfs on 2.6.
4512 Severity : enhancement
4513 Description: The default stripe count has been changed to 1
4514 Details : The interpretation of the default stripe count (0, to lfs
4515 or lmc) has been changed to mean striping across a single
4516 OST, rather than all available. For general usage we have
4517 found a stripe count of 1 or 2 works best.
4519 Severity : enhancement
4520 Description: Add support for compiling against Cray portals.
4521 Details : Conditional compiling for some areas that are different
4525 Frequency : occasional
4526 Bugzilla : 6409, 6834
4527 Description: Creating files with an explicit stripe count may lead to
4528 a failed assertion on the MDS
4529 Details : If some OSTs are full or unavailable, creating files may
4530 trigger a failed assertion on the MDS. Now, Lustre will
4531 try to use other servers or return an error to the
4535 Frequency : occasional
4537 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4538 caused lock timeout during readahead (since 1.4.2).
4539 Details : Processes doing readahead might match a lock that hasn't been
4540 granted yet if there are overlapping and conflicting lock
4541 requests. The readahead process waits on ungranted lock
4542 (original lock is CBPENDING), while OST waits for that process
4543 to cancel CBPENDING read lock and eventually evicts client.
4545 Severity : enhancement
4547 Description: Initial enabling of flock support for clients
4548 Details : Implements fcntl advisory locking and file status functions.
4549 This feature is provided as an optional mount flag (default
4550 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4551 locking are implemented yet, and those that are are not guaranteed
4552 to be completely correct in production environments.
4553 mount -t lustre -o [flock|noflock] ...
4556 Frequency : occasional
4558 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4559 assertion in the ext3 JBD (journaling) layer.
4560 Details : The b_committed_data struct is protected by the big kernel lock
4561 in 2.4 kernels, serializing journal_commit_transaction() and
4562 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4563 access to this struct. In 2.6 kernels there is finer grained
4564 locking to improve SMP performance of the JBD layer.
4568 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4573 Description: merge and cleanup kernel patches.
4576 -----------------------------------------------------------------------------
4578 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4580 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4581 versions of Lustre. Please update all servers and clients to
4582 version 1.4.2 or later at the same time. You must also run
4583 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4584 to update the configuration logs.
4586 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4587 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4588 - swab llog records of type '0' so we get proper header size/idx (5861)
4589 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4590 - fix rename of one directory over another leaking an inode (5953)
4591 - avoid SetPageDirty on 2.6 (5981)
4592 - don't re-add just-being-destroyed locks to the waiting list (5653)
4593 - when creating new directories, inherit the parent's custom
4594 striping settings if present parent (3048)
4595 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4596 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4597 - don't hold client locks on temporary worklist from l_lru (5666)
4598 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4599 - automatically evict dead clients (5921)
4600 - Update file size properly in create+truncate+fstat case (6196)
4601 - Do not unhash mountpoint dentries, do not allow removal of
4603 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4604 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4605 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4606 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4607 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4608 - Correctly handle reads of files with no objects (6243)
4609 - lctl recover will also mark a device active if deactivate used (5933)
4611 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4612 - allow --write-conf on an MDS with different nettype than client (5619)
4613 - don't write config llogs to MDS for mounts not from that MDS (5617)
4614 - lconf should create multiple TCP connections from a client (5201)
4615 - init scripts are now turned off by default; run chkconfig --on
4616 lustre and chkconfig --on lustrefs to use them
4617 - upcalls are no longer needed for clients to recover to failover
4619 - add --abort-recovery option to lconf to abort recovery on device
4621 - add support for an arbitrary number of OSTs (3026)
4622 - Quota support protocol changes.
4623 - forward compatibility changes to wire structs (6007)
4624 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4625 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4626 - improved "lustre status" script
4627 - initialize blocksize for non-regular files (6062)
4628 - added --disable-server and --disable-client configure options (5782)
4629 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4630 - Vanilla 2.4.29 support
4631 - increase maximum number of obd devices to 520 (6242)
4632 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4633 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4634 SLES 9 2.6.5 kernels
4635 - limit stripes per file to 160 (the maximum EA size) (6093)
4637 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4640 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4641 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4642 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4643 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4644 - small ext3 extents cleanups and fixes (5733)
4645 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4646 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4647 - avoid needless client->OST connect, fix handle mismatch (5317)
4648 - fix DLM error path that led to out-of-sync client, long delays (5779)
4649 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4650 - fix several locking issues related to i_size (5492,5624,5654,5672)
4651 - don't move pending lock onto export if it is already evicted (5683)
4652 - fix kernel oops when creating .foo in unlinked directory (5548)
4653 - fix deadlock in obdfilter statistics vs. object create (5811)
4654 - use time_{before,after} to avoid timer jiffies wrap (5882)
4655 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4656 - Do not leak request if server was not able to process it (5154)
4657 - If mds_open unable to find parent dir, make that negative lookup(5154)
4658 - don't create new directories with extent-mapping (5909, 5936)
4660 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4661 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4662 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4663 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4664 - support for RHEL4 kernel on i686 (5773)
4665 - provide error messages when incompatible logs are encountered (5898)
4667 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4668 * version 1.4.0.10 (1.4.1 release candidate 1)
4670 - don't keep a lock reference when lock is not granted (4238)
4671 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4672 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4673 - drop import inflight refcount on signal_completed_replay error (5255)
4674 - unlock page after async write error during send (3677)
4675 - handle missing objects in filter_preprw_read properly (5265)
4676 - no transno return for symlink open, don't save no-trasno open (3440)
4677 - don't try to complete elan receive that already failed (4012)
4678 - free RPC server reply state on error (5406)
4679 - clean up thread from ptlrpc_start_thread() on error (5160)
4680 - readahead could read extra page into cache that wasn't ejected (5388)
4681 - prevent races in class_attach/setup/cleanup/detach (5260)
4682 - don't dereference de->d_inode after l_dput of de (5458)
4683 - use "int" for stripe value returned from lock_to_stripe (5544)
4684 - mballoc allocation and error-checking fixes in 2.6 (5504)
4685 - block device patches to fix I/O request sizes in 2.6 (5482)
4686 - look up hostnames for IB nals (5602)
4687 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4688 - don't start multiple acceptors for the same port (5277)
4689 - fix incorrect LASSERT in mds_getattr_name (5635)
4690 - export a proc file for general "ping" checking (5628)
4691 - fix "lfs check" to not block when the MDS is down (5628)
4693 - service request history (4965)
4694 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4695 - create an "evict_client" /proc entry on OSTs, like the MDS has
4696 - fix mount usage message, return errors per mount(8) (5168)
4697 - change grep [] to grep "[]" in tests so they work in more UMLs
4698 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4699 - remove ancient LOV_MAGIC_V0 EA support (5047)
4700 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4701 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4702 - allow readahead allocations to fail when low on memory (5383)
4703 - mmap locking landed again, after considerable improvement (2828)
4704 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4706 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4709 - send OST transaction number in read/write reply to free req (4966)
4710 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4711 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4712 - the watchdog thread now runs as interruptible (5246)
4713 - flock/lockf fixes (but it's still disabled, pending 5135)
4714 - don't use EXT3 constants in llite code (5094)
4715 - memory shortage at startup could cause assertion (5176)
4717 - reorganization of lov code
4718 - single portals codebase
4720 - add extents/mballoc support (5025)
4721 - direct I/O reads in the obdfilter (4048)
4722 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4724 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4727 - send OST transaction number in read/write reply to free req (4966)
4728 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4729 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4730 - the watchdog thread now runs as interruptible (5246)
4731 - handle missing objects in filter_preprw_read properly (5265)
4732 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4733 - drop import inflight refcount on signal_completed_replay error (5255)
4734 - unlock page after async write error during send (3677)
4735 - return original error code on reconstructed replies (3761)
4736 - no transno return for symlink open, don't save no-trasno open (3440)
4738 - add pid to ldlm debugging output (4922)
4739 - bump the watchdog timeouts -- we can't handle 30sec yet
4740 - extra debugging for orphan dentry/inode bug (5259)
4742 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4745 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4746 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4747 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4748 - fix NULL dereference in /proc/sys/portals/routes (4827)
4749 - allow failed mdc_close() operations to be interrupted (4561)
4750 - stop precreate on OST before MDS would time out on it (4778)
4751 - don't send partial-page writes before EOF from client (4410)
4752 - discard client grant for sub-page writes on large-page clients (4520)
4753 - don't free dentries not owned by NFS code, check generation (4806)
4754 - fix lsm leak if mds_create_objects() fails (4801)
4755 - limit debug_daemon file size, always print CERROR messages (4789)
4756 - use transno after validating reply (3892)
4757 - process timed out requests if import state changes (3754)
4758 - update mtime on OST during writes, return in glimpse (4829)
4759 - add mkfsoptions to LDAP (4679)
4760 - use ->max_readahead method instead of zapping global ra (5039)
4761 - don't interrupt __l_wait_event() during strace
4763 - add software watchdogs to catch hung threads quickly (4941)
4764 - make lustrefs init script start after nfs is mounted
4765 - fix CWARN/ERROR duplication (4930)
4766 - return async write errors to application if possible (2248)
4767 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4768 - print NAL number in %x format (4645)
4769 - update barely-supported suse-2.4.21-171 series (4842)
4770 - support for sles 9 %post scripts
4771 - support for building 2.6 kernel-source packages
4772 - support for sles km_* packages
4774 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4777 - ignore -ENOENT errors in osc_destroy (3639)
4778 - notify osc create thread that OSC is being cleaned up (4600)
4779 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4780 - reconstruct ost_handle() like mds_handle() (4657)
4781 - create a new thread to do import eviction to avoid deadlock (3969)
4782 - let lconf resolve symlinked-to devices (4629)
4783 - don't unlink "objects" from directory with default EA (4554)
4784 - hold socknal file ref over connect in case target is down (4394)
4785 - allow more than 32000 subdirectories in a single directory (3244)
4786 - fix blocks count for O_DIRECT writes (3751)
4787 - OST returns ENOSPC from object create when no space left (4539)
4788 - don't send truncate RPC if file size isn't changing (4410)
4789 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4790 - bind to privileged port in socknal and tcpnal (3689)
4792 - rate limit CERROR/CWARN console message to avoid overload (4519)
4793 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4794 - basic mmap support (3918)
4795 - kernel patch series update from b1_4 (4711)
4797 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4800 - avoid crash during MDS cleanup with OST shut down (2775)
4801 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4802 - don't use bad inodes on the MDS (3744)
4803 - dynamic object preallocation to improve recovery speed (4236)
4804 - don't hold spinlock over lock dumping or change debug flags (4401)
4805 - don't zero obd_dev when it is force cleaned (3651)
4806 - print grants to console if they go negative (4431)
4807 - "lctl deactivate" will stop automatic recovery attempts (3406)
4808 - look for existing locks in ldlm_handle_enqueue() (3764)
4809 - don't resolve lock handle twice in recovery avoiding race (4401)
4810 - revalidate should check working dir is a directory (4134)
4812 - don't always mark "slow" obdfilter messages as errors (4418)
4814 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4817 - don't close LustreDB during write_conf until it is done (3860)
4818 - fix typo in lconf for_each_profile (3821)
4819 - allow dumping logs from multiple threads at one time (3820)
4820 - don't allow multiple threads in OSC recovery (3812)
4821 - fix debug_size parameters (3864)
4822 - fix mds_postrecov to initialize import for llog ctxt (3121)
4823 - replace config semaphore with spinlock (3306)
4824 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4825 - don't allow enqueue to complete on a destroyed export (3822)
4826 - down write_lock before checking llog header bitmap (3825)
4827 - recover from lock replay timeout (3764)
4828 - up llog sem before sending rpc (3652)
4829 - reduce ns lock hold times when setting kms (3267)
4830 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4831 - fix NULL deref and obd_dev leak on setup error (3312)
4832 - replace some LBUG about llog ops with error handling (3841)
4833 - don't match INVALID dentries from d_lookup and spin (3784)
4834 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4835 - fix invalid assertion in ptlrpc_set_wait (3880)
4837 - add libwrap support for the TCP acceptor (3996)
4838 - add /proc/sys/portals/routes for non-root route listing (3994)
4839 - allow setting MDS UUID in .xml (2580)
4840 - print the stack of a process that LBUGs (4228)
4842 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4845 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4846 - make sure to unset replay flag from failed open requests (3440)
4847 - if default stripe count is 0, use OST count for inode size (3636)
4848 - update parent mtime/ctime on client for create/unlink (2611)
4849 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4850 - free recovery state on server during a forced cleanup (3571)
4851 - unregister_reply for resent reqs (3063)
4852 - loop back devices mounting and status check on 2.6 (3563)
4853 - fix resource-creation race that can provoke i_size == 0 (3513)
4854 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4855 - more debugging for page-accounting assertion (3746)
4856 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4857 - avoid deadlock after precreation failure (3758)
4858 - fix race and lock order deadlock in orphan handling (3450, 3750)
4859 - add validity checks when grabbing inodes from l_ast_data (3599)
4861 - add /proc/.../recovery_status to obdfilter (3428)
4862 - lightweight CDEBUG infrastructure, debug daemon (3668)
4863 - change default OSC RPC parameters to be better on small clusters
4864 - turn off OST read cache for files smaller than 32MB
4865 - install man pages and include them in rpms (3100)
4866 - add new init script for (un)mounting lustre filesystems (2593)
4867 - run chkconfig in %post for init scripts (3701)
4868 - drop scimac NAL (unmaintained)
4870 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4873 - clean kiobufs before and after use (3485)
4874 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4875 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4876 - change init script to fail more clearly if not run as root (1528)
4877 - allow clients to reconnect during replay (1742)
4878 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4879 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4880 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4881 - fix cond_resched (3554)
4882 - only evict unfinished clients after recovery (3515)
4883 - allow bulk resend, prevent data loss (3570)
4884 - dynamic ptlrpc request buffer allocation (2102)
4885 - don't allow unlinking open directory if it isn't empty (2904)
4886 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4887 - remove extraneous obd dereference causing LASSERT failure (3334)
4888 - don't use get_cycles() when creating temp. files on the mds (3156)
4889 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4890 - handle EEXIST for set-stripe, set proper directory name (3336)
4892 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4893 - fix ksocknal_fmb_callback() error messages (2918)
4895 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4898 - don't copy lvb into (possibly NULL) reply on error (2983)
4899 - don't deref dentry after dput, don't free lvb on error (2922)
4900 - use the kms to determine writeback rpc length (2947)
4901 - increment oti_logcookies when osc is inactive (2948)
4902 - update client's i_blocks count via lvb messages (2543)
4903 - handle intent open/close of special files properly (1557)
4904 - mount MDS with errors=remount-ro, like obdfilter (2009)
4905 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4906 - don't use cancelling-locks' kms values (2947)
4907 - use highest lock extent for kms, not last one (2925)
4908 - don't dereference ERR_PTR() dentry in error handling path (3107)
4909 - fix thread race in portals_debug_dumplog() (3122)
4910 - create lprocfs device entries at setup instead of at attach (1519)
4911 - common AST error handler, don't evict client on completion race (3145)
4912 - zero nameidata in detach_mnt in 2.6 (3118)
4913 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4914 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4915 - initialize RPC timeout timer earlier for 2.6 (3219)
4916 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4917 - print nal/nid for unknown nid (3258)
4918 - additional checks for oscc recovery before doing precreate (3284)
4919 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4920 - don't crash in mdc_close for bad permissions on open (3285)
4921 - zero i_rdev for non-device files (3147)
4922 - clear page->private before handing to FS, better assertion (3119)
4923 - tune the read pipeline (3236)
4924 - fix incorrect decref of invalidated dentry (2350)
4925 - provide read-ahead stats and refine rpc in flight stats (3328)
4926 - don't hold journal transaction open across create RPC (3313)
4927 - update atime on MDS at close time (3265)
4928 - close LDAP connection when recovering to avoid server load (3315)
4929 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4930 - don't leak open file on MDS after open resend (3325)
4931 - serialize filter_precreate and filter_destroy_precreated (3329)
4932 - loop device shouldn't call sync_dev() for nul device (3092)
4933 - clear page cache after eviction (2766)
4934 - resynchronize MDS->OST in background (2824)
4935 - refuse to mount the same filesystem twice on same mountpoint (3394)
4936 - allow llmount to create routes for mounting behind routers (3320)
4937 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4938 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4939 - fix rare problem with rename on htree directories (3417)
4941 - allow default OST striping configuration per directory (1414)
4942 - fix compilation for qswnal for 2.6 kernels (3125)
4943 - increase maximum number of MDS request buffers for large systems
4944 - change liblustreapi to be useful for external progs like lfsck (3098)
4945 - increase local configuration timeout for slow disks (3353)
4946 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4948 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4951 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4952 - don't overwrite extent policy data in reply if lock was blocked (2901)
4953 - drop filter export grants atomically with removal from device (2663)
4954 - del obd_self_export from work_list in class_disconnect_exports (2908)
4955 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4956 - swab reply message in mdc_close, other PPC fixes (2464)
4957 - fix destroying of named logs (2325)
4958 - overwrite old logs when running lconf --write_conf (2264)
4959 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4960 - fix race in target_handle_connect (2898)
4961 - mds_reint_create() should take same inode create lock (2926)
4962 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4963 - don't close files for self_export to avoid uninitialized obd (2936)
4964 - allow MDS with the same name as client node (2939)
4965 - hold dentry reference for closed log files for unlink (2325)
4966 - reserve space for all logs during transactions (2059)
4967 - don't evict page beyond end of stripe extent (2925)
4968 - don't oops on a deleted current working directory (2399)
4969 - handle hard links to targets without a parent properly (2517)
4970 - don't dereference NULL lock when racing during eviction (2867)
4971 - don't grow lock extents when lots of conflicting locks (2919)
4973 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4976 - account for cache space usage on clients to avoid data loss (974)
4977 - lfsck support in lustre kernel code (2349)
4978 - reduce journal credits needed for BRW writes (2370)
4979 - orphan handling to avoid losing space on client/server crashes
4980 - ptlrpcd can be blocked, stopping ALL progress (2477)
4981 - use lock value blocks to assist in proper KMS, faster stat (1021)
4982 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4983 - recovery for initial connections (2355)
4984 - fixes for mds_cleanup_orphans (1934)
4985 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4986 - block all file creations until orphan recovery completes (1901)
4987 - client remove rq_connection from request struct (2423)
4988 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4989 - recovery timer race (2670)
4990 - mdc_close recovey bug (2532)
4991 - ptlrpc cleanup bug (2710)
4992 - mds timeout on local locks (2588)
4993 - namespace lock held during RPCs (2431)
4994 - handle interrupted sync write properly (2503)
4995 - don't try to handle a message that hasn't been replied to (2699)
4996 - client assert failure during cleanup after abort recovery (2701)
4997 - leak mdc device after failed mount (2712)
4998 - ptlrpc_check_set allows timedout requests to complete (2714)
4999 - wait for inflight reqs when ptlrpcd finishes (2710)
5000 - make sure unregistered services are removed from the srv_list
5001 - reset bulk XID's when resending them (caught by 1138 test)
5002 - unregister_bulk after timeout
5003 - fix lconf error (2694)
5004 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5005 - readahead locks pages, leaves pending causing memory pressure (2673)
5006 - increase OST request buffers to 4096 on large machines (2729)
5007 - fix up permission of existing directories in simple_mkdir (2661)
5008 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5009 - don't assign transno to errored transactions (2742)
5010 - don't delete objects on OST if given a bogus objid from MDS (2751)
5011 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5012 - if rq_no_resend, then timeout request after recovery (2432)
5013 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5014 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5015 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5016 - MDS recovery shouldn't delete objects that it has given out (2730)
5017 - if enqueue arrives after completion, don't clobber LVB (2819)
5018 - don't unlock pages twice when trigger_group_io returns error (2814)
5019 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5020 - don't write pages to disk if there was an error (1450)
5021 - don't ping imports that have recovery disabled (2676)
5022 - take buffered bytes into account when balancing socknal conn (2817)
5023 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5024 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5025 - remove little-endian swabbing of llog records (1987)
5026 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5027 - retry reposting request buffers if they fail (1191)
5028 - grow extent at grant time to avoid granting a revoked lock (2809)
5029 - lock revoke doesn't evict page if covered by a second lock (2765)
5030 - disable VM readahead to avoid reading outside lock extents (2805)
5032 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5033 - updated kernel patches for hp-2.4.20 kernel (2681)
5035 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5038 - fix truncated write corruption (2366)
5039 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5041 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5042 - ensure len(uuid) < 37 in lmc (1171)
5043 - fix ia64 OOPS in llog_test (2255)
5044 - zero end of page at obdfilter for partial page writes (2648)
5045 - don't leave stale dentries around after renames (bug 2428)
5046 - fix timeouts when evicting a client with a single lock held (2642)
5047 - set deadline for the initial HELLO message to drain (2634)
5048 - print out dotted-quad IP addresses in the socknal (2302)
5050 - additional debugging for MDS client eviction problem (2443)
5051 - fix mkfsoptions support for osts (2603, 2604)
5053 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5056 - add series for the vanilla 2.6.0 kernel
5057 - add series for the vanilla 2.4.24 kernel
5058 - add series for a cray x86/64 UL kernel drop
5059 - fix xattr patches for the vanilla 2.4.19 series
5061 - generate true UUIDs in lmc (1171)
5062 - have portals stack dumping break in UML (2466)
5063 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5064 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5065 - wake callback waiting threads on client eviction (2460)
5066 - Add --ptldebug and --subsystem to lmc (1719)
5067 - update assertion to allow safe interrupt allocation
5068 - set rq_no_resend for cancel requests (2432)
5069 - recalculate ptlrpcd timeout after resend (2494)
5070 - call vfs_rmdir when removing pending directories (2368)
5071 - fix renaming a file to itself (2429)
5072 - lmc creates a default one-stripe lov (2454)
5073 - expand procfs space to handle large clusters (2326)
5074 - increase UML stack to avoid overflow
5075 - update lconf's list of debug and subsystem masks
5076 - fix lfs find --obd (2510)
5077 - /proc tunable for disabling filter read caching (2591)
5078 - stop rpm packages from altering slapd.conf (2301)
5079 - disable nagle in the socknal under 0conf (2578)
5080 - choose mds inode size based on stripe count (2572)
5081 - fix kernel-source rpm problems (2516)
5083 - add --disable-doc to avoid pdf generation (2421)
5084 - update documentation, tests, type-os, comments
5085 - avoid format warnings on ia64
5086 - remove the TOE NAL
5087 - tiny code cleanups by removing unused fields
5089 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5092 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5093 - remove the most blatant lies from BUILDING file (2371)
5094 - change default debug level to reasonable production setting
5095 - reduce client side cache size to reduce cache flush time
5096 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5097 - make TCP zerocopy and pinger support enabled by default (2476)
5098 - sync writes completed after process exits caused crashes (2319)
5099 - maintain correct mount count on the MDS (2356)
5100 - backout 1557, because 2316 wasn't really fixed
5101 - better file I/O statistics gathering in /proc
5102 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5103 - another kernel patch to fix zero-copy TCP function export
5104 - don't take duplicate lock when processing re-sent getattr (2420)
5105 - lctl uses obd_self_export instead of creating new conn (2353)
5106 - MDS/OST recovery case which requires object creation asserted (2425)
5107 - move lfs from /usr/sbin to /usr/bin in packages
5108 - fix race between mds_client_add and mds_client_free (2417)
5109 - use kmalloc instead of slabs in portals (2430)
5110 - don't create duplicate records when a failover MDS is present (2442)
5111 - remove unnecessary mount age check (2332)
5112 - don't remove directory inodes from locks prematurely (2451)
5113 - don't break if MDS service name is the same as hostname (2103)
5114 - fix races in client write RPC generation when cache full (2482)
5116 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5119 - remove now-unused request->rq_obd (278)
5120 - if an allocation fails, print out how much memory we've used (1933)
5121 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5122 - add the "configurable stack size" patch to most series files (1256)
5123 - ability to write large log records, for 100+ OST configs (2306)
5124 - fix NULL deref when filter_prep fails (2314)
5125 - fix operator precedence error in filter_sync
5126 - dynamic allocation of socknal TX descriptors (2315)
5127 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5128 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5129 - ensure that CWARN messages in Portals always get to the syslog
5130 - __init/__exit are not for prototype decls (ldlm_init/exit)
5131 - x86-64 compile warning fixes
5132 - fix gateway LMC keyword conflict (2318)
5133 - fix MDS lock inversions in getattr/reint paths (1844)
5134 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5135 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5136 - DLM race condition prevented some lock evictions (2328)
5137 - ENOMEM detection and retry on socknal sends (2230)
5138 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5139 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5140 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5141 - better error messages when a client is rejected during recovery (1505)
5142 - avoid cancelling locks which were never granted, after failure (2330)
5143 - fix i_sem/journal inversion in mds_client_add (2333)
5144 - fix truncate/getattr lock cycle deadlock (2334)
5145 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5146 - fix two rare exit paths which could leak an l_lock() ref (2321)
5147 - fencepost error in MDS/OST orphan recovery (2226)
5148 - make log record alignment 8 bytes (1988)
5149 - lstripe now fails when requested offset > ost_count (2237)
5150 - ensure that all kernel series have a complete list.h (1607)
5151 - fix crashes in special-file operations (2316)
5152 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5154 - allow configurable automake binary, for testing new versions
5155 - small update to the lfs documentation
5157 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5159 * fix negative export reference count in fsfilt_sync (2312)
5161 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5162 * release candidate 0.9.1
5164 - orphans are moved into the PENDING directory for possible recovery
5165 - replayed opens now open by fid for orphan/rename safety (1042)
5166 - last close of an orphan inode generates a transno (683)
5167 - chdir() and mount() now pin the directory entry (1020)
5168 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5169 - discard very old requests without processing them (1502)
5170 - remove some common, well-understood CERRORs (1505)
5171 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5172 - clear "grant" flags in OST replies until OST grant code lands (1644)
5173 - fix read performance by not clobbering i_blksize on client (1598)
5174 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5175 - make LRU size a /proc tunable, clears locks when reduced (707)
5176 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5177 - server locks take references on exports now (1558)
5178 - build fixes for 2.4.20-rh trees (1663)
5179 - return an error from lov_create if all OSCs are inactive (1751)
5180 - fix import levels when a reconnect happens without a timeout (1597)
5181 - exit early from mds_open if we get a lookup error (1749)
5182 - partial page read at EOF wouldn't wait for disk before sending (1642)
5183 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5184 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5185 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5186 - write/truncate lock inversion (1639)
5187 - Don't auto-load obdclass, portals modules during cleanup (1495)
5188 - fix timestamps from jumping to "now" (1763)
5189 - extra journal assertions (1648)
5190 - add an extra multiunlink test (1771)
5191 - fix read_record/write_record API (1776)
5192 - fix leak of offset_extent, possible incorrect i_size later (1772)
5193 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5194 - Debugging for the kqswnal_get_idle_tx problems (1820)
5195 - Allow recovery to be attempted multiple times (1536)
5196 - Write out MDS last_rcvd file after it is first created (1600)
5197 - Fix tx_descriptor leak in failed transmit situations (1827)
5198 - ext3 journaling fixes for assertion failure after IO error (1871)
5199 - class_export_put() on freed export after completion AST error (1896)
5200 - Fix revalidate looping in VFS (1322)
5201 - Don't access a freed export during MDS_REINT timeout (1521)
5202 - Add open-unlink recovery support on the MDS (1673,1764)
5203 - Return an error if no MDS data was read from last_rcvd (1946)
5204 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5205 - Don't setup a disk device that doesn't match exported UUID (317)
5206 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5207 - avoid committing NULL handle in force close
5208 - local.sh is now a one-stripe LOV configuration
5209 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5210 - fix ext3 htree duplicate directory entry corruption (1516)
5211 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5212 - update to new LOV EA format (2097)
5213 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5215 2003-06-15 Phil Schwan <phil@clusterfs.com>
5218 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5219 - per-import/export recovery handling (958, 931, 959)
5220 - multiple last-rcvd slots, for serving multiple FSes (949)
5221 - connections are again shared between multiple imp/exports (963, 964)
5222 - "umount -f" would hang if any requests needed to be sent (393, 978)
5223 - avoid pinning large req buffer by copying for queued messages (989)
5224 - add "uuid" to "lctl device" command to help upcalls (991)
5225 - "open" RPCs with transnos would confuse recovery counters (1037)
5226 - do proper endian conversion of all wire messages (288, 340, 891)
5227 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5228 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5229 - fix ptlrpc_connection leak in target_handle_connect (1174)
5230 - fix import refcounting bug in OST and MDS cleanup (1134)
5231 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5232 - fix very unlikely obd_types race condition (501)
5233 - remove osc_open hack for echo_client (1187)
5234 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5235 - a failure in read_inode2 leads to deadlock (1139)
5236 - cancel ack-locks as soon as transaction is committed (1072)
5237 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5238 - make sure to commitrw after any preprw to avoid deadlock (1162)
5239 - failing to execute a file in a lustre FS would lock inode (1203)
5240 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5241 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5242 - obd_uuid2tgt would walk off the end of the list (1255)
5243 - on IA64 the osc would give portals incorrect bulk size (1258)
5244 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5245 - fix lock inversion caused by new llite matching code (1282)
5246 - limit the number of dirty pages on a client to 10MB (1286)
5247 - timed out locks were not being corrected cancelled (1289)
5248 - fix O_DIRECT above 4GB on IA-32 (1292)
5249 * major user-visible changes
5250 - fail out/fail over policy now controlled by the upcall (993)
5252 - add OBD_PING to check server availability and failure (954)
5253 - lustre messages are now sent in sending host order (288, 340, 891)
5254 - add eadatalen to MDS getattr reply (340)
5255 - OST read replies may contain second buffer, with per-page status (593)
5257 2003-03-11 Phil Schwan <phil@clusterfs.com>
5260 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5261 - failed open()s could cause deadlock; fixed (867, 869)
5262 - stop cancelling OST locks when files are closed (481)
5263 - overlapping XID spaces caused network corruption (851, 853)
5264 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5265 - setattr_raw added, to do single-RPC, server-side setattrs
5266 - lmc/lconf syntax change for OST UUIDs
5267 - fix crashy race condition between ptlrpc_free_req and osc_close
5268 - don't use request in mdc_enqueue if we hit a timeout (889)
5269 - don't set the inode i_size for regular files from the MDS (896)
5270 - handle out of order completion AST (842)
5271 - don't LBUG if a lock request times out after receiving AST (913)
5272 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5273 - if a bad lock AST arrives, send an error instead of dropping entirely
5274 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5275 - fix leak in bulk IO when only partially completed (899, 900, 926)
5276 - fix O_DIRECT for ia64 (55)
5277 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5278 - C-z after timeout could hang a process forever; fixed (977)
5280 - client-side I/O cache (678, 924, 929, 941, 970)
5282 - READPAGE and SETATTRs which don't take server-side locks get
5285 2003-02-11 Phil Schwan <phil@clusterfs.com>
5288 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5289 - Fix fsfilt_extN_readpage() to read a full page of directory
5290 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5291 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5292 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5293 - Most intent operations were removed, in favour of a new RPC mode
5294 that does a single RPC to the server and bypasses most of the VFS
5295 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5296 - Aggressively cancel local locks on DLM servers
5297 - mds_reint_unlink sends EA to the client if it's the last nlink.
5298 client uses that EA to unlink OST objects.
5299 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5300 - recursive symlinks were fixed (439)
5301 - fixed NULL deref in DEBUG_REQ
5302 - filter_update_lastobjid no longer calls sync, which annoyed extN
5303 - fixed multi-client small-writes to a single file problem (445)
5304 - fixed mtime updates during file writes (607)
5305 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5306 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5307 - fixed Linux OST size reporting problem (444, 656)
5308 - OST now updates object mtime with writes or setattr (607, 619)
5309 - client verifies file size before zeroing page past EOF (445)
5310 - OST now writes last allocated objid to disk with allocation (108)
5311 - LOV on echo now works (409)
5313 - mds_reint_unlink sends a new buffer, with the EA included. this
5314 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5315 set if a regular file was being unlinked, and it was the last link
5316 - use PtlGet from the target for bulk writes (315)
5317 - OST now updates object mtime with writes or setattr (607, 619)
5318 - LDLM now has a grant-time callback to revalidate locked items, if
5320 - Many MDS operations were reorganized to combat race conditions
5322 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5323 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5325 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5328 - Fully reactivate OST imports after reconnection (512, others)
5329 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5330 - More graceful error handling for truncating on dead OST (515)
5331 - Don't error out unless we're actually accessing dead stripes (474)
5332 - Fix garbage sizes when stripes are missing (410)
5333 - LRU counters were broken, causing constant lock purge (433, 432)
5334 - garbage on read from stripes with failed OSTs (441)
5335 - mark OSCs as active before reconnecting during recovery (438)
5336 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5337 - lfind did not preserve OST order in output (443)
5338 - symlinks cause hung clients, incorrect data (439)
5339 - stop dereferencing request after dropping refcount (457)
5340 - don't LASSERT(spin_is_locked) on non-SMP (455)
5341 - fixes for many rename() bugs
5342 - fstat didn't correctly synchronize attributes (399)
5343 - server must handle lock cancellation during blocking AST prep (487)
5344 - bulk descriptors were free()d too soon (511)
5345 - fix paths in lconf, which would load incorrect modules (451, 507)
5346 - fix confusing lconf 'host not found' error message (386)
5347 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5348 - fix race condition in mdc_blocking_ast() for inode access (526)
5349 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5350 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5351 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5352 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5353 - fix lstripe in several ways (kernel side) (527)
5354 - fix request leak in ldlm_cli_enqueue (262)
5355 - incorrect OSC was marked inactive after OST failure
5356 - call mds_fs_cleanup before unmounting filesystem (524)
5357 - fix races between taking ns_lock and ldlm_lock_change_resource
5358 - fix races updating LOV export open file list
5359 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5360 - fix recovery NULL deref in ldlm_cli_cancel_unused
5361 - fix some DLM races by using new hash table for lock handles (419)
5362 - permit the client to specify desired inodes, at replay
5363 - duplicate requests when we queue them for replay reintegration
5364 - fix last_rcvd offset calculation
5365 - sync after each recovered transaction, so we always make progress
5366 - never, not always, ERESTART requests without transnos
5367 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5369 - skip replay if the MDS says that the client is already connected
5370 - don't check for a recovery-enabled export to match lctl's UUID
5371 - don't INC_USE_COUNT for phantom exports
5372 - don't crash when cleaning up phantom exports (567)
5373 - don't double-finish or set replay data for errored mdc_open requests
5374 - abort requests when they time out, so we don't get old replies
5375 - send/receive replies for AST messages again
5376 - if the client says that it doesn't have the lock, cancel it on the
5378 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5379 mark it as destroyed, it will all work out when decref is called
5380 - fix module use counts (22, 581)
5382 - ASTs now expect a reply (server cancels lock on error reply)
5384 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5387 - fix many simultaneous client startup (392)
5388 - fix dentry->d_it clobbering
5389 - credentials weren't being shipped for readdir/getattr operations
5390 - remove invalid assertions triggered during some concurrent MD
5392 - proper Lustre versions added (336, 389)
5393 - fix memory leak for create error case (398)
5394 - fix LOV locking bug that would get cli/srv out of sync
5395 - fix echo client over LOV (409)
5396 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5397 - fix double-O_EXCL intent crash (424)
5398 - avoid sending multiple lock CANCELs (352)
5400 - MDS can do multi-client recovery (modulo bugs in new code)
5402 - many updates, edits, cleanups
5404 2002-11-18 Phil Schwan <phil@clusterfs.com>
5407 - fix null d_it dereference (346)
5408 - fix full OST/dbench hang (333)
5409 - fix permission problem with file removal (286)
5410 - fix removal of OSCs from LOV when they fail
5411 - fix NULL deref during bulk timeout (214)
5412 - fix problems related to multiple filesystems on one MDS (241)
5413 - fixed serious subtle metadata locking bugs
5414 - free locks on clients when inodes are removed due to memory
5416 - fix inode pointer in lock data (285)
5417 - partial support for multiple MDS on a single host (241)
5418 - data locks weren't cancelled at clear_inode time (290, 311)
5419 - intent locks could lead to unbounded lock growth (205)
5420 - added a maximum lock count, an LRU list, and a flusher
5421 - fix multiple rename (365)
5422 - properly abstracted the echo client
5423 - OSC locked 1 byte too many; fixed
5424 - rewrote brw callback code:
5425 - fixed recovery bugs related to LOVs (306)
5426 - fixed too-many-pages-in-one-write crash (191)
5427 - fixed (again) crash in sync_io_timeout (214)
5428 - probably fixed callback-related race (385)
5430 - Add capability to MDS protocol
5431 - LDLM cancellations and callbacks on different portals
5433 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5436 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5437 - defer open object destruction to close time (601981, 138)
5438 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5439 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5440 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5441 - don't call obd_setattr in ll_file_release for destroyed objects
5442 * protocol change to lustre_msg: move |version| and add |flags|
5443 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5444 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5445 * added replay of create, unlink, link and rename operations during
5446 MDS failover; recovery should be much more robust now
5447 * remove failed OSCs from LOVs (only lov_create uses this so far)
5448 * the lustre-HOWTO was brought (more) up to date (582544)
5450 2002-10-23 Phil Schwan <phil@clusterfs.com>
5453 - in-use dentries weren't being reused properly (617851)
5454 - prevent multiple LDLM setup (599178)
5455 - fix LOV size calculations for truncate (617853)
5456 - fix client handling of MDS intent errors (POSIX)
5457 - fix permission bug in lovstripe.c test (624321)
5458 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5459 - truncate past end of file could corrupt data
5460 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5461 - a race in recovery could return ETIMEDOUT to apps (623947)
5462 - building outside the source directory was fixed
5463 * the lustre-HOWTO was brought (more) up to date (582544)
5464 * major progress was made on recovery functionality
5466 2002-10-10 Phil Schwan <phil@clusterfs.com>
5469 - recovery deadlock fix
5470 - rm -rf causes LBUG fix (617817)
5471 - file open by multiple tasks fix (618962)
5472 - directory permissions bugs (602707 and 620007)
5473 - journal_stop fixed with locking (611313)
5474 - O_APPEND failures resolved (618273, perhaps 614459)
5475 - lconf PATH fix (619770)
5476 - IA64 build fix (621450)
5477 - RPC buffer sizes scale with amount of memory
5479 2002-10-01 Phil Schwan <phil@clusterfs.com>
5482 - locks would be cancelled without throwing away data pages,
5483 resulting in inconsistent data (605627)
5484 - inode attributes were not always being refreshed (605627, 612449)
5485 - lconf now continues to cleanup after lctl reports an error
5486 - MDS now enforces user permissions (602707)
5487 - lprocfs cleanup fixed, but not yet enabled (614157)
5488 - fixed infinite server hang, should a client not respond to an AST
5489 - avoid going into recovery if user calls readlink() with a buffer
5490 that's too small (613941)
5491 - AST RPCs no longer require replies (614867) -- this may be changed
5492 - don't crash server if client sends an IOV that's too big (611336)
5493 - fixed lock conversion deadlock (611892)
5494 - fixed the following of symlinks (614622)
5495 * recovery: the server can remove locks from a client that dies, other
5496 clients can make progress
5497 * more extN patch fixes
5498 * compile-time configurable ptlrpc buffer allocations
5500 - collaborative read cache document
5501 - Lustre Lite Performance CDR document-in-progress
5503 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5506 - fix typo in patch-2.4.18
5508 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5511 - clear ptlrpc request each time in handle_incoming_request()
5512 - unlink of files now destroys the object on the OST
5514 2002-09-19 Peter Braam <braam@clusterfs.com>
5516 * add hard link support
5517 * change obdfile creation method
5518 * kernel patch changed
5520 2002-09-19 Peter Braam <braam@clusterfs.com>
5523 - stack overflow bug in extN fixed
5525 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5527 * documentation updates
5528 - add man pages for config tools
5529 - update tests/README to describe testing with new config tools
5530 - finish metadata API descriptions
5531 * bug fixes and cleanups
5532 - statfs workaround for 16TB limit
5533 - LOV stripe allocation improved, can stripe on subset of OSTs
5534 - LOV file size/IO offset was wrong for files > 4GB in size
5535 - object EA data was being dropped, caused files to be unreadable
5536 - memory overflow with non-LOV OST caused memory corruption
5537 - fixed regression tests to work with new config tools, obdfilter
5538 - fixed bug when directory size became larger than 1 block
5539 - fixed bug (for single client case) when PWD was deleted
5540 - invalidate local directory pages when doing intent-based ops
5541 - avoid LDLM oops when lock callback contained bad data
5543 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5545 * documentation updates
5546 * bug fixes and cleanups
5547 - configuration tools
5550 - 64-bit compile warnings
5551 - 64-bit internal statfs data
5553 * test_brw on persistent OST devices
5555 * lprocfs (disabled)
5557 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5559 * documentation updates
5560 * bug fixes and cleanups
5561 * configuration tools
5563 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5566 * many small fixes to 0_5_4
5567 * io/network handling
5568 * thinkos in MDS operations
5570 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5573 * crucial basic fixes to 0.5.3
5574 * IOR, Iozone work over Elan
5577 2002-08-07 Phil Schwan <phil@clusterfs.com>
5578 * version 0_5_3, our first alpha
5579 * we use the new Portals iovs
5580 * documentation updates
5581 * bug fixes and cleanups
5582 * small changes in the DLM wire protocol
5584 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5585 * version 0_5_1 with some initial stability,
5586 * locking on MD and file I/O.
5587 * documentation updates
5588 * several bug fixes since 0.5.0
5589 * small changes in wire protocol
5591 2002-07-18 Phil Schwan <phil@clusterfs.com>
5593 * delivered as Lustre Light Alpha
5594 * fixed a crash after handling invalid MDS requests
5595 * fixed directory pages for architectures with non-4k pages sizes
5597 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5598 * release version v0_4_4
5599 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5601 * Fixes a number of interruption problems with OST operations.
5602 * Update documentation for portals header changes
5603 * Move all wire protocol structs/defines to lustre_idl.h
5604 * Fixes symlink length bug.
5605 * Add tcpdump to repository.
5607 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5608 * release version v0_4_3
5609 * Fixes statfs for inodes on extN.
5610 * Fixes bug in runtests which would delete /etc/hosts.
5611 * Use 64-bit object IDs wherever possible (not into VFS though)
5612 Remove ost_get_info, which is unused by lustre, and out of date.
5614 2002-07-03 Peter Braam <braam@clusterfs.com>
5615 * release version v0_4_2 Fixes a lookup error (type not passed)
5616 * move forward to head of Portals
5617 * move forward to latest Lustre kernel
5619 2002-06-25 Peter Braam <braam@clusterfs.com>
5620 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob