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: shutdown vs evict race
20 Details : client_disconnect_export vs connect request race.
21 if client will evicted at this time - we start invalidate
22 thread without referece to import and import can be freed
26 Frequency : rare, need acl's on inode.
28 Description: client can't handle ost additional correctly
29 Details : if ost was added after client connected to mds client can have
30 hit lnet_try_match_md ... to big messages to wide striped files.
31 in this case need teach client to handle config events about add
32 lov target and update client max ea size at that event.
34 Severity : enhancement
36 Description: Changelogs
37 Details : Changelogs are a lightweight mechanism to track filesystem
38 metadata and namespace changes. The changelog is recorded
39 permanently on the MDTs, and is periodically "consumed" / purged
40 when records are no longer needed.
42 Severity : enhancement
44 Description: compact fld format with extents
45 Details : Store range of seq rather than every seq in FLD. Seq
46 controller update FLD rather than clients. In Case of CMD, mdt0
47 has FLD, all other metadata server act as non persistent proxy
48 for FLD queries and cache fld entries in fld cache.
53 Description: don't skip ost target if they assigned to file
54 Details : Drop slow OSCs if we can, but not for requested start idx.
55 This means "if OSC is slow and it is not the requested
56 start OST, then it can be skipped, otherwise skip it only
57 if it is inactive/recovering/out-of-space.
61 Description: more cleanup in mds_lov
62 Details : not send LOV EA under replay, we can't know about they size at this
63 time. Don't allow client connect to mds before any ost connected,
64 for avoid problems with LOV EA size and returning EIO to client.
66 Severity : enhancement
68 Description: Interoperability at server side (Disk interoperability)
70 Severity : enhancement
72 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
74 Severity : enhancement
76 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
78 Severity : enhancement
80 Description: New client IO stack (CLIO).
82 Severity : enhancement
84 Description: Commit on sharing. Eliminate inter-client dependencies between
85 uncommitted transactions by doing transaction commits.
86 Thereby clients may recovery independently.
89 Frequency : Create a symlink file with a very long name
91 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
92 Details : If there is no extra space in the request for early cancels,
93 ldlm_req_handles_avail() returns 0 instead of a negative value.
95 Severity : enhancement
97 Description: Add /proc entry for import status
98 Details : The mdc, osc, and mgc import directories now have
99 an import directory that contains useful import data for debugging
102 Severity : enhancement
104 Description: Re-disable certain /proc logging
105 Details : Enable and disable client's offset_stats, extents_stats and
106 extents_stats_per_process stats logging on the fly.
109 Frequency : Only on FC kernels 2.6.22+
111 Description: oops in statahead
112 Details : Do not drop reference count for the dentry from VFS when lookup,
113 VFS will do that by itself.
115 Severity : enhancement
117 Description: Generic /proc file permissions
118 Details : Set /Proc file permissions in a more generic way to enable non-
119 root users operate on some /proc files.
123 Description: Hitting mdc_commit_close() ASSERTION
124 Details : Properly handle request reference release in
125 ll_release_openhandle().
129 Description: quota recovery deadlock during mds failover
130 Details : This patch includes att18982, att18236, att18237 in bz14840.
132 1. fix osts hang when mds does failover with quotaon
133 2. prevent watchdog storm when osts threads wait for the
138 Frequency : only patchless client
139 Description: add workaround for race between add/remove dentry from hash
141 Severity : enhancement
143 Description: Allow OST glimpses to return PW locks
147 Description: LBUG when llog conf file is full
148 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
152 Description: Prevent import from entering FULL state when server in recovery
156 Description: service mount cannot take device name with ":"
157 Details : Only when device name contains ":/" will mount treat it as
163 Description: replace ptlrpcd with the statahead thread to interpret the async
164 statahead RPC callback
168 Frequency : on recovery
169 Description: I/O failures after umount during fail back
170 Details : if client reconnected to restarted server we need join to recovery
171 instead of find server handler is changed and process self
172 eviction with cancel all locks.
174 Severity : enhancement
176 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
178 Severity : enhancement
180 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
182 Severity : enhancement
184 Description: Upcall on Lustre log has been dumped
185 Details : Allow for a user mode script to be called once a Lustre log has
186 been dumped. It passes the filename of the dumped log to the
187 script, the location of the script can be specified via
188 /proc/sys/lnet/debug_log_upcall.
193 Description: avoid idr_remove called for id which is not allocated.
194 Details : Move assigment s_dev for clustered nfs to end of initialization,
195 to avoid problem with error handling.
200 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
201 Details : When connection is reused this not moved from CONN_UNUSED_HASH
202 into CONN_USED_HASH and this prodice warning when put connection
203 again in unused hash.
206 Severity : enhancement
208 Description: File striping can now be set to use an arbitrary pool of OSTs.
210 Severity : enhancement
212 Description: Export bytes_read/bytes_write count on OSC/OST.
216 Description: Early reply size mismatch, MGC loses connection
217 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
218 the connect flags are properly negotiated.
223 Description: quotas are not honored with O_DIRECT
224 Details : all writes with the flag O_DIRECT will use grants which leads to
225 this problem. Now using OBD_BRW_SYNC to guard this.
229 Description: add quota statistics
230 Details : 1. sort out quota proc entries and proc code.
231 2. add quota statistics
233 Severity : enhancement
235 Description: enable quota support for HEAD.
239 Description: Properly propagate oinfo flags from lov to osc for statfs
240 Details : restore missing copy oi_flags to lov requests.
242 Severity : enhancement
244 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
245 : l_getgroups(8), lst(8), routerstat(8)
246 Details : included man pages for llobdstat(8), llstat(8),
247 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
249 Severity : enhancement
251 Description: configure's --enable-quota should check the
252 : kernel .config for CONFIG_QUOTA
253 Details : configure is terminated if --enable-quota is passed but
254 : no quota support is in kernel
258 Description: Remove portals compatibility
259 Details : Remove portals compatibility, not interoperable with releases
264 Description: Resolve device initialization race
265 Details : Prevent proc handler from accessing devices added to the
266 obd_devs array but yet be intialized.
268 Severity : enhancement
270 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
272 Severity : enhancement
274 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
279 Description: Using wrong pointer in osc_brw_prep_request
280 Details : Access to array[-1] can produce panic if kernel compiled with
281 CONFIG_PAGE_ALLOC enabled
285 Description: Client runs out of low memory
286 Details : Consider only lowmem when counting initial number of llap pages
290 Description: *optional* service tags registration
291 Details : if the "service tags" package is installed on a Lustre node
292 When the filesystem is mounted, a local-node service tag will
293 be created. See http://inventory.sun.com/ for more information
294 about the Service Tags asset management system.
298 Description: Kernel BUG tries to release flock
299 Details : Lustre does not destroy flock lock before last reference goes
300 away. So always drop flock locks when client is evicted and
301 perform unlock regardless of successfulness of speaking to MDS.
305 Description: add refcount protection for osc callbacks, avoid panic on shutdown
309 Description: sanity test 65a fails if stripecount of -1 is set
310 Details : handle -1 striping on filesystem in ll_dirstripe_verify
315 Description: ASSERTION(CheckWriteback(page,cmd)) failed
316 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
319 Severity : enhancement
321 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
325 Description: do not process already freed flock
326 Details : flock can possibly be freed by another thread before it reaches
327 to ldlm_flock_completion_ast.
331 Description: LBUG during stress test
332 Details : Need properly lock accesses the flock deadlock detection list.
336 Description: oops in page fault handler
337 Details : kernel page fault handler can return two special 'pages' in error
338 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
342 Description: timeout with invalidate import.
343 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
344 should be handled by ptlrpcd. This produce long age waiting and
345 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
347 Severity : enhancement
349 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
353 Description: enable MGS and MDT services start separately
354 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
355 a co-located MGS without starting the MGS, which is a complement
356 to 'nosvc' mount option.
360 Frequency : after recovery
361 Description: precreate to many object's after del orphan.
362 Details : del orphan st in oscc last_id == next_id and this triger growing
363 count of precreated objects. Set flag LOW to skip increase count
364 of precreated objects.
368 Frequency : rare, on clear nid stats
369 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
370 Details : when clean nid stats sometimes try destroy live entry,
371 and this produce panic in free.
375 Description: Stack overflow during MDS log replay
376 ease stack pressure by using a thread dealing llog_process.
380 Description: wait until IO finished before start new when do lock cancel.
381 Details : VM protocol want old IO finished before start new, in this case
382 need wait until PG_writeback is cleared until check dirty flag and
383 call writepages in lock cancel callback.
385 Severity : enhancement
387 Description: using special macro for print time and cleanup in includes.
391 Description: mds_mfd_close() ASSERTION(rc == 0)
392 Details : In mds_mfd_close(), we need protect inode's writecount change
393 within its orphan write semaphore to prevent possible races.
397 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
401 Frequency : rare, on shutdown ost
402 Description: don't hit live lock with umount ost.
403 Details : shrink_dcache_parent can be in long loop with destroy dentries,
404 use shrink_dcache_sb instead.
408 Description: don't panic with use echo client
409 Details : echo client pass NULL as client nid pointer and this produce null
414 Description: fix build on ppc32
415 Details : compile code with -m64 flag produce wrong object file for ppc32.
419 Description: add message levels for liblustreapi
423 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
424 Details : Move the 'good_osts' check before the 'total_bavail' check. This
425 will result in an -EAGAIN and in the exit call path we call
426 alloc_rr() which will with increasing aggressiveness attempt to
427 aquire precreated objects on the minimum number of required OSCs.
431 Description: Use old size assignment to avoid deadlock
432 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
433 the scheduling while holding a spinlock. We do not need locking
434 for size in ll_update_inode() because size is only updated from
435 the MDS for directories or files without objects, so there is no
436 other place to do the update, and concurrent access to such inodes
437 are protected by the inode lock.
441 Description: resolve "_IOWR redefined" build error on SLES10
445 Description: dump the memory debugging after all modules are unloaded to
446 suppress false negative in conf_sanity test 39
448 Severity : enhancement
450 Description: build kernel-ib packages for OFED 1.3 in our release cycle
455 Description: fix SLES kernel versioning
456 Details : the kernel version for our SLES 10 kernel did not include a "-"
457 before the "smp" at the end. while this was not a problem in
458 general, it did mean that software trying to use the kernel
459 version to try to detect a vendor specific kernel would fail.
460 this was most evident by the OFED build scripts.
464 Description: Don't update lov_desc members until making sure they are valid
465 Details : When updating lov_desc members via proc fs, need fix their
466 validities before doing the real update.
470 Description: don't put request into delay list while invalidate in flight.
471 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
472 import in flight. this produce timeout for invalidate and sometimes
473 can cause stale data.
477 Frequency : on ppc only
478 Description: not convert ost objects for directory because it's not exist.
479 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
483 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
485 Severity : enhancement
487 Description: Update to support 2.6.22.14 vanilla kernel.
491 Frequency : rare, on recovery
492 Description: read procfs can produce deadlock in some situation
493 Details : Holding lprocfs lock which send rpc can produce block for destroy
494 obd objects and this also block reconnect with -EALREADY. This
495 isn't fix all lprocfs bugs - but make it rare.
497 Severity : enhancement
499 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
502 Frequency : frequent on X2 node
504 Description: mdc_set_open_replay_data LBUG
505 Details : Set replay data for requests that are eligible for replay.
509 Description: lustre_mgs: operation 101 on unconnected MGS
510 Details : When MGC is disconnected from MGS long enough, MGS will evict the
511 MGC, and late on MGC cannot successfully connect to MGS and a lot
512 of the error messages complaining that MGS is not connected.
516 Frequency : on network error
517 Description: panic with double free request if network error
518 Details : mdc_finish_enqueue is finish request if any network error ocuring,
519 but it's true only for synchronus enqueue, for async enqueue
520 (via ptlrpcd) this incorrect and ptlrpcd want finish request
523 Severity : enhancement
525 Description: client-side metadata stat-ahead during readdir(directory readahead)
526 Details : perform client-side metadata stat-ahead when the client detects
527 readdir and sequential stat of dir entries therein
530 Frequency : on start mds
532 Description: Implement get_info(last_id) in obdfilter.
535 Frequency : occasional
537 Description: Correctly check stale fid, not start epoch if ost not support SOM
538 Details : open with flag O_CREATE need set old fid in op_fid3 because
539 op_fid2 was overwritten with new generated fid, but mds can answer
540 with one of these two fids and both is not stale. Setattr
541 incorrectly started an epoch and assume will be called
542 done_writing, but without SOM done_writing ever being called.
545 Frequency : rare, depends on device drivers and load
547 Description: MDS or OSS nodes crash due to stack overflow
548 Details : Code changes in 1.8.0 increased the stack usage of some functions.
549 In some cases, in conjunction with device drivers that use a lot
550 of stack the MDS (or possibly OSS) service threads could overflow
551 the stack. One change which was identified to consume additional
552 stack has been reworked to avoid the extra stack usage.
555 Frequency : occasional
557 Description: Do not fail import if osc_interpret_create gets -EAGAIN
558 Details : If osc_interpret_create got -EAGAIN it immediately exits and
559 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
560 oscc_has_objects and see OSC has no objests and call
561 oscc_internal_create to resend create request.
563 Severity : enhancement
565 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
567 Severity : enhancement
569 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
572 Frequency : very rare
574 Description: Fix replay if there is an un-replied request and open
575 Details : In some cases, older replay request will revert the
576 mcd->mcd_last_xid on MDS which is used to record the client's
579 Severity : enhancement
581 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
583 Severity : enhancement
585 Description: Add rhel5 support to HEAD.
587 Serverity : enhancement
589 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
594 Description: Don't allow skipping OSTs if index has been specified.
595 Details : Don't allow skipping OSTs if index has been specified, make
596 locking in internal create lots better.
600 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
601 Details : access to bitfield in structure is always rounded to long
602 and this produce problem with not atomic change any bit.
606 Description: Lustre make rpms failed.
607 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
608 builds Lustre rpms from tarball.
611 Frequency : rare on shutdown ost
613 Description: If llog cancel was not send before clean_exports phase, this can
614 produce deadlock in llog code.
615 Details : If llog thread has last reference to obd and call class_import_put
616 this produce deadlock because llog_cleanup_commit_master wait when
617 last llog_commit_thread exited, but this never success because was
618 called from llog_commit_thread.
622 Description: allow userland application know is lost one of stripes.
623 Details : fill lvb_blocks with error code on ost and return it to
624 application if error flag found.
628 Description: NULL lov_tgts causing MDS oops
629 Details : more safe checks for NULL lov_tgts for avoid oops.
631 Severity : enhancement
633 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
637 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
638 Details : If an OST is down the MDS will hang indefinitely in
639 obd_statfs_rqset() waiting for the statfs data. While for
640 MDS QOS usage of statfs, it should not stuck in waiting.
642 Severity : enhancement
644 Description: remote_acl support
645 Details : Support ACL-based permission check for remote user.
646 Support setfacl/getfacl for remote user with the utils
647 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
648 the system "{s,g}etfacl" utils.
650 Severity : enhancement
652 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
654 Severity : enhancement
656 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
660 Description: make mgs_setparam() handle fsname containing dash
661 Details : fsname containing a dash does not work with lctl conf_param
663 Severity : enhancement
665 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
667 Severity : enhancement
669 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
673 Frequency : rare, at shutdown
674 Description: access already free / zero obd_namespace.
675 Details : if client_disconnect_export was called without force flag set,
676 and exist connect request in flight, this can produce access to
677 NULL pointer (or already free pointer) when connect_interpret
678 store ocd flags in obd_namespace.
682 Frequency : only at startup
683 Description: not alloc memory with spinlock held.
684 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
685 if any spinlock held.
687 Severity : enhancement
689 Description: make lustre randomly fail allocating memory
690 Details : Make lustre randomly failed allocating memory for testing purpose.
692 Severity : enhancement
694 Description: lost problems with lov objid file
695 Details : Fixes some scability and access to not inited memory problems
696 in work with lov objdid file.
701 Description: lfs find does not continue on file error
702 Details : Continue other files processing when a file/dir is absent.
706 Description: Inconsistent usage of lustre_pack_reply()
707 Details : Standardize the usage of lustre_pack_reply() such that it
708 always generate a CERROR on failure.
711 Frequency : occasional
713 Description: MDS hang or stay in waiting lock
714 Details : If client receive lock with CBPENDING flag ldlm need send lock
715 cancel as separate rpc, to avoid situation when cancel request
716 can't processed due all i/o threads stay in wait lock.
720 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
724 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
728 Description: Update to SLES9 kernel-2.6.5-7.287.3.
732 Description: Update to RHEL4 latest kernel.
734 Severity : enhancement
736 Description: Build SLES10 patchless client fails
737 Details : The configure was broken by run ./configure with
738 --with-linux-obj=.... argument for patchless client. When the
739 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
740 can't search header adequately. Use absolute path such as
741 -I($LINUX)/include instead.
745 Description: interrupt oig_wait produce painc on resend.
746 Details : brw_redo_request can be used for resend requests from ptlrpcd and
747 private set, and this produce situation when rq_ptlrpcd_data not
748 copyed to new allocated request and triggered LBUG on assert
749 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
750 ptlrpcd set if request is changed and can be safety changed to use
755 Description: LASSERT_{REQ,REP}SWAB macros are buggy
756 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
757 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
758 these macros and replace them with their difinitions instead.
762 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
763 Details : Update which_patch & target file for SLES10 latest kernel.
766 Description: Remove client patches from SLES 10 kernel.
767 Details : This causes SLES 10 clients to behave as patchless clients
768 even on a Lustre-patched (server) kernel.
770 Severity : enhancement
772 Description: self-adjustable client's lru lists
773 Details : use adaptive algorithm for managing client cached locks lru
774 lists according to current server load, other client's work
775 pattern, memory activities, etc. Both, server and client
776 side namespaces provide number of proc tunables for controlling
779 Severity : enhancement
781 Description: light-weight GSS support
782 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
783 and reduce performance overhead.
785 Severity : enhancement
787 Description: Linux keyring support
788 Details : Support using service of Linux keyring for Lustre GSS internal
789 context refresh/cache mechanism.
793 Description: Fix errors in lfs documentation
794 Details : Fixes man pages
798 Description: don't use GFP_* in generic Lustre code.
799 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
803 Description: obdclass is limited by single OBD_ALLOC(idarray)
804 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
808 Description: warnings with build patchless client with vanila 2.6.19 and up
809 Details : change old ctl_table style and replace ctl_table/ctl_table_header
810 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
814 Frequency : Only for SLES
815 Description: sanity_quota fail test_1
816 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
817 file, which makes TSTID[2] inunique.
821 Frequency : Only for SLES9
822 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
823 Details : Update target/ChangeLog/which_patch .
827 Frequency : Only for test_10 in sanity-quota.sh
828 Description: a bug in quota test code
829 Details : checking a test flag in a wrong place
833 Frequency : Only for test_13 in sanity-quota.sh
834 Description: a bug in quota test script
835 Details : shouldn't check overfull allocated quota
839 Description: jbd statistics
840 Details : Port older jbd statistics patch for sles10
842 Severity : enhancement
844 Description: Add printing inode info into message about error in writepage.
849 Description: reply_lock_interpret crash due to race with it and lock cancel.
850 Details : Do not replay locks that are being cancelled. Do not reference
851 locks by their address during replay, just by their handle.
855 Frequency : When flocks are used.
856 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
857 Details : Flock locks might destroy just granted lock if it could be merged
858 with another existing flock, this is done in completion handler,
859 so teach ldlm_cli_enquque_fini that this is a valid case for
865 Description: Oops in read and write path when failing to allocate lock.
866 Details : Check if lock allocation failed and return error back.
870 Description: lstripe command fails for valid OST index
871 Details : The stripe offset is compared to lov->desc.ld_tgt_count
872 instead of lov->desc.ld_active_tgt_count.
876 Description: sanity.sh failed test 103
877 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
882 Description: df doesn't work properly if diskfs blocksize != 4K
883 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
887 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
888 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
890 Severity : enhancement
892 Description: Adaptive timeouts
893 Details : RPC timeouts adapt to changing server load and network
894 conditions to reduce resend attempts and improve recovery time.
898 Description: llapi_file_create() does not allow some changes
899 Details : add llapi_file_open() that allows specifying the mode and
900 open flags, and also returns an open file handle.
904 Description: merge and cleanup kernel patches.
905 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
909 Description: Add journal checksum support.(Kernel part)
910 Details : The journal checksum feature adds two new flags i.e
911 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
912 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
913 indicates that the commit block contains the checksum for
914 the blocks described by the descriptor blocks. Now commit
915 record can be sent to disk without waiting for descriptor
916 blocks to be written to disk. This behavior is controlled
917 using JBD2_FEATURE_ASYNC_COMMIT flag.
921 Description: OSS needs mutliple precreate threads
922 Details : Add ability to start more than one create thread per OSS.
926 Description: Sanity.sh test_65k failed: setstripe should have succeeded
927 Details : execute lfs setstripe on client
931 Description: mds_obd_create error creating tmp object
932 Details : When the user sets quota on root, llog will be affected and can't
933 create files and write files.
936 Frequency : Always on ia64 patchless client, and possibly others.
938 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
939 Details : This allows the patchless client to be loaded on architectures
944 Description: RedHat Update kernel for RHEL5
945 Details : Add the kernel config file for RHEL5.
949 Description: RedHat Update kernel for RHEL5
950 Details : Modify the kernel config file more closer RHEL5.
954 Description: Build failure against Centos5 (RHEL5)
955 Details : Define PAGE_SIZE when it isn't present.
959 Description: test_7 on sanity-quota.sh can't be used on separate machines
960 Details : add facet to handle it
964 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
965 Details : with NFS, the anon dentry's parent was set to itself in
966 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
967 corresponding dentry other than use rec->ur_name.
969 Severity : enhancement
971 Description: lfs setstripe enhancement
972 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
976 Description: enable data checksumming by default
977 Details : enable checksum by default, allow --disable-checksum
978 configure option and "-o nochecksum" mount option. Checksums
979 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
983 Description: lustre support for RHEL5
984 Details : Add support for RHEL5.
988 Description: Client eviction due to failover config
989 Details : after a connection loss, the lustre client should attempt to
990 reconnect to the last active server first before trying the
991 other potential connections.
995 Description: when mds and osts use different quota unit(32bit and 64bit),
996 quota will be released repeatly.
997 Details : void sending multiple quota reqs to mds, which will keep the
998 status between the reqs.
1002 Description: osts not allocated evenly to files
1003 Details : change the condition to increase offset_idx
1006 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1008 Description: Data corruption for OSTs that are formatted larger than 2TB
1010 Details : When generating the bio request for lustre file writes the
1011 sector number would overflow a temporary variable before being
1012 used for the IO. The data reads correctly from Lustre (which
1013 will overflow in a similar manner) but other file data or
1014 filesystem metadata may be corrupted in some cases.
1018 Description: Tune the kernel for good SCSI performance.
1019 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1020 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1025 Description: rewrite ext2-derived code in obdclass/uuid.c
1026 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1027 from scratch preserving functionality.
1031 Description: "lfs find -obd UUID" prints directories
1032 Details : "lfs find -obd UUID" will return all directory names instead
1033 of just file names. It is incorrect because the directories
1034 do not reside on the OSTs.
1038 Description: use i_size_read and i_size_write in 2.6 port
1039 Details : replace inode->i_size access with i_size_read/write()
1042 Frequency : when removing large files
1044 Description: scheduling issue during removal of large Lustre files
1045 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1046 It causes scheduling issues when removing large files (17TB in the
1050 Frequency : only with liblustre clients on XT3
1052 Description: evictions taking too long
1053 Details : allow llrd to evict clients directly on OSTs
1056 Frequency : only on ppc
1058 Description: /proc/fs/lustre/devices broken on ppc
1059 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1060 We should make sure the type of 'index' is loff_t and then cast
1061 explicitly as needed below. Do not assign an explicitly cast
1065 Frequency : only for rhel5
1067 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1068 Details : Modify the target file & which_kernel.
1070 Severity : enhancement
1072 Description: omit set fsid for export NFS
1073 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1074 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1078 Frequency : Always, for kernels after 2.6.16
1079 Description: Fix warning idr_remove called for id=.. which is not allocated.
1080 Details : Last kernels save old s_dev before kill super and not allow
1081 to restore from callback - restore it before call kill_anon_super.
1085 Description: buffer overruns could theoretically occur
1086 Details : llapi_semantic_traverse() modifies the "path" argument by
1087 appending values to the end of the origin string, and a
1088 overrun may occur. Adding buffer overrun check in liblustreapi.
1092 Description: Fix error on 'ls .' at the top of the Lustre mount.
1093 Details : Don't revalidate dentry if it is a root dentry.
1097 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1098 Details : Modify vm-tunables-rhel4.patch.
1102 Description: Kernel config for 2.6.18-vanilla.
1103 Details : Modify targets/2.6-vanilla.target.in.
1104 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1105 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1106 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1107 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1111 Description: improve handling recoverable errors
1112 Details : if request processig with error which can be recoverable on server
1113 request should be resend, otherwise page released from cache and
1118 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1119 Details : Modify target file & which_patch.
1120 A flaw was found in the IA32 system call emulation provided
1121 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1122 value could be stored in the %RAX register, which could trigger an
1123 out-of-bounds system call table access. An untrusted local user
1124 could exploit this flaw to run code in the kernel
1125 (ie a root privilege escalation). (CVE-2007-4573).
1129 Description: O_DIRECT bypasses client statistics.
1130 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1131 read_ahead_stats were updated. copied the stats section from
1132 osc_send_oap_rpc() into async_internal().
1136 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1140 Description: change order of libsysio includes
1141 Details : '#include sysio.h' should always come before '#include xtio.h'
1145 Description: handle "serious error: objid * already exists" more gracefully
1146 Details : If LAST_ID value on disk is smaller than the objects existing in
1147 the O/0/d* directories, it indicates disk corruption and causes an
1148 LBUG(). If the object is 0-length, then we should use the existing
1149 object. This will help to avoid a full fsck in most cases.
1151 Severity : enhancement
1153 Description: adapt the lustre_config script to support the upgrade case
1154 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1155 server targets to 1.6 in parallel.
1159 Description: To avoid grant space > avaible space when the disk is almost
1160 full. Without this patch you might see the error "grant XXXX >
1161 available" or some LBUG about grant, when the disk is almost
1163 Details : In filter_check_grant, for non_grant cache write, we should
1164 check the left space by if (*left > ungranted + bytes), instead
1165 of (*left > ungranted), because only we are sure the left space
1166 is enough for another "bytes", then the ungrant space should be
1167 increase. In client, we should update cl_avail_grant only there
1168 is OBD_MD_FLGRANT in the reply.
1172 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1173 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1174 described in CVE-2007-4573. This problem could allow untrusted
1175 local users to gain root access.
1178 Frequency : when using O_DIRECT and quotas
1180 Description: Incorrect file ownership on O_DIRECT output files
1181 Details : block usage reported by 'lfs quota' does not take into account
1182 files that have been written with O_DIRECT.
1187 Description: touch file failed when fs is not full
1188 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1189 otherwise we can get ENOSP while fs is not full.
1193 Description: parallel lock callbacks
1194 Details : Instead of sending blocking and completion callbacks as separated
1195 requests, adding them to a set and sending in parallel.
1198 Frequency : only for Cray XT3
1199 Bugzilla : 12829/13455
1200 Description: Changing primary group doesn't change the group lustre assigns to
1202 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1203 overridden with the primary group provided by the group upcall,
1204 whereas the supplied fsgid can be trusted if it is in the list of
1205 supplementary groups returned by the group upcall.
1207 Severity : enhancement
1209 Description: Allow masking D_WARNING, D_ERROR messages from console
1210 Details : Console messages can now be disabled via lnet.printk.
1214 Description: User code with malformed file open parameter crashes client node
1215 Details : Before packing join_file req, all the related reference should be
1216 checked carefully in case some malformed flags cause fake
1217 join_file req on client.
1221 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1222 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1223 mode is equal to the granted mode. However, it should also check
1224 that LDLM_FL_CP_REQD is not set.
1228 Description: Heavy nfs access might result in deadlocks
1229 Details : After ELC code landed, it is now improper to enqueue any mds
1230 locks under och_sem, because enqueue might want to decide to
1231 cancel open locks for same inode we are holding och_sem for.
1235 Description: Client eviction while running blogbench
1236 Details : A lot of unlink operations with concurrent I/O can lead to a
1237 deadlock causing evictions. To address the problem, the number of
1238 oustanding OST_DESTROY requests is now throttled to
1239 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1240 callbacks are processed in priority.
1244 Description: enable ACLs on MDS by default
1245 Details : ACLs must be enabled on MDS by default.
1248 Frequency : PPC/PPC64 only
1250 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1251 Details : fix duplicated definitions between asm-ppc64/types.h and
1252 lustre_types.h on PPC.
1255 Frequency : PPC/PPC64 only
1257 Description: asm-ppc/segment.h does not exist
1258 Details : fix compile issue on PPC.
1262 Description: better handle error messages in extents code
1265 Frequency : RHEL4 only
1267 Description: mkfs is very slow on IA64/RHEL4
1268 Details : A performance regression has been discovered in the MPT Fusion
1269 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1270 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1271 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1273 Severity : enhancement
1275 Description: SNMP support enhancement
1276 Details : Adding total number of sampled request for an MDS node in snmp
1279 Severity : enhancement
1281 Description: Optimize ldlm waiting list processing for PR extent locks
1282 Details : When processing waiting list for read extent lock and meeting read
1283 lock that is same or wider to it that is not contended, skip
1284 processing rest of the list and immediatelly return current
1285 status of conflictness, since we are guaranteed there are no
1286 conflicting locks in the rest of the list.
1290 Description: Time out and refuse to reconnect
1291 Details : When the failover node is the primary node, it is possible
1292 to have two identical connections in imp_conn_list. We must
1293 compare not conn's pointers but NIDs, otherwise we can defeat
1294 connection throttling.
1298 Description: port llog fixes from b1_6 into HEAD
1299 Details : Port llog reference couting and some llog cleanups from b1_6
1300 (bug 10800) into HEAD, for protect from panic and access to already
1301 free llog structures.
1305 Description: Detect stride IO mode in read-ahead
1306 Details : When a client does stride read, read-ahead should detect that and
1307 read-ahead pages according to the detected stride pattern.
1311 Description: data checksumming impacts single node performance
1312 Details : add support for several checksum algorithm. Currently, only CRC32
1313 and Adler-32 are supported. The checksum type can be changed on
1314 the fly via /proc/fs/lustre/osc/*/checksum_type.
1318 Description: use adler32 for page checksums
1319 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1324 Description: build for x2 fails
1325 Details : fix compile issue on Cray systems.
1329 Description: Properly match for duplicate locks
1330 Details : Due to different lock order from skiplists code, we need to
1331 traverse entire list for now
1334 Frequency : only on PPC/SLES10
1336 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1337 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1338 be defined. Add a hack in mkfs_lustre.c to work around this compile
1343 Description: LASSERT on MDS when client holding flock lock dies
1344 Details : ldlm pool logic depends on number of granted locks equal to
1345 number of released locks which is not true for flock locks, so
1346 just exclude such locks from consideration.
1350 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1351 Details : Use fsfilt_sync() for both the whole filesystem sync and
1352 individual file sync to eliminate dangerous inode locking
1353 with I_LOCK that can lead to a deadlock.
1357 Description: performance in 1.6.3
1358 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1359 sure that 1MB requests are not fragmented by the block layer.
1361 Severity : enhancement
1363 Description: organize the server-side client stats on per-nid basis
1364 Details : Change the structure of stats under obdfilter and mds to
1372 The "uuid"s file would list the uuids of _active_ exports.
1373 And the clear entry is to clear all stats and stale nids.
1375 Severity : enhancement
1377 Description: eliminate client locks in face of contention
1378 Details : file contention detection and lockless i/o implementation
1379 for contended files.
1383 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1386 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1388 Description: MDS may lose file striping (and hence file data) in some cases
1389 Details : If there are additional extended attributes stored on the MDS,
1390 in particular ACLs, SELinux, or user attributes (if user_xattr
1391 is specified for the client mount options) then there is a risk
1392 of attribute loss. Additionally, the Lustre file striping
1393 needs to be larger than default (e.g. striped over all OSTs),
1394 and an additional attribute must be stored initially in the
1395 inode and then increase in size enough to be moved to the
1396 external attribute block (e.g. ACL growing in size) for file
1401 Description: skiplist implementation simplification
1402 Details : skiplists are used to group compatible locks on granted list
1403 that was implemented as tracking first and last lock of each
1404 lock group the patch changes that to using doubly linked lists
1408 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1409 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1410 because LASSERT on both the data supplied by a client, and the
1411 data on disk is dangerous and incorrect.
1413 Severity : enhancement
1415 Description: Slow truncate/writes to huge files at high offsets.
1416 Details : Directly associate cached pages to lock that protect those pages,
1417 this allows us to quickly find what pages to write and remove
1418 once lock callback is received.
1422 Description: more ldlm soft lockups
1423 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1424 starve other threads from the resource lock for a long time in
1425 case of long waiting queue, so change the debug level from
1426 D_OTHER to the less frequently used D_INFO.
1428 Severity : enhancement
1430 Description: add -gid, -group, -uid, -user options to lfs find
1434 Description: Hung threads in invalidate_inode_pages2_range
1435 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1436 once one is completed. As a result, some RPCs are stuck in the
1437 queue and are never sent.
1441 Description: filter threads hungs on waiting journal commit
1442 Details : Cleanup filter group llog code, then only filter group llog will
1443 be only created in the MDS/OST syncing process.
1447 Description: Procfs and llog threads access destoryed import sometimes.
1448 Details : Sync the import destoryed process with procfs and llog threads by
1449 the import refcount and semaphore.
1451 Severity : enhancement
1453 Description: openlock cache of b1_6 port to HEAD
1458 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1459 Details : If insertion of an extent fails, then discard the inode
1460 preallocation and free data blocks else it can lead to duplicate
1465 Description: don't always update ctime in ext3_xattr_set_handle()
1466 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1467 In some cases the ctime should not be updated, for example for
1468 2.0->1.8 compatibility it is necessary to delete an xattr and it
1469 should not update the ctime.
1473 Bugzilla : 15713/16362
1474 Description: Assertion in iopen_connect_dentry in 1.6.3
1475 Details : looking up an inode via iopen with the wrong generation number can
1476 populate the dcache with a disconneced dentry while the inode
1477 number is in the process of being reallocated. This causes an
1478 assertion failure in iopen since the inode's dentry list contains
1479 both a connected and disconnected dentry.
1483 Description: assertion failure in ldlm_handle2lock()
1484 Details : fix a race between class_handle_unhash() and class_handle2object()
1485 introduced in lustre 1.6.5 by bug 13622.
1490 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1491 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1496 Description: oops in sd_iostats_seq_show()
1497 Details : unloading/reloading the scsi low level driver triggers a kernel
1498 bug when trying to access the sd iostat file.
1503 Description: Kernel panics during QLogic driver reload
1504 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1505 patch, causing memory corruption.
1510 Description: journal_dev option does not work in b1_6
1511 Details : pass mount option during pre-mount.
1513 Severity : enhancement
1515 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1516 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1517 extent information of a file. It can be used to map logical blocks
1518 in a file to physical blocks in the block device.
1522 Description: LDLM soft lockups - improvement
1523 Details : It is be possible to send the lock handle along with each read
1524 or write request because the client is already doing a lock match
1525 itself so there isn't any reason the OST should have to re-do that
1529 Frequency : only on Cray X2
1531 Description: X2 build failures
1532 Details : fix build failures on Cray X2.
1536 Description: xid & resent requests
1537 Details : Initialize RPC XID from clock at startup (randomly if clock is
1540 Severity : enhancement
1542 Description: Add lustre_start utility to start or stop multiple Lustre servers
1547 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1548 Details : In case of memory pressure, list_del() can be called twice on
1549 req->rq_history_list, causing a kernel oops.
1553 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1554 Details : fix stack overflow in the distributed lock manager by defering
1555 export eviction after a failed AST to the elt thread instead of
1556 handling it in the dlm interpret routine.
1560 Description: Convert some comments to new format.
1561 Details : Update documenting comments to match doxygen conventions.
1565 Description: Grammar fixes.
1566 Details : A couple of trivial spelling fixes.
1570 Description: OSD_COUNTERS-mandatory
1571 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1572 to simplify the code.
1576 Description: simplify cmm_device freeing logic.
1577 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1581 Description: Add lockdep support to dt_object_operations locking interface.
1582 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1583 parameter indicating lock ordering. Update mdd code to use new
1588 Description: Introduce failloc constants for lockless IO tests.
1589 Details : Add two new failloc constants to test lockless IO. Only one of
1590 them in implemented---another is checked in yet to be landed
1595 Description: Add lockdep support for inode mutex.
1596 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1597 in the situations where Lustre has to lock more than one inode
1602 Description: Add optional invariants checking support.
1603 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1604 configure switch. This macro is to be used for consistency and
1605 sanity checks that are too expensive to be left in `production'
1610 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1611 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1616 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1617 Details : Use spin_lock_nested() in (the only) situation where more than
1618 one ldlm_lock is locked simultaneously. Also, fix possible
1619 dead-lock in ldlm_lock_change_resource() by enforcing particular
1624 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1625 Details : Instead of specifying each ldlm_lock call-back through separate
1626 parameter, wrap them into struct ldlm_callback_suite.
1630 Description: Kill join_lru obd method and its callers.
1631 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1632 in memory---a requirement implemented through ->o_join_lru() obd
1633 method in HEAD. Remove this method and its users.
1637 Description: Add asynchronous ldlm ENQUEUE completion handler.
1638 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1639 case that stock ldlm_completion_ast() cannot handle as it waits
1640 until lock is granted. Introduce new ldlm_completion_ast_async()
1641 for this. Also comment ldlm_completion_ast().
1645 Description: ldlm_error <-> errno conversion.
1646 Details : Add functions to map (rather arbitrary) between LDLM error codes
1647 and standard errno values. CLIO needs this to prevent LDLM specific
1648 constants from escaping ldlm and osc.
1652 Description: Kill unused ldlm_handle2lock_ns() function.
1653 Details : Kill unused ldlm_handle2lock_ns() function.
1657 Description: Add lu_ref support to ldlm_lock
1658 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1659 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1660 and ldlm_resource. LDLM interface has to be changed, because code that
1661 releases a reference on a lock, has to "know" what reference this is.
1662 In the most frequent case
1664 lock = ldlm_handle2lock(handle);
1666 LDLM_LOCK_PUT(lock);
1668 no changes are required. When any other reference (received _not_ from
1669 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1670 instead of LDLM_LOCK_PUT().
1672 Arguably, changes are pervasive, and interface requires some discipline
1673 for proper use. On the other hand, it was very instrumental in finding
1674 a few leaked lock references.
1678 Description: Add ldlm_lock_addref_try().
1679 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1680 attempts to addref a lock that might be being canceled
1685 Description: Add ldlm_weigh_callback().
1686 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1687 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1688 hard-coded `number of pages' logic.
1692 Description: Add lockdep annotations to llog code.
1693 Details : Use appropriately tagged _nested() locking calls in the places
1694 where llog takes more than one ->lgh_lock lock.
1698 Description: Add loi_kms_set().
1699 Details : Wrap kms updates into a helper function.
1703 Description: Constify instances of struct lsm_operations.
1704 Details : Constify instances of struct lsm_operations.
1708 Description: lu_conf support.
1709 Details : On a server, a file system object is uniquely identified
1710 by a fid, which is sufficient to locate and load all object
1711 state (inode). On a client, on the other hand, more data are
1712 necessary instantiate an object. Change lu_object_find() and
1713 friends to take additional `lu_conf' argument describing object.
1714 Typically this includes layout information.
1718 Description: lu_context fixes.
1719 Details : Introduce new lu_context functions that are needed on the client
1720 side, where some system threads (ptlrpcd) are shared by multiple
1721 modules, and so cannot be stopped during module shutdown.
1725 Description: Add start and stop methods to lu_device_type_operations.
1726 Details : Introduce two new methods in lu_device_type_operations, that are
1727 invoked when first instance of a given type is created and last one
1728 is destroyed respectively. This is need by CLIO.
1732 Description: Add lu_ref support to struct lu_device.
1733 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1734 track leaked references.
1738 Description: Introduce lu_kmem_descr.
1739 Details : lu_kmem_descr and its companion interface allow to create
1740 and destroy a number of kmem caches at once.
1744 Description: Fix lu_object finalization race.
1745 Details : Fix a race between lu_object_find() finding an object and its
1746 concurrent finalization. This race is (most likely) not possible
1747 on the server, but might happen on the client.
1751 Description: Introduce lu_ref interface.
1752 Details : lu_ref is a debugging module allowing to track references to
1753 a given object. It is quite cpu expensive, and has to be
1754 explicitly enabled with --enable-lu_ref. See usage description
1759 Description: Factor lu_site procfs stats into a separate function.
1760 Details : Separate lu_site stats printing code into a separate function
1761 that can be reused on a client.
1765 Description: Constify instances of struct {lu,dt,md}_device_operations.
1766 Details : Constify instances of struct {lu,dt,md}_device_operations.
1770 Description: Introduce struct md_site and move meta-data specific parts of
1771 struct lu_site here.
1772 Details : Move md-specific fields out of struct lu_site into special struct
1773 md_site, so that lu_site can be used on a client.
1777 Description: Kill mdd_lov_destroy().
1778 Details : Remove unused mdd code.
1782 Description: Add st_block checking to multistat.c.
1783 Details : Add st_block checking to multistat.c.
1787 Description: Add lu_ref support to struct obd_device.
1788 Details : Add lu_ref tracking to obd_device.
1792 Description: Kill obd_set_fail_loc().
1793 Details : Remove unused code.
1797 Description: Add special type for ptlrpc_request interpret functions.
1798 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1799 there. Actual usage will be in CLIO.
1803 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1804 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1805 doesn't work with lockdep.
1809 Description: Add rwv.c test program.
1810 Details : New testing program exercising readv(2) and writev(2) (Qian).
1814 Description: Add sendfile.c test program.
1815 Details : New testing program exercising sendfile(2) (Jay).
1819 Description: Ratelimit a message that can be very frequent.
1820 Details : Ratelimit a memory allocation failure message that can
1825 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1826 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1830 Description: Liblustre build fixes.
1831 Details : Liblustre build fixes.
1835 Description: libcfs: add cfs_{need,cond}_resched() interface.
1836 Details : libcfs: add cfs_{need,cond}_resched() definition and
1837 implementations for Linux, NT, and liblustre.
1839 Severity : enhancement
1841 Description: More exported tunables for mballoc
1842 Details : Add support for tunable preallocation window and new tunables for
1843 large/small requests
1847 Description: Detect corruption of block bitmap and checking for preallocations
1848 Details : Checks validity of on-disk block bitmap. Also it does better
1849 checking of number of applied preallocations. When corruption is
1850 found, it turns filesystem readonly to prevent further corruptions.
1854 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1855 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1856 window to 0. If it does hit cache miss, and read-pattern is still
1857 stride-io mode, does not reset the stride window, but also does not
1858 increase the stride window length in this case.
1862 Frequency : only for big-endian servers
1863 Description: Check if system is big-endian while mounting fs with extents feature
1864 Details : Mounting a filesystem with extents feature will fail on big-endian
1865 systems since ext3-based ldiskfs is not supported on big-endian
1866 systems. This can be over-riden with "bigendian_extents" mount option.
1868 Severity : enhancement
1870 Description: The root squash functionality
1871 Details : A security feature, which is to prevent users from being able
1872 to mount lustre on their desktop, run as root, and delete
1873 all of the files in the filesystem. The goal is accomplished by
1874 remapping user id (UID) and group id (GID) of the root user to
1875 a UID and GID specified by the system administartor via Lustre
1876 configuration management server (MGS). The functionality also
1877 allows to specify sets of clients for which the remapping does
1882 Description: Excessive recovery window
1883 Details : With AT enabled, the recovery window can be excessively long (6000+
1884 seconds). To address this problem, we no longer use
1885 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1886 timeout no longer depends on the service time, it is set to
1887 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1888 time via pb_service_time.
1890 --------------------------------------------------------------------------------
1892 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1894 * Support for kernels:
1895 2.6.5-7.283 (SLES 9),
1896 2.6.9-55.EL (RHEL 4),
1897 2.6.16.46-0.14 (SLES 10),
1898 2.6.18.8 vanilla (kernel.org)
1899 * Client support for unpatched kernels:
1900 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1901 2.6.16 - 2.6.22 vanilla (kernel.org)
1902 * Due to recently discovered recovery problems, we do not recommend
1903 using patchless RHEL 4 clients with this or any earlier release.
1904 * Recommended e2fsprogs version: 1.39.cfs8
1905 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1906 * Starting with this release, the ldiskfs backing filesystem required
1907 by Lustre is now in its own package, lustre-ldiskfs. This package
1908 should be installed. It is versioned separately from Lustre and
1909 may be released separately in future.
1913 Description: block reactivating mgc import until all deactivates complete
1914 Details : Fix race when failing back MDT/MGS to itself (testing)
1917 Frequency : at statup only
1919 Description: mds_lov_synchronize race leads to various problems
1920 Details : simultaneous MDT->OST connections at startup can cause the
1921 sync to abort, leaving the OSC in a bad state.
1923 Severity : enhancement
1925 Description: add optional extra BUILD_VERSION info
1926 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1927 to override the lustre version.
1930 Frequency : 2.6.18 servers only
1932 Description: ll_kern_mount() doesn't release the module reference
1933 Details : The ldiskfs module reference count never drops down to 0
1934 because ll_kern_mount() doesn't release the module reference.
1939 Description: server LBUG when using old ost_num_threads parameter
1940 Details : Accept the old ost_num_threads parameter but warn that it
1941 is deprecated, and fix an off-by-one error that caused an LBUG.
1946 Description: Transient SCSI error results in persistent IO issue
1947 Details : iobuf->dr_error is not reinitialized to 0 between two
1951 Frequency : sometimes when underlying device returns I/O errors
1953 Description: OSTs not going read-only during write failures
1954 Details : OSTs are not remounted read-only when the journal commit threads
1955 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1956 instead of the ext3 wrappers.
1960 Description: poor connect scaling with increasing client count
1961 Details : Don't run filter_grant_sanity_check for more than 100 exports
1962 to improve scaling for large numbers of clients.
1965 Frequency : SLES10 only
1967 Description: sanity-quota.sh quotacheck failed: rc = -22
1968 Details : Quotas cannot be enabled on SLES10.
1971 Frequency : liblustre clients only
1973 Description: getdirentries does not give error when run on compute nodes
1974 Details : getdirentries does not fail when the size specified as an argument
1975 is too small to contain at least one entry
1977 Severity : enhancement
1979 Description: Add LNET router traceability for debug purposes
1980 Details : If a checksum failure occurs with a router as part of the
1981 IO path, the NID of the last router that forwarded the bulk data
1982 is printed so it can be identified.
1987 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1988 Details : Due to a race condition, liblustre clients were occasionally
1989 evicted incorrectly.
1991 Severity : enhancement
1993 Description: lfs setstripe use optional parameters instead of postional
1996 Severity : enhancement
1998 Description: Nanosecond timestamp support for ldiskfs
1999 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2000 resolution timestamps. There is not yet support for this at
2001 the Lustre filesystem level.
2004 Frequency : during server recovery
2006 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2007 Details : request with rq_no_resend flag not awake l_wait_event if they get
2011 Frequency : nfs export on patchless client
2013 Description: connectathon hang when test nfs export over patchless client
2014 Details : Disconnected dentry cannot be found with lookup, so we do not need
2015 to unhash it or make it invalid
2018 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2019 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2020 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2025 Description: open req refcounting wrong on reconnect
2026 Details : If reconnect happened between getting open reply from server and
2027 call to mdc_set_replay_data in ll_file_open, we will schedule
2028 replay for unreferenced request that we are about to free.
2029 Subsequent close will crash in variety of ways.
2030 Check that request is still eligible for replay in
2031 mdc_set_replay_data().
2036 Description: disable writes to filesystem when reading health_check file
2037 Details : the default for reading the health_check proc file has changed
2038 to NOT do a journal transaction and write to disk, because this
2039 can cause reads of the /proc file to hang and block HA state
2040 checking on a healthy but otherwise heavily loaded system. It
2041 is possible to return to the previous behaviour during configure
2042 with --enable-health-write.
2044 Severity : enhancement
2046 Description: 64-bit inode version
2047 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2048 to the inode. This will be required for version-based recovery.
2053 Description: MDS fails to start if a duplicate client export is detected
2054 Details : in some rare cases it was possible for a client to connect to
2055 an MDS multiple times. Upon recovery the MDS would detect this
2056 and fail during startup. Handle this more gracefully.
2058 Severity : enhancement
2060 Description: Add -o localflock option to simulate old noflock
2062 Details : This will achieve local-only flock/fcntl locks
2068 Description: log_commit_thread vs filter_destroy race leads to crash
2069 Details : Take import reference before releasing llog record semaphore
2074 Description: Wrong request locking in request set processing
2075 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2076 imp_delayed_list, in this place should be used imp_lock.
2079 Frequency : when reconnection
2081 Description: Grant Leak when osc reconnect to OST
2082 Details : When osc reconnect ost, OST(filter) should check whether it
2083 should grant more space to client by comparing fed_grant and
2084 cl_avail_grant, and return the granted space to client instead
2085 of "new granted" space, because client will call osc_init_grant
2086 to update the client grant space info.
2089 Frequency : when client reconnect to OST
2091 Description: Grant Leak when osc do resend and replay bulk write
2092 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2093 bulk write request, because the grant info while be sync between
2094 OSC and OST when reconnect, and we should ignore the grant info
2095 these of resend/replay write req.
2100 Description: Grant space more than avaiable left space sometimes.
2101 Details : When then OST is about to be full, if two bulk writing from
2102 different clients came to OST. Accord the avaliable space of the
2103 OST, the first req should be permitted, and the second one
2104 should be denied by ENOSPC. But if the seconde arrived before
2105 the first one is commited. The OST might wrongly permit second
2106 writing, which will cause grant space > avaiable space.
2109 Frequency : when client is evicted
2111 Description: Grant might be wrongly erased when osc is evicted by OST
2112 Details : when the import is evicted by server, it will fork another
2113 thread ptlrpc_invalidate_import_thread to invalidate the
2114 import, where the grant will be set to 0. While the original
2115 thread will update the grant it got when connecting. So if
2116 the former happened latter, the grant will be wrongly errased
2117 because of this race.
2122 Description: Checking Stale with correct fid
2123 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2124 to check whether it is stale, because sometimes, we want the
2125 enqueue happened anyway, and op_data.fid2 will not be initialized.
2127 Severity : enhancement
2129 Description: update patchless client
2130 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2133 Frequency : only with 2.4 kernel
2135 Description: random memory corruption
2136 Details : size of struct ll_inode_info is to big for union inode.u and this
2137 can be cause of random memory corruption.
2142 Description: Memory leak in recovery
2143 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2144 It should also check obd_fail before fsfilt_start, otherwise if
2145 fsfilt_start return -EROFS,(failover mds during mds recovery).
2146 then the req will return with repmsg->transno = 0 and rc = EROFS.
2147 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2148 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2149 be freed no matter whether fsfilt_commit success or not.
2152 Frequency : only with huge count clients
2154 Description: Prevents from taking the superblock lock in llap_from_page for
2156 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2157 for llap_from_page call in ll_removepage prevents from taking the
2158 superblock lock for a soon died page.
2163 Description: Not check open intent error before release open handle
2164 Details : in some rare cases, the open intent error is not checked before
2165 release open handle, which may cause
2166 ASSERTION(open_req->rq_transno != 0), because it tries to release
2167 the failed open handle.
2172 Description: Set cat log bitmap only after create log success.
2173 Details : in some rare cases, the cat log bitmap is set too early. and it
2174 should be set only after create log success.
2178 Description: Accessing a block bevice can re-enable I/O when Lustre is
2179 tearing down a device.
2180 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2185 Description: service threads may hog cpus when there are a lot of requests
2186 Details : Insert cond_resched to give other threads a chance to use some CPU
2191 Description: the cat log was not initialized in recovery
2192 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2193 unlink log on mds will not be initialized until mds post
2194 recovery. And also in mds post recovery, the unlink log will
2195 initialization will be done asynchronausly, so there will be race
2196 between add unlink log and unlink log initialization.
2200 Description: brw_stats were being printed incorrectly
2201 Details : brw_stats were being printed as log2 but all of them were not
2202 recorded as log2. Also remove some code duplication arising from
2203 filter_tally_{read,write}.
2207 Frequency : rare, only in recovery.
2208 Description: ASSERTION(req->rq_type != LI_POISON) failed
2209 Details : imp_lock should be held while iterating over imp_sending_list for
2210 prevent destroy request after get timeout in ptlrpc_queue_wait.
2214 Description: replay-single.sh test 52 fails
2215 Details : A lock's skiplist need to be cleanup when it being unlinked
2216 from its resource list.
2220 Description: Short directio read returns full requested size rather than
2222 Details : Direct I/O operations should return actual amount of bytes
2223 transferred rather than requested size.
2225 Severity : enhancement
2227 Description: metadata RPC reduction (e.g. for rm performance)
2228 Details : decrease the amount of synchronous RPC between clients and servers
2229 by canceling conflicing lock before the operation on the client
2230 and packing thier handles into the main operation RPC to server.
2232 Severity : enhancement
2234 Description: Async OSC create to avoid the blocking unnecessarily.
2235 Details : If a OST has no remain object, system will block on the creating
2236 when need to create a new object on this OST. Now, ways use
2237 pre-created objects when available, instead of blocking on an
2238 empty osc while others are not empty. If we must block, we block
2239 for the shortest possible period of time.
2243 Description: improve handling recoverable errors
2244 Details : if request processig with error which can be recoverable on server
2245 request should be resend, otherwise page released from cache and
2248 Severity : enhancement
2250 Description: refine locking for avoid write wrong info into lov_objid file
2251 Details : fix possible races with add new target and write/update data in
2254 --------------------------------------------------------------------------------
2256 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2261 Frequency : on some architectures
2263 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2264 Details : Uninitialized flags sometimes cause configuration commands to
2268 Frequency : patchless clients only
2270 Description: missing __iget() symbol export
2271 Details : The __iget() symbol export is missing. To avoid the need for
2272 this on patchless clients the deathrow inode reaper is turned
2273 off, and we depend on the VM to clean up old inodes. This
2274 dependency was during via the fix for bug 12181.
2276 --------------------------------------------------------------------------------
2278 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2280 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2281 INTEROPERATE with older versions automatically. In many cases a
2282 special upgrade step is needed. Please read the
2283 user documentation before upgrading any part of a live system.
2284 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2285 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2286 * WARNING: Lustre configuration and startup changes are required with
2287 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2291 Severity : enhancement
2293 Description: MountConf
2294 Details : Lustre configuration is now managed via mkfs and mount
2295 commands instead of lmc and lconf. New obd types (MGS, MGC)
2296 are added for dynamic configuration management. See
2297 https://mail.clusterfs.com/wikis/lustre/MountConf for
2300 Severity : enhancement
2302 Description: dynamic OST addition
2303 Details : OSTs can now be added to a live filesystem
2305 Severity : enhancement
2307 Description: startup order invariance
2308 Details : MDTs and OSTs can be started in any order. Clients only
2309 require the MDT to complete startup.
2311 Severity : enhancement
2313 Description: parallel, asynchronous orphan cleanup
2314 Details : orphan cleanup is now performed in separate threads for each
2315 OST, allowing parallel non-blocking operation.
2317 Severity : enhancement
2319 Description: optimized stripe assignment
2320 Details : stripe assignments are now made based on ost space available,
2321 ost previous usage, and OSS previous usage, in order to try
2322 to optimize storage space and networking resources.
2324 Severity : enhancement
2326 Description: Permanently set tunables
2327 Details : All writable /proc/fs/lustre tunables can now be permanently
2328 set on a per-server basis, at mkfs time or on a live system.
2330 Severity : enhancement
2332 Description: Lustre message v2
2333 Details : Add lustre message format v2.
2335 Severity : enhancement
2337 Description: client OST exclusion list
2338 Details : Clients can be started with a list of OSTs that should be
2339 declared "inactive" for known non-responsive OSTs.
2343 Description: SPEC SFS validation failure on NFS v2 over lustre.
2344 Details : Changes the blocksize for regular files to be 2x RPC size,
2345 and not depend on stripe size.
2347 Severity : enhancement
2349 Description: Multiple MD RPCs in flight.
2350 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2351 To avoid overly-overloading MDS, introduce a limit on number of
2352 MDS RPCs in flight for a single client and add /proc controls
2353 to adjust this limit.
2355 Severity : enhancement
2357 Description: client read/write statistics
2358 Details : Add client read/write call usage stats for performance
2359 analysis of user processes.
2360 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2361 file access. extents_stats shows chunk size distribution.
2362 extents_stats_per_process show chunk size distribution per
2365 Severity : enhancement
2367 Description: mds statistics
2368 Details : Add detailed mds operations statistics in
2369 /proc/fs/lustre/mds/*/stats.
2373 Description: Failure of copying files with lustre special EAs.
2374 Details : Client side always return success for setxattr call for lustre
2375 special xattr (currently only "trusted.lov").
2380 Description: Refcount LNET uuids
2381 Details : The global LNET uuid list grew linearly with every startup;
2382 refcount repeated list entries instead of always adding to
2385 Severity : enhancement
2387 Description: Dynamic service threads
2388 Details : Within a small range, start extra service threads
2389 automatically when the request queue builds up.
2392 Frequency : mixed-endian client/server environments
2394 Description: mixed-endian crashes
2395 Details : The new msg_v2 system had some failures in mixed-endian
2398 Severity : enhancement
2400 Description: Easy OST removal
2401 Details : OSTs can be permanently deactivated with e.g. 'lctl
2402 conf_param lustre-OST0001.osc.active=0'
2404 Severity : enhancement
2406 Description: MGS proc entries
2407 Details : Added basic proc entries for the MGS showing what filesystems
2410 Severity : enhancement
2412 Description: provide MGS failover
2413 Details : Added config lock reacquisition after MGS server failover.
2415 Severity : enhancement
2417 Description: add Linux 2.4 support
2418 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2422 Description: a large application tries to do I/O to the same resource and dies
2423 in the middle of it.
2424 Details : Check the req->rq_arrival time after the call to
2425 ost_brw_lock_get(), but before we do anything about
2426 processing it & sending the BULK transfer request. This
2427 should help move old stale pending locks off the queue as
2428 quickly as obd_timeout.
2431 Frequency : when an incorrect nid is specified during startup
2433 Description: ptlrpc connect to non-existant node causes kernel crash
2434 Details : LNET can't be re-entered from an event callback, which
2435 happened when we expire a message after the export has been
2436 cleaned up. Instead, hand the zombie cleanup off to another
2439 Severity : enhancement
2441 Description: plain/inodebits lock performance improvement
2442 Details : Grouping plain/inodebits in granted list by their request modes
2443 and bits policy, thus improving the performance of search through
2447 Frequency : only if OST filesystem is corrupted
2449 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2450 Details : for a short time RPCs with bulk IO are in the replay list,
2451 but replay of bulk IOs is unimplemented. If the OST filesystem
2452 is corrupted due to disk cache incoherency and then replay is
2453 started it is possible to trip an assertion. Avoid putting
2454 committed RPCs into the replay list at all to avoid this issue.
2457 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2459 Description: System hang on startup
2460 Details : This bug allowed the liblustre (e.g. catamount) client to
2461 return to the app before handling all startup RPCs. This
2462 could leave the node unresponsive to lustre network traffic
2463 and manifested as a server ptllnd timeout.
2465 Severity : enhancement
2467 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2468 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2469 boolean to control whether to print peer debug info when a
2470 client's RPC times out.
2473 Frequency : only for kernels with patches from Lustre below 1.4.3
2475 Description: Remove old rdonly API
2476 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2479 Frequency : only for devices with external journals
2481 Description: Set external device read-only also
2482 Details : During a commanded failover stop, we set the disk device
2483 read-only while the server shuts down. We now also set any
2484 external journal device read-only at the same time.
2487 Frequency : when upgrading from 1.4 while trying to change parameters
2489 Description: The wrong (new) MDC name was used when setting parameters for
2490 upgraded MDT's. Also allows changing of OSC (and MDC)
2491 parameters if --writeconf is specified at tunefs upgrade time.
2494 Frequency : when setting specific ost indicies
2496 Description: QOS code breaks on skipped indicies
2497 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2498 created with --index need not be sequential.
2500 Severity : enhancement
2502 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2503 Details : The uninit_groups feature works in conjunction with the kernel
2504 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2505 up the pass1 processing of e2fsck. This is a read-only feature
2506 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2507 mount filesystems that have had this feature enabled.
2509 Severity : enhancement
2511 Description: Improve multi-block allocation algorithm to avoid fragmentation
2512 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2513 allocation locality and avoid filesystem fragmentation.
2515 ------------------------------------------------------------------------------
2517 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2519 * Support for kernels:
2520 2.4.21-47.0.1.EL (RHEL 3)
2521 2.6.5-7.283 (SLES 9)
2522 2.6.9-42.0.10.EL (RHEL 4)
2523 2.6.12.6 vanilla (kernel.org)
2524 2.6.16.27-0.9 (SLES 10)
2525 * Recommended e2fsprogs version: 1.39.cfs5
2527 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2531 Frequency : occasional, depends on client load and configuration
2532 Bugzilla : 12181, 12203
2533 Description: data loss for recently-modified files
2535 Details : In some cases it is possible that recently written or created
2536 files may not be written to disk in a timely manner (this should
2537 normally be within 30s unless client IO load is very high).
2538 The problem appears as zero-length files or files that are a
2539 multiple of 1MB in size after a client crash or client eviction
2540 that are missing data at the end of the file.
2542 This problem is more likely to be hit on clients where files are
2543 repeatedly created and unlinked in the same directory, clients
2544 have a large amount of RAM, have many CPUs, the filesystem has
2545 many OSTs, the clients are rebooted frequently, and/or the files
2546 are not accessed by other nodes after being written.
2548 The presence of the problem can be detected by looking at
2549 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2550 smaller than the second (nr_unused) then dirty files will not
2551 be flushed automatically to disk. "sync; sleep 10" should be
2552 run several times on the node before unmounting it to update
2553 Lustre (this is also safe to run on nodes without this problem).
2555 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2556 that can cause this same problem, so customers using that kernel
2557 also need to update the kernel in addition to Lustre. In order
2558 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2560 It is normal that files written just before a client crash (less
2561 than 30s) may not yet have been flushed to disk, even for local
2565 Frequency : frequent on thin XT3 nodes
2567 Description: UUID collision on thin XT3 Linux nodes
2568 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2569 randomly, since we relied on an insufficiently-seeded PRNG.
2574 Description: OSS hangs after "All ost request buffers busy"
2575 Details : A deadlock between quota and journal operations caused OSS
2576 hangs after printing "All ost request buffers busy."
2579 Frequency : always on liblustre builds
2581 Description: Cleanup compiler warnings on liblustre
2584 Frequency : always on liblustre builds on XT3
2586 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2590 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2591 > 32000 subdirectories
2592 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2593 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2594 subdirectory count crosses 32000. This will aid e2fsck to
2595 correctly handle more than 32000 subdirectories.
2598 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2600 Description: System hang on startup
2601 Details : This bug allowed the liblustre (e.g. catamount) client to
2602 return to the app before handling all startup RPCs. This
2603 could leave the node unresponsive to lustre network traffic
2604 and manifested as a server ptllnd timeout.
2606 Severity : enhancement
2608 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2609 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2610 boolean to control whether to print peer debug info when a
2611 client's RPC times out.
2616 Description: make O_SYNC working on 2.6 kernels
2617 Details : 2.6 kernels use different method for mark pages for write,
2618 so need add a code to lustre for O_SYNC work.
2623 Description: Failure to close file and release space on NFS
2624 Details : Put inode details into lock acquired in ll_intent_file_open.
2625 Use mdc_intent_lock in ll_intent_open to properly
2626 detect all kind of errors unhandled by mdc_enqueue.
2631 Description: proc file read during shutdown sometimes raced obd removal,
2633 Details : Add lock to prevent obd access after proc file removal.
2636 Frequency : Only for files larger than 4GB on 32-bit clients.
2638 Description: improperly doing page alignment of locks
2639 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2640 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2645 Description: RPCs being resent when they shouldn't be
2646 Details : Some RPCs that should not be resent are being resent. This
2647 can cause inconsistencies in the RPC state machine. Do not
2648 resend such requests.
2651 Frequency : rare, only with NFS export
2653 Description: Crash on NFS re-export node
2654 Details : under very unusual load conditions an assertion is hit in
2655 ll_intent_file_open()
2658 Frequency : only if OST filesystem is corrupted
2660 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2661 Details : for a short time RPCs with bulk IO are in the replay list,
2662 but replay of bulk IOs is unimplemented. If the OST filesystem
2663 is corrupted due to disk cache incoherency and then replay is
2664 started it is possible to trip an assertion. Avoid putting
2665 committed RPCs into the replay list at all to avoid this issue.
2670 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2671 Details : Large single O_DIRECT read and write calls can fail to allocate
2672 a sufficiently large buffer to process the request. In case of
2673 allocation failure the allocation is retried with a smaller
2674 buffer and broken into smaller requests.
2676 Severity : enhancement
2678 Description: Add -o localflock option to simulate old noflock behaviour.
2679 Details : This will achieve local-only flock/fcntl locks coherentness.
2684 Description: versioning check is incomplete
2685 Details : Checking the version difference of client vs. server, report
2686 error if the gap is too big.
2691 Description: add support PG_writeback bit
2692 Details : add support for PG_writeback bit for Lustre, for more carefull
2693 work with page cache in 2.6 kernel. This also fix some deadlocks
2694 and remove hack for work O_SYNC with 2.6 kernel.
2696 Severity : enhancement
2698 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2699 Details : The uninit_groups feature works in conjunction with the kernel
2700 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2701 up the pass1 processing of e2fsck. This is a read-only feature
2702 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2703 mount filesystems that have had this feature enabled.
2705 Severity : enhancement
2707 Description: Improve multi-block allocation algorithm to avoid fragmentation
2708 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2709 allocation locality and avoid filesystem fragmentation.
2711 ------------------------------------------------------------------------------
2713 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2715 * Support for kernels:
2716 2.6.9-42.0.3.EL (RHEL 4)
2717 2.6.5-7.276 (SLES 9)
2718 2.4.21-47.0.1.EL (RHEL 3)
2719 2.6.12.6 vanilla (kernel.org)
2720 2.6.16.21-0.8 (SLES10)
2721 * Recommended e2fsprogs version: 1.39.cfs2-0
2723 * The backwards-compatible /proc/sys/portals symlink has been removed
2724 in this release. Before upgrading, please ensure that you change
2725 any configuration scripts or /etc/sysctl.conf files that access
2726 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2727 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2728 in advance of the upgrade on any system running Lustre 1.4.6 or
2729 newer, since /proc/sys/lnet was added in that version.
2730 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2734 Frequency : only when quota is used
2736 Description: avoid scanning export list for quota master
2737 Details : Change the algorithms to avoid scanning export list in order
2738 to improve the efficiency.
2741 Frequency : MDS failover only, very rarely
2743 Description: "went back in time" messages on mds failover
2744 Details : The greatest transno may be lost when the current operation
2745 finishes with an error (transno==0) and the client's last_rcvd
2746 record is over-written. Save the greatest transno in the
2747 mds_last_transno for this case.
2750 Frequency : always for specific kernels and striping counts
2752 Description: client may get "Matching packet too big" without ACL support
2753 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2754 when trying to access files in certain configurations. The
2755 clients should in fact be denied when mounting because they do
2756 not understand ACLs.
2759 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2761 Description: many clients connecting with IO in progress causes connect timeouts
2762 Details : Avoid synchronous journal commits to avoid delays caused by many
2763 clients connecting/disconnecting when bulk IO is in progress.
2764 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2765 OST_IO_PORTAL to avoid delays behind potentially many pending
2769 Frequency : occasionally with multiple writers to a single file
2771 Description: shared writes to file may result in wrong size reported by stat()
2772 Details : Allow growing of kms when extent lock is cancelled
2775 Frequency : always with random mmap IO to multi-striped file
2777 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2778 Details : Only if the hole is at the end of OST object so that kms is too
2779 small. Fix is to increase kms accordingly in ll_nopage.
2782 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2784 Description: writes to a missing object would leak memory on the OST
2785 Details : If there is an inconsistency between the MDS and OST filesystems,
2786 such that the MDS references an object that doesn't exist, writes
2787 to that object will leak memory due to incorrect cleanup in the
2788 error handling path, eventually running out of memory on the OST.
2793 Description: Creating too long symlink causes lustre errors
2794 Details : Check symlink and name lengths before sending requests to MDS.
2797 Frequency : only if flock is enabled (not on by default)
2799 Description: posix locks not released on fd closure on 2.6.9+
2800 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2801 kernels, this caused such locks not to be released on fd close and
2802 then assertions on fs unmount about still used locks.
2805 Frequency : MDS failover only, very rarely
2807 Description: clients may get ASSERTION(granted_lock != NULL)
2808 Details : When request was taking a long time, and a client was resending
2809 a getattr by name lock request. The were multiple lock requests
2810 with the same client lock handle and
2811 mds_getattr_name->fixup_handle_for_resent_request found one of the
2812 lock handles but later failed with ASSERTION(granted_lock != NULL).
2817 Description: handle->h_buffer_credits > 0, assertion failure
2818 Details : h_buffer_credits is zero after truncate, causing assertion
2819 failure. This patch extends the transaction or creates a new
2823 Frequency : NFS re-export or patchless client
2824 Bugzilla : 11179, 10796
2825 Description: Crash on NFS re-export node (__d_move)
2826 Details : We do not want to hash the dentry if we don't have a lock.
2827 But if this dentry is later used in d_move, we'd hit uninitialised
2828 list head d_hash, so we just do this to init d_hash field but
2829 leave dentry unhashed.
2832 Frequency : NFS re-export or patchless client
2834 Description: NFS exports has problem with symbolic link
2835 Details : lustre client didn't properly install dentry when re-exported
2836 to NFS or running patchless client.
2839 Frequency : NFS re-export or patchless client
2841 Description: Various nfs/patchless fixes.
2842 Details : fixes reuse disconected alias for lookup process - this fixes
2843 warning "find_exported_dentry: npd != pd",
2844 fix permission error with open files at nfs.
2845 fix apply umask when do revalidate.
2848 Frequency : occasional
2850 Description: Crash on NFS re-export node
2851 Details : calling clear_page() on the wrong pointer triggered oops in
2852 generic_mapping_read().
2855 Frequency : rarely, using O_DIRECT IO
2857 Description: unaligned directio crashes client with LASSERT
2858 Details : check for unaligned buffers before trying any requests.
2861 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2863 Description: stale data returned from RAID cache
2864 Details : If only a small amount of IO is done to the RAID device before
2865 reading it again it is possible to get stale data from the RAID
2866 cache instead of reading it from disk.
2869 Frequency : always for sles10 kernel
2871 Description: sles10 support
2872 Details : ll_follow_link: compile fixes and using of nd_set_link
2873 under newer kernels.
2876 Frequency : depends on arch, kernel and compiler version, always on sles10
2879 Description: recursive or deep enough symlinks cause stack overflow
2880 Details : getting rid of large stack-allocated variable in
2884 Frequency : depends on hardware
2886 Description: lustre write performance loss in the SLES10 kernel
2887 Details : the performance loss is caused by using of write barriers in the
2888 ext3 code. The SLES10 kernel turns barrier support on by
2889 default. The fix is to undo that change for ldiskfs.
2891 ------------------------------------------------------------------------------
2893 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2895 * Support for kernels:
2896 2.6.9-42.0.3EL (RHEL 4)
2897 2.6.5-7.276 (SLES 9)
2898 2.4.21-47.0.1.EL (RHEL 3)
2899 2.6.12.6 vanilla (kernel.org)
2903 Frequency : quota enabled and large files being deleted
2905 Description: releasing more than 4GB of quota at once hangs OST
2906 Details : If a user deletes more than 4GB of files on a single OST it
2907 will cause the OST to spin in an infinite loop. Release
2908 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2913 Description: statfs data retrieved from /proc may be stale or zero
2914 Details : When reading per-device statfs data from /proc, in the
2915 {kbytes,files}_{total,free,avail} files, it may appear
2916 as zero or be out of date.
2919 Frequency : systems with MD RAID1 external journal devices
2921 Description: lconf's call to blkid is confused by RAID1 journal devices
2922 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2923 of returning all block devices that match the journal UUID.
2926 Frequency : always, for aggregate stripe size over 4GB
2928 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2929 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2930 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2933 Frequency : always if "lfs find" used on a local file/directory
2935 Description: "lfs find" segfaults if used on a local file/directory
2936 Details : The case where a directory component was not specified wasn't
2937 handled correctly. Handle this properly.
2940 Frequency : always on ppc64
2942 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2943 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2944 perform endianness conversion.
2947 Frequency : rarely (truncate to non-zero file size after write under load)
2948 Bugzilla : 10730, 10687
2949 Description: Files padded with zeros to next 4K multiple
2950 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2951 it is possible that files that are truncated to a non-zero size
2952 immediately after being written are filled with zero bytes beyond
2953 the truncated size. No file data is lost.
2955 Severity : enhancement
2957 Description: Allow recovery/failover for liblustre clients.
2958 Details : liblustre clients were unaware of failover configurations until
2961 Severity : enhancement
2963 Description: user file locks should fail when not mounting with flock option
2964 Details : Set up an error-returning stub in ll_file_operations.lock field
2965 to prevent incorrect behaviour when client is mounted without
2966 flock option. Also, set up properly f_op->flock field for
2970 Frequency : always on ia64
2972 Description: "lfs df" loops on printing out MDS statfs information
2973 Details : The obd_ioctl_data was not initialized and in some systems
2974 this caused a failure during the ioctl that did not return
2975 an error. Initialize the struct and return an error on failure.
2978 Frequency : SLES 9 only
2980 Description: Error of copying files with lustre special EAs as root
2981 Details : Client side always return success for setxattr call for lustre
2982 special xattr (currently only "trusted.lov").
2985 Frequency : rarely on clusters with both ia64+i386 clients
2987 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2988 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2989 clients are doing concurrent writes to the same file it is
2990 possible that the ia64 clients may LASSERT because the OST
2991 extent locks are not PAGE_SIZE aligned. Ensure that grown
2992 locks are always aligned on the request boundary.
2995 Frequency : specific use, occasional
2997 Description: Overwriting in use executable truncates on-disk binary image
2998 Details : If one node attempts to overwrite an executable in use by
2999 another node, we now correctly return ETXTBSY instead of
3000 truncating the file.
3002 Severity : enhancement
3004 Description: Async OSC create to avoid the blocking unnecessarily.
3005 Details : If a OST has no remain object, system will block on the creating
3006 when need to create a new object on this OST. Now, ways use
3007 pre-created objects when available, instead of blocking on an
3008 empty osc while others are not empty. If we must block, we block
3009 for the shortest possible period of time.
3014 Description: chmod on Lustre root is propagated to other clients
3015 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3016 invalid by the follow_mount time.
3019 Frequency : liblustre clients only
3021 Description: Race in 'instant cancel' lock handling could lead to such locks
3022 never to be granted in case of SMP MDS
3023 Details : Do not destroy not yet granted but cbpending locks in
3027 Frequency : replay/resend of open
3029 Description: non null lock assetion failure in mds_intent_policy
3030 Details : Trying to replay/resend lockless open requests resulted in
3031 mds_open() returning 0 with no lock. Now it sets a flag if
3032 a lock is going to be returned.
3034 Severity : enhancement
3036 Description: Checksum enhancements
3037 Details : New checksum enhancements allow for resending RPCs that failed
3040 Severity : enhancement
3042 Description: Tunables on number of dirty pages in cacche
3043 Details : Allow to set limit on number of dirty pages cached.
3048 Description: client crash on unmount - lock still has references
3049 Details : In some error handling cases it was possible to leak a lock
3050 reference on a client while accessing a file. This was not
3051 harmful to the client during operation, but would cause the
3052 client to crash when the filesystem is unmounted.
3055 Frequency : specific case, rare
3057 Description: ETXTBSY on mds though file not in use by client
3058 Details : ETXTBSY is no longer incorrectly returned when attempting to
3059 chmod or chown a directory that the user previously tried to
3060 execute or a currently-executing binary.
3063 Frequency : extremely rare except on liblustre-based clients
3065 Description: Lustre space not freed when files are deleted
3066 Details : Clean up open-unlinked files after client eviction. Previously
3067 the unlink was skipped and the files remained as orphans.
3072 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3073 Details : In some cases it was possible to send a blocking callback to
3074 a client doing a glimpse, even though that client didn't get
3075 a lock granted. When the glimpse lock is cancelled on the OST
3076 the freed lock is left on the waiting list and corrupted the list.
3079 Frequency : all core dumps
3081 Description: Broke core dumps to lustre
3082 Details : Negative dentry may be unhashed if parent does not have UPDATE
3083 lock, but some callers, e.g. do_coredump, expect dentry to be
3084 hashed after successful create, hash it in ll_create_it.
3086 ------------------------------------------------------------------------------
3088 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3090 * Support for kernels:
3091 2.6.9-42.0.2.EL (RHEL 4)
3092 2.6.5-7.276 (SLES 9)
3093 2.4.21-40.EL (RHEL 3)
3094 2.6.12.6 vanilla (kernel.org)
3098 Frequency : always on RHEL 3
3100 Description: Number of open files grows over time
3101 Details : The number of open files grows over time, whether or not
3102 Lustre is started. This was due to a filp leak introduced
3103 by one of our kernel patches.
3105 ------------------------------------------------------------------------------
3107 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3109 * Support for kernels:
3110 2.6.9-42.EL (RHEL 4)
3111 2.6.5-7.276 (SLES 9)
3112 2.4.21-40.EL (RHEL 3)
3113 2.6.12.6 vanilla (kernel.org)
3118 Bugzilla : 5719, 9635, 9792, 9684
3119 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3120 Details : If a server is under heavy load and cannot reply to new
3121 connection requests before the client resends the (re)connect,
3122 the connection handling code can behave badly if two service
3123 threads are concurrently handing separate (re)connections from
3124 the same client. Add better locking to the connection handling
3125 code, and ensure that only a single connection will be processed
3126 for a given client UUID, even if the lock is dropped.
3128 Severity : enhancement
3130 Description: add TCP zero-copy support to kernel
3131 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3132 sends if the hardware supports scatter-gather and checksumming.
3133 This allows socklnd to do client-write and server-read more
3134 efficiently and reduce CPU utilization from skbuf copying.
3137 Frequency : only if NFS exporting from client
3139 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3140 Details : If setting attributes on a file created under NFS that had
3141 never been opened it would be possible to oops the client
3142 if the file had no objects.
3145 Frequency : always for liblustre
3147 Description: liblustre client does MDS+OSTs setattr RPC for each write
3148 Details : When doing a write from a liblustre client, the client
3149 incorrectly issued an RPC to the MDS and each OST the file was
3150 striped over in order to update the timestamps. When writing
3151 with small chunks and many clients this could overwhelm the MDS
3152 with RPCs. In all cases it would slow down the write because
3153 these RPCs are unnecessary.
3155 Severity : enhancement
3157 Description: allow number of MDS service threads to be changed at module load
3158 Details : It is now possible to change the number of MDS service threads
3159 running. Adding "options mds mds_num_threads={N}" to the MDS's
3160 /etc/modprobe.conf will set the number of threads for the next
3161 time Lustre is restarted (assuming the "mds" module is also
3162 reloaded at that time). The default number of threads will
3163 stay the same, 32 for most systems.
3168 Description: OST crash if filesystem is unformatted or corrupt
3169 Details : If an OST is started on a device that has never been formatted
3170 or if the filesystem is corrupt and cannot even mount then the
3171 error handling cleanup routines would dereference a NULL pointer.
3176 Description: NULL pointer deref in llap_from_page.
3177 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3178 incorrect) mapping if the page was truncated/reclaimed while it was
3179 searched for. Check for this condition and skip such pages when
3180 doing readahead. Introduce extra check to llap_from_page() to
3181 verify page->mapping->host is non-NULL (so page is not anonymous).
3184 Frequency : Sometimes when using sys_sendfile
3186 Description: "page not covered by a lock" warnings from ll_readpage
3187 Details : sendfile called ll_readpage without right page locks present.
3188 Now we introduced ll_file_sendfile that does necessary locking
3189 around call to generic_file_sendfile() much like we do in
3193 Frequency : with certain MDS communication failures at client mount time
3195 Description: NULL pointer deref after failed client mount
3196 Details : a client connection request may delayed by the network layer
3197 and not be sent until after the PTLRPC layer has timed out the
3198 request. If the client fails the mount immediately it will try
3199 to clean up before the network times out the request. Add a
3200 reference from the request import to the obd device and delay
3201 the cleanup until the network drops the request.
3204 Frequency : occasionally during client (re)connect
3206 Description: assertion failure during client (re)connect
3207 Details : processing a client connection request may be delayed by the
3208 client or server longer than the client connect timeout. This
3209 causes the client to resend the connection request. If the
3210 original connection request is replied in this interval, the
3211 client may trip an assertion failure in ptlrpc_connect_interpret()
3212 which thought it would be the only running connect process.
3215 Frequency : only with obd_echo servers and clients that are rebooted
3217 Description: kernel BUG accessing uninitialized data structure
3218 Details : When running an obd_echo server it did not start the ping_evictor
3219 thread, and when a client was evicted an uninitialized data
3220 structure was accessed. Start the ping_evictor in the RPC
3221 service startup instead of the OBD startup.
3223 Severity : enhancement
3224 Bugzilla : 10193 (patchless)
3225 Description: Remove dependency on various unexported kernel interfaces.
3226 Details : No longer need reparent_to_init, exit_mm, exit_files,
3227 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3230 Frequency : rare (only users of deprecated and unsupported LDAP config)
3232 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3233 Details : LDAP apparently contains 'lustreName' attributes instead of
3234 'name'. A simple remapping of the name is sufficient.
3237 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3239 Description: waiting_locks_callback trips kernel BUG if client is evicted
3240 Details : Running with the dump_on_timeout debug flag turned on makes
3241 it possible that the waiting_locks_callback() can try to dump
3242 the Lustre kernel debug logs from an interrupt handler. Defer
3243 this log dumping to the expired_lock_main() thread.
3245 Severity : enhancement
3247 Description: Support NFS exporting on 2.6 kernels.
3248 Details : Implement non-rawops metadata methods for NFS server to use without
3249 changing NFS server code.
3252 Frequency : very rare (synthetic metadata workload only)
3254 Description: two racing renames might cause an MDS thread to deadlock
3255 Details : Running the "racer" program may cause one MDS thread to rename
3256 a file from being the source of a rename to being the target of
3257 a rename at exactly the same time that another thread is doing
3258 so, and the second thread has already enqueued these locks after
3259 doing a lookup of the target and is trying to relock them in
3260 order. Ensure that we don't try to re-lock the same resource.
3263 Frequency : only very large systems with liblustre clients
3265 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3266 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3267 all OSTs in parallel. This allows the eviction of stale liblustre
3268 clients to proceed much faster than if they were done in series,
3269 and also offers similar improvements for other set_info RPCs.
3274 Description: excessive CPU usage during initial read phase on client
3275 Details : During the initial read phase on a client, it would agressively
3276 retry readahead on the file, consuming too much CPU and impacting
3277 performance (since 1.4.5.8). Improve the readahead algorithm
3278 to avoid this, and also improve some other common cases (read
3279 of small files in particular, where "small" is files smaller than
3280 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3285 Description: MDS crash when receiving packet with unknown intent.
3286 Details : Do not LBUG in unknown intent case, just return -EFAULT
3288 Severity : enhancement
3289 Bugzilla : 9293, 9385
3290 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3291 Details : Do not serialize getattr (non-intent version) and statfs.
3294 Frequency : occasional, when OST network is overloaded/intermittent
3296 Description: client evicted by OST after bulk IO timeout
3297 Details : If a client sends a bulk IO request (read or write) the OST
3298 may evict the client if it is unresposive to its data GET/PUT
3299 request. This is incorrect if the network is overloaded (takes
3300 too long to transfer the RPC data) or dropped the OST GET/PUT
3301 request. There is no need to evict the client at all, since
3302 the pinger and/or lock callbacks will handle this, and the
3303 client can restart the bulk request.
3306 Frequency : Always when mmapping file with no objects
3308 Description: client crashes when mmapping file with no objects
3309 Details : Check that we actually have objects in a file before doing any
3310 operations on objects in ll_vm_open, ll_vm_close and
3316 Description: Request leak when working with deleted CWD
3317 Details : Introduce advanced request refcount tracking for requests
3318 referenced from lustre intent.
3320 Severity : Enhancement
3322 Description: Cache open file handles on client.
3323 Details : MDS now will return special lock along with openhandle, if
3324 requested and client is allowed to hold openhandle, even if unused,
3325 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3326 closing files for every read/write openration.
3328 Severity : Enhancement
3330 Description: Cache open negative dentries on client when possible.
3331 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3332 negative dentries on lock revocation.
3337 Description: Remounting a client read-only wasn't possible with a zconf mount
3338 Details : It wasn't possible to remount a client read-only with llmount.
3340 Severity : enhancement
3341 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3342 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3343 MPICH is included to add Lustre-specific ADIO interfaces.
3344 This is based closely on the UFS ADIO layer and only differs
3345 in file creation, in order to allow the OST striping to be set.
3346 This is user-contributed code and not supported by CFS.
3351 Description: extended inode attributes (immutable, append-only) work improperly
3352 when 2.4 and 2.6 kernels are used on client/server or vice versa
3353 Details : Introduce kernel-independent values for these flags.
3355 Severity : enhancement
3358 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3359 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3360 granularity. Now, Fractional MB granularity is be supported,
3361 this is very useful for low memory system.
3363 Severity : enhancement
3365 Description: Getattr by fid
3366 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3367 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3368 have locks and avoids getattr rpc requests for it on every stat.
3371 Frequency : Always, for filesystems larger than 2TB
3373 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3374 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3375 filesystems larger than 2TB to be mounted with mballoc enabled.
3378 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3380 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3381 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3382 filesystems larger than 2TB is mounted then the kernel will
3383 silently corrupt the start of the filesystem. CONFIG_LBD is
3384 enabled for all CFS-supported kernels, but the possibility of
3385 this happening with a modified kernel config exists.
3387 Severity : enhancement
3389 Description: add client O_DIRECT support for 2.6 kernels
3390 Details : It is now possible to do O_DIRECT reads and writes to files
3391 in the Lustre client mountpoint on 2.6 kernel clients.
3393 Severity : enhancement
3395 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3396 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3397 parallel, not waiting for response from every OST before sending
3398 a rpc to the next OST.
3403 Description: setattr vs write race when updating file timestamps
3404 Details : Client processes that update a file timestamp into the past
3405 right after writing to the file (e.g. tar) it is possible that
3406 the updated file modification time can be reset to the current
3407 time due to a race between processing the setattr and write RPC.
3409 Severity : enhancement
3411 Description: Bring 'lfs find' closer in line with regular Linux find.
3412 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3413 -print0 options and obtains all the needed info through the lustre
3416 Severity : enhancement
3418 Description: support up to 1024 configured devices on one node
3419 Details : change obd_dev array from statically allocated to dynamically
3420 allocated structs as they are first used to reduce memory usage
3425 Description: Flush dirty partially truncated pages during truncate
3426 Details : Immediatelly flush partially truncated pages in filter_setattr,
3427 this way we completely avoid having any pages in page cache on OST
3428 and can retire ugly workarounds during writes to flush such pages.
3433 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3434 Details : protect inode from truncation within vfs_unlink() context
3435 just take a reference before calling vfs_unlink() and release it
3436 when parent's i_sem is free.
3441 Description: last_id value checked outside lock on OST caused LASSERT failure
3442 Details : If there were multiple MDS->OST object precreate requests in
3443 flight, it was possible that the OST's last object id was checked
3444 outside a lock and incorrectly tripped an assertion. Move checks
3445 inside locks, and discard old precreate requests.
3448 Frequency : always, if extents are used on OSTs
3450 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3451 Details : OSTs using the extents format would not zero the high 16 bits of
3452 the index physical block number. This is not a problem for any
3453 OST filesystems smaller than 16TB, and no kernels support ext3
3454 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3455 new/modified files) and can be fixed for existing filesystems
3456 with e2fsprogs-1.39-cfs1.
3461 Description: import connection selection may be incorrect if timer wraps
3462 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3463 import connections to be ignored if the 32-bit jiffies counter
3464 wraps. Use a 64-bit jiffies counter.
3467 Frequency : very large clusters immediately after boot
3469 Description: LNET request buffers exhausted under heavy short-term load
3470 Details : If a large number of client requests are generated on a service
3471 that has previously never seen so many requests it is possible
3472 that the request buffer growth cannot keep up with the spike in
3473 demand. Instead of dropping incoming requests, they are held in
3474 the LND until the RPC service can accept more requests.
3477 Frequency : Sometimes during replay
3479 Description: Assertion failure in ll_local_open after replay.
3480 Details : If replay happened on an open request reply before we were able
3481 to set replay handler, reply will become not swabbed tripping the
3482 assertion in ll_local_open. Now we set the handler right after
3483 recognising of open request
3486 Frequency : very rare
3488 Description: kernel reports "badness in vsnprintf"
3489 Details : Reading from the "recovery_status" /proc file in small chunks
3490 may cause a negative length in lprocfs_obd_rd_recovery_status()
3491 call to vsnprintf() (which is otherwise harmless). Exit early
3492 if there is no more space in the output buffer.
3494 Severity : enhancement
3496 Description: clear OBD RPC statistics by writing to them
3497 Details : It is now possible to clear the OBD RPC statistics by writing
3498 to the "stats" file.
3503 Description: Client mtime is not the same on different clients after utimes
3504 Details : In some cases, the client was using the utimes() syscall on
3505 a file cached on another node. The clients now validate the
3506 ctime from the MDS + OSTs to determine which one is right.
3511 Description: Inability to activate failout mode
3512 Details : lconf script incorrectly assumed that in pythong string's numeric
3513 value is used in comparisons.
3516 Frequency : always with multiple stripes per file
3518 Description: Inefficient object allocation for mutli-stripe files
3519 Details : When selecting which OSTs to stripe files over, for files with
3520 a stripe count that divides evenly into the number of OSTs,
3521 the MDS is always picking the same starting OST for each file.
3522 Return the OST selection heuristic to the original design.
3527 Description: mount failures may take full timeout to return an error
3528 Details : Under some heavy load conditions it is possible that a
3529 failed mount can wait for the full obd_timeout interval,
3530 possibly several minutes, before reporting an error.
3531 Instead return an error as soon as the status is known.
3533 Frequency : quota enabled and large files being deleted
3535 Description: releasing more than 4GB of quota at once hangs OST
3536 Details : If a user deletes more than 4GB of files on a single OST it
3537 will cause the OST to spin in an infinite loop. Release
3538 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3543 Description: statfs data retrieved from /proc may be stale or zero
3544 Details : When reading per-device statfs data from /proc, in the
3545 {kbytes,files}_{total,free,avail} files, it may appear
3546 as zero or be out of date.
3549 Frequency : systems with MD RAID1 external journal devices
3551 Description: lconf's call to blkid is confused by RAID1 journal devices
3552 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3553 of returning all block devices that match the journal UUID.
3556 Frequency : always, for aggregate stripe size over 4GB
3558 Description: assertion fails when trying to use 4GB stripe size
3559 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3560 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3563 Frequency : always on ppc64
3565 Description: the first write on an ext3 filesystem with mballoc got stuck
3566 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3567 perform endianness conversion.
3569 ------------------------------------------------------------------------------
3571 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3573 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3574 INTEROPERATE with older versions automatically. Please read the
3575 user documentation before upgrading any part of a live system.
3576 * WARNING: Lustre networking configuration changes are required with
3577 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3580 * Support for kernels:
3581 2.6.9-22.0.2.EL (RHEL 4)
3582 2.6.5-7.244 (SLES 9)
3583 2.6.12.6 vanilla (kernel.org)
3586 Severity : enhancement
3587 Bugzilla : 7981/8208
3588 Description: Introduced Lustre Networking (LNET)
3589 Details : LNET is new networking infrastructure for Lustre, it includes
3590 a reorganized network configuration mode (see the user
3591 documentation for full details) as well as support for routing
3592 between different network fabrics. Lustre Networking Devices
3593 (LNDs) for the supported network fabrics have also been
3594 created for this new infrastructure.
3596 Severity : enhancement
3597 Description: Introduced Access control lists
3598 Details : clients can set ACLs on files and directories in order to have
3599 more fine-grained permissions than the standard Unix UGO+RWX.
3600 The MDS must be started with the "-o acl" mount option.
3602 Severity : enhancement
3603 Description: Introduced filesystem quotas
3604 Details : Administrators may now establish per-user quotas on the
3607 Severity : enhancement
3609 Description: Configuration change for the XT3
3610 The PTLLND is now used to run Lustre over Portals on the XT3
3611 The configure option(s) --with-cray-portals are no longer used.
3612 Rather --with-portals=<path-to-portals-includes> is used to
3613 enable building on the XT3. In addition to enable XT3 specific
3614 features the option --enable-cray-xt3 must be used.
3619 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3620 Details : A race between allocating a new llcd and re-getting the llcd_lock
3621 allowed another thread to grab newly-allocated llcd.
3623 Severity : enhancement
3625 Description: 2.6 OST async journal commit and locking fix to improve performance
3626 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3627 2.6 kernels are now async as they already were in 2.4 kernels so
3628 that they can commit concurrently with the network bulk transfer.
3629 For block-allocated files the filter allocation semaphore is held
3630 to avoid filesystem fragmentation during allocation. BKL lock
3631 removed for 2.6 xattr operations where it is no longer needed.
3636 Description: lconf incorrectly determined whether two IP networks could talk
3637 Details : In some more complicated routing and multiple-network
3638 configurations, lconf will avoid trying to make a network
3639 connection to a disjoint part of the IP space. It was doing the
3640 math incorrectly for one set of cases.
3645 Description: Fix for potential infinite loop processing records in an llog.
3646 Details : If an llog record is corrupted/zeroed, it is possible to loop
3647 forever in llog_process(). Validate the llog record length
3648 and skip the remainder of the block on error.
3651 Frequency : occasional (liblustre only)
3653 Description: liblustre could not open files whose last component is a symlink
3654 Details : sysio_path_walk() would incorrectly pass the open intent to
3655 intermediate path components.
3658 Frequency : rare (liblustre only with non-standard tuning)
3659 Bugzilla : 7201 (7350)
3660 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3661 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3662 after it has been released, only do this for referenced locks
3664 Severity : enhancement
3666 Description: specifying an (invalid) directory default stripe_size of -1
3667 would reset the directory default striping
3668 Details : stripe_size -1 was used internally to signal directory stripe
3669 removal, now use "all default" to signal dir stripe removal
3670 as a directory striping of "all default" is not useful
3673 Frequency : common for large clusters running liblustre clients
3675 Description: doing an ls when liblustre clients are running is slow
3676 Details : sending a glimpse AST to a liblustre client waits for every AST
3677 to time out, as liblustre clients will not respond. Since they
3678 cannot cache data we refresh the OST lock LVB from disk instead.
3680 Severity : enhancement
3682 Description: doing an ls at the same time as file IO can be slow
3683 Details : enqueue and other "small" requests can be blocked behind many
3684 large IO requests. Create a new OST IO portal for non-IO
3685 requests so they can be processed faster.
3688 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3690 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3691 created file with mode 000
3692 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3693 allow an ftruncate() as a non-root user to a file with mode 000.
3694 Lustre now respects this flag to disable mode checks when
3695 truncating a file owned by the user
3698 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3701 Description: Revoking locks from clients that went dead or catatonic might take
3703 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3704 cancellation of such locks instant on servers without waiting for
3705 any reply from clients. Clients drops these locks when cancel
3706 notification from server is received without replying.
3709 Frequency : liblustre-only, when liblustre client dies or becomes busy
3711 Description: Doing ls on Linux clients can take a long time with active
3713 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3714 granting such locks to it in the first place if possible. Locks
3715 are taken by proxy on the OST during the read or write and
3716 dropped immediately afterward. Add connect flags handling, do
3717 not grant locks to liblustre clients for glimpse ASTs.
3719 Severity : enhancement
3721 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3722 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3723 sequential reads and behaves badly with applications doing random
3724 reads. Improve it by reading ahead at least read region, and
3725 avoiding excessive large RPC for small reads.
3727 Severity : enhancement
3729 Description: Creating more than 1000 files for a single job may cause a load
3730 imbalance on the OSTs if there are also a large number of OSTs.
3731 Details : qos_prep_create() uses an OST index reseed value that is an
3732 even multiple of the number of available OSTs so that if the
3733 reseed happens in the middle of the object allocation it will
3734 still utilize the OSTs as uniformly as possible.
3739 Description: OST or MDS may oops in ping_evictor_main()
3740 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3741 but doesn't restart at beginning of obd_exports_timed list
3744 Severity : enhancement
3746 Description: improve by-nid export eviction on the MDS and OST
3747 Details : allow multiple exports with the same NID to be evicted at one
3748 time without re-searching the exports list.
3751 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3753 Description: MDS may oops in groups_free()
3754 Details : in rare race conditions a newly allocated group_info struct is
3755 freed again, and this can be NULL. The 2.4 compatibility code
3756 for groups_free() checked for a NULL pointer, but 2.6 did not.
3759 Frequency : common for liblustre clients doing little filesystem IO
3760 Bugzilla : 9352, 7313
3761 Description: server may evict liblustre clients accessing contended locks
3762 Details : if a client is granted a lock or receives a completion AST
3763 with a blocking AST already set it would not reply to the AST
3764 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3767 Frequency : lfs setstripe, only systems with more than 160 OSTs
3769 Description: unable to set striping with a starting offset beyond OST 160
3770 Details : llapi_create_file() incorrectly limited the starting stripe
3771 index to the maximum single-file stripe count.
3774 Frequency : LDAP users only
3776 Description: lconf did not handle in-kernel recovery with LDAP properly
3777 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3779 Severity : enhancement
3781 Description: bind OST threads to NUMA nodes to improve performance
3782 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3783 node and do their allocations there to localize memory access
3785 Severity : enhancement
3787 Description: llmount can determine client NID directly from Myrinet (GM)
3788 Details : the client NID code from gmnalnid was moved directly into
3789 llmount, removing the need to use this or specifying the
3790 client NID explicitly when mounting GM clients with zeroconf
3793 Frequency : if client is started with down MDS
3795 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3796 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3797 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3798 ptlrpc_import_delayed_req() because we didn't check if the
3799 request was interrupted, and we also didn't break out of the
3800 event loop if there was no timeout
3805 Description: data loss during non-page-aligned writes to a single file from
3806 both multiple nodes and multiple threads on one node at same time
3807 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3808 inconsistency led to false short-reads, that were cached and later
3809 used by ->prepare_write() to fill in partially written page,
3810 leading to data loss.
3813 Frequency : always, if lconf --abort_recovery used
3815 Description: lconf --abort_recovery fails with 'Operation not supported'
3816 Details : lconf was attempting to abort recovery on the MDT device and not
3819 Severity : enhancement
3821 Description: remove cleanup logs
3822 Details : replace lconf-generated cleanup logs with lustre internal
3823 cleanup routines. Eliminates the need for client-cleanup and
3826 Severity : enhancement
3828 Description: add support for EAs (user and system) on lustre filesystems
3829 Details : it is now possible to store extended attributes in the Lustre
3830 client filesystem, and with the user_xattr mount option it
3831 is possible to allow users to store EAs on their files also
3833 Severity : enhancement
3835 Description: Add possibility (config option) to show minimal available OST free
3837 Details : When compiled with --enable-mindf configure option, statfs(2)
3838 (and so, df) will return least minimal free space available from
3839 all OSTs as amount of free space on FS, instead of summary of
3840 free spaces of all OSTs.
3842 Severity : enhancement
3844 Description: do not expand extent locks acquired on OST-side
3845 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3846 by server: they are not cached anyway.
3849 Frequency : when mmap is used/binaries executed from Lustre
3851 Description: Unmmap pages before throwing them away from read cache.
3852 Details : llap_shrink cache now attempts to unmap pages before discarding
3853 them (if unmapping failed - do not discard). SLES9 kernel has
3854 extra checks that trigger if this unmapping is not done first.
3859 Description: lconf didn't resolve symlinks before checking to see whether a
3860 given mountpoint was already in use
3863 Frequency : when migrating failover services
3864 Bugzilla : 6395, 9514
3865 Description: When migrating a subset of services from a node (e.g. failback
3866 from a failover service node) the remaining services would
3867 time out and evict clients.
3868 Details : lconf --force (implied by --failover) sets the global obd_timeout
3869 to 5 seconds in order to quickly disconnect, but this caused
3870 other RPCs to time out too quickly. Do not change the global
3871 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3873 Severity : enhancement
3874 Frequency : if MDS is started with down OST
3875 Bugzilla : 9439,5706
3876 Description: Allow startup/shutdown of an MDS without depending on the
3877 availability of the OSTs.
3878 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3879 Add appropriate locking and lov-osc refcounts for safe
3880 cleaning. Add osc abort_inflight calls in case the
3881 synchronize never started.
3884 Frequency : occasional (Cray XT3 only)
3886 Description: root not authorized to access files in CRAY_PORTALS environment
3887 Details : The client process capabilities were not honoured on the MDS in
3888 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3889 been accessed by an authorized user then root was able to access
3890 the file on the local client also. The root user capabilities
3891 are now allowed on the MDS, as this environment has secure UID.
3894 Frequency : occasional
3896 Description: ldiskfs "too long searching" message happens too often
3897 Details : A debugging message (otherwise harmless) prints too often on
3898 the OST console. This has been reduced to only happen when
3899 there are fragmentation problems on the filesystem.
3904 Description: Division by zero in statfs when all OSCs are inactive
3905 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3906 lov_statfs divides by value returned by lov_get_stripecnt().
3910 Bugzilla : 9489, 3273
3911 Description: First write from each client to each OST was only 4kB in size,
3912 to initialize client writeback cache, which caused sub-optimal
3913 RPCs and poor layout on disk for the first writen file.
3914 Details : Clients now request an initial cache grant at (re)connect time
3915 and so that they can start streaming writes to the cache right
3916 away and always do full-sized RPCs if there is enough data.
3917 If the OST is rebooted the client also re-establishes its grant
3918 so that client cached writes will be honoured under the grant.
3923 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3924 Details : Now I/O RPCs go to different portal number and (presumably) fast
3925 lock requests (and glimses) and other RPCs get their own service
3926 threads pool that should be able to service those RPCs
3929 Severity : enhancement
3931 Description: Ability to exchange lustre version between client and servers and
3932 issue warnings at client side if client is too old. Also for
3933 liblustre clients there is ability to refuse connection of too old
3935 Details : New 'version' field is added to connect data structure that is
3936 filled with version info. That info is later checked by server and
3940 Frequency : rare, liblustre only.
3941 Bugzilla : 9296, 9581
3942 Description: Two simultaneous writes from liblustre at offset within same page
3943 might proceed at the same time overwriting eachother with stale
3945 Details : I/O lock withing llu_file_prwv was released too early, before data
3946 actually was hitting the wire. Extended lock-holding time until
3947 server acknowledges receiving data.
3950 Frequency : extremely rare. Never observed in practice.
3952 Description: avoid generating lustre_handle cookie of 0.
3953 Details : class_handle_hash() generates handle cookies by incrementing
3954 global counter, and can hit 0 occasionaly (this is unlikely, but
3955 not impossible, because initial value of cookie counter is
3956 selected randonly). Value of 0 is used as a sentinel meaning
3957 "unassigned handle" --- avoid it. Also coalesce two critical
3958 sections in this function into one.
3960 Severity : enhancement
3962 Description: allow liblustre clients to delegate truncate locking to OST
3963 Details : To avoid overhead of locking, liblustre client instructs OST to
3964 take extent lock in ost_punch() on client's behalf. New connection
3965 flag is added to handle backward compatibility.
3967 Severity : enhancement
3968 Bugzilla : 4928, 7341, 9758
3969 Description: allow number of OST service threads to be specified
3970 Details : a module parameter allows the number of OST service threads
3971 to be specified via "options ost ost_num_threads={N}" in the
3972 OSS's /etc/modules.conf or /etc/modprobe.conf.
3976 Bugzilla : 6146, 9635, 9895
3977 Description: servers crash with bad pointer in target_handle_connect()
3978 Details : In rare cases when a client is reconnecting it was possible that
3979 the connection request was the last reference for that export.
3980 We would temporarily drop the export reference and get a new
3981 one, but this may have been the last reference and the export
3982 was just destroyed. Get new reference before dropping old one.
3984 Severity : enhancement
3985 Frequency : if client is started with failover MDS
3987 Description: Allow multiple MDS hostnames in the mount command
3988 Details : Try to read the configuration from all specified MDS
3989 hostnames during a client mount in case the "primary"
3992 Severity : enhancement
3994 Description: Stop sending data to evicted clients as soon as possible.
3995 Details : Check if the client we are about to send or are sending data to
3996 was evicted already. (Check is done every second of waiting,
3997 for which l_wait_event interface was extended to allow checking
3998 of exit condition at specified intervals).
4001 Frequency : rare, normally only when NFS exporting is done from client
4003 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4005 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4008 Frequency : always, if srand() is called before liblustre initialization
4010 Description: Liblustre uses system PRNG disturbing its usage by user application
4011 Details : Introduce internal to lustre fast and high-quality PRNG for
4012 lustre usage and make liblustre and some other places in generic
4013 lustre code to use it.
4015 Severity : enhancement
4016 Bugzilla : 9477, 9557, 9870
4017 Description: Verify that the MDS configuration logs are updated when xml is
4018 Details : Check if the .xml configuration logs are newer than the config
4019 logs stored on the MDS and report an error if this is the case.
4020 Request --write-conf, or allow starting with --old_conf.
4022 Severity : enhancement
4024 Description: Handle symlinks in the path when checking if Lustre is mounted.
4025 Details : Resolve intermediate symlinks when checking if a client has
4026 mounted a filesystem to avoid duplicate client mounts.
4031 Description: lconf can hit an error exception but still return success.
4032 Details : The lconf command catches the Command error exception at the top
4033 level script context and will exit with the associated exit
4034 status, but doesn't ensure that this exit status is non-zero.
4039 Description: failure of ptlrpc thread startup can cause oops
4040 Details : Starting a ptlrpc service thread can fail if there are a large
4041 number of threads or the server memory is very fragmented.
4042 Handle this without oopsing.
4045 Frequency : always, only if liblustre and non-default acceptor port was used
4047 Description: liblustre cannot connect to servers with non-default acceptor port
4048 Details : tcpnal_set_default_params() was not called and was therefore
4049 ignoring the environment varaible TCPNAL_PORT, as well as other
4050 TCPNAL_ environment variables
4055 Description: two objects could be created on the same OST for a single file
4056 Details : If an OST is down, in some cases it was possible to create two
4057 objects on a single OST for a single file. No problems other
4058 than potential performance impact and spurious error messages.
4062 Bugzilla : 5681, 9562
4063 Description: Client may oops in ll_unhash_aliases
4064 Details : Client dcache may become inconsistent in race condition.
4065 In some cases "getcwd" can fail if the current directory is
4071 Description: Inode refcounting problems in NFS export code
4072 Details : link_raw functions used to call d_instantiate without obtaining
4073 extra inode reference first.
4077 Bugzilla : 9942, 9903
4078 Description: Referencing freed requests leading to crash, memleaks with NFS.
4079 Details : We used to require that call to ll_revalidate_it was always
4080 followed by ll_lookup_it. Also with revalidate_special() it is
4081 possible to call ll_revalidate_it() twice for the same dentry
4082 even if first occurence returned success. This fix changes semantic
4083 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4084 reference on a request referred from the intent.
4085 ll_intent_release() then releases such a request.
4088 Frequency : rare, normally benchmark loads only
4090 Description: unlinked inodes were kept in memory on the client
4091 Details : If a client is repeatedly creating and unlinking files it
4092 can accumulate a lot of stale inodes in the inode slab cache.
4093 If there is no other client load running this can cause the
4094 client node to run out of memory. Instead flush old inodes
4095 from client cache that have the same inode number as a new inode.
4098 Frequency : SLES9 2.6.5 kernel and long filenames only
4099 Bugzilla : 9969, 10379
4100 Description: utime reports stale NFS file handle
4101 Details : SLES9 uses out-of-dentry names in some cases, which confused
4102 the lustre dentry revalidation. Change it to always use the
4106 Frequency : rare, unless heavy write-truncate concurrency is continuous
4107 Bugzilla : 4180, 6984, 7171, 9963, 9331
4108 Description: OST becomes very slow and/or deadlocked during object unlink
4109 Details : filter_destroy() was holding onto the parent directory lock
4110 while truncating+unlinking objects. For very large objects this
4111 may block other threads for a long time and slow overall OST
4112 responsiveness. It may also be possible to get a lock ordering
4113 deadlock in this case, or run out of journal credits because of
4114 the combined truncate+unlink. Solution is to do object truncate
4115 first in one transaction without parent lock, and then do the
4116 final unlink in a new transaction with the parent lock. This
4117 reduces the lock hold time dramatically.
4120 Frequency : rare, 2.4 kernels only
4122 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4123 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4124 before unmounting their filesystem in order to avoid deadlock.
4125 The kernel_locked() function in 2.4 kernels only checks whether
4126 the kernel lock is held, not whether it is this process that is
4127 holding it as 2.6 kernels do.
4132 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4133 Details : The client ptlrpc code may be trying to reconnect to a down
4134 server before a previous connection attempt has timed out.
4135 Increase the reconnect interval to be longer than the connection
4136 timeout interval to avoid sending duplicate connections to
4140 Frequency : echo_client brw_test command
4142 Description: fix echo_client to work with OST preallocated code
4143 Details : OST preallocation code (5137) didn't take echo_client IO path
4144 into account: echo_client calls filter methods outside of any
4145 OST thread and, hence, there is no per-thread preallocated
4146 pages and buffers to use. Solution: hijack pga pages for IO. As
4147 a byproduct, this avoids unnecessary data copying.
4151 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4152 Description: Client can oops in mdc_commit_close() after open replay
4153 Details : It was possible for the MDS to return an open request with no
4154 transaction number in mds_finish_transno() if the client was
4155 evicted, but without actually returning an error. Clients
4156 would later try to replay that open and may trip an assertion
4157 Simplify the client close codepath, and always return an error
4158 from the MDS in case the open is not successful.
4161 Frequency : rare, 2.6 OSTs only
4163 Description: OST may deadlock under high load on fragmented files
4164 Details : If there was a heavy load and highly-fragmented OST filesystems
4165 it was possible to have all the OST threads deadlock waiting on
4166 allocation of biovecs, because the biovecs were not released
4167 until the entire RPC IO was completed. Instead, release biovecs
4168 as soon as they are complete to ensure forward IO progress.
4170 Severity : enhancement
4172 Description: Support for specifying external journal device at mount
4173 Details : If an OST or MDS device is formatted with an external journal
4174 device, this device major/minor is stored in the ext3 superblock
4175 and may not be valid for failover. Allow detecting and
4176 specifying the external journal at mount time.
4181 Description: Mounting an MDS with pending unlinked files may cause oops
4182 Details : target_finish_recovery() calls mds_postrecov() which returned
4183 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4184 considers this an error and immediately begins cleaning up the
4185 lov, just after starting the mds_lov process
4187 Severity : enhancement
4189 Description: Implement 'lfs df' to report actual free space on per-OST basis
4190 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4191 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4192 usage of objects; '-h' to report in human readable format.
4194 ------------------------------------------------------------------------------
4196 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4203 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4204 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4205 at startup during orphan unlinking. Instead, skip buddy update
4206 before setup, as it will be regenerated after recovery is complete.
4211 Description: If an OST is inactive, its locks might reference stale inodes.
4212 Details : lov_change_cbdata() must iterate over all namespaces, even if
4213 they are inactive to clear inode references from the lock.
4215 Severity : enhancement
4216 Frequency : occasional, if non-standard max_dirty_mb used
4218 Description: Client will block write RPCs if not enough grant
4219 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4220 then the client will block writes while waiting for another RPC
4221 to complete instead of consuming its dirty limit. With change
4222 we get improved performance when max_dirty_mb is small.
4224 Severity : enhancement
4225 Bugzilla : 3389, 6253
4226 Description: Add support for supplementary groups on the MDS.
4227 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4228 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4229 lookups for user supplementary groups into a cache.
4233 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4234 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4235 for liblustre O_EXCL is enforced by MDS.
4239 Description: readdir never returns NULL in liblustre.
4240 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4241 dentry in struct dirent.
4245 Frequency : liblustre only, depends on application IO pattern
4246 Description: liblustre clients evicted if not contacting servers
4247 Details : Don't put liblustre clients into the ping_evictor list, so
4248 they will not be evicted by the pinger ever.
4250 Severity : enhancement
4252 Description: Add ability to evict clients by NID from MDS.
4253 Details : By echoing "nid:$NID" string into
4254 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4255 $NID would be instantly evicted from this MDS and from all active
4256 OSTs connected to it.
4260 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4261 Details : lookup_it_finish() used to query file size from OSTs that was not
4266 Description: service threads change working directory to that of init
4267 Details : Starting lustre service threads may pin the working directory
4268 of the parent thread, making that filesystem busy. Threads
4269 now change to the working directory of init to avoid this.
4273 Frequency : during shutdown only
4274 Description: shutdown with a failed MDS or OST can cause unmount to hang
4275 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4280 Frequency : 2.6 only
4281 Description: chmod/chown may include an extra supplementary group
4282 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4283 supplementary group and if none is specified this is used.
4286 Bugzilla : 5479 (6816)
4288 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4289 Details : If lookup is in progress on a file that is unlinked we might try
4290 to revalidate the inode and fail in revalidate after lookup is
4291 complete and ll_file_open() enqueues the open again but
4292 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4295 Frequency : always, if lconf --abort_recovery used
4297 Description: lconf --abort_recovery fails with 'Operation not supported'
4298 Details : lconf was attempting to abort recovery on the MDT device and not
4301 ------------------------------------------------------------------------------
4303 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4308 Frequency : rare (only unsupported configurations with a node running as an
4310 Bugzilla : 6514, 5137
4311 Description: Mounting a Lustre file system on a node running as an OST could
4313 Details : OSTs now preallocates memory needed to write out data at
4314 startup, instead of when needed, to avoid having to
4315 allocate memory in possibly low memory situations.
4316 Specifically, if the file system is mounted on on OST,
4317 memory pressure could force it to try to write out data,
4318 which it needed to allocate memory to do. Due to the low
4319 memory, it would be unable to do so and the node would
4320 become unresponsive.
4322 Severity : enhancement
4324 Description: Addition of lconf --service command line option
4325 Details : lconf now accepts a '--service <arg>' option, which is
4326 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4328 Severity : enhancement
4330 Description: Failover mode is now the default for OSTs.
4331 Details : By default, OSTs will now run in failover mode. To return to
4332 the old behaviour, add '--failout' to the lmc line for OSTs.
4334 Severity : enhancement
4336 Description: Health checks are now provided for MDS and OSTs
4337 Details : Additional detailed health check information on MSD and OSTs
4338 is now provided through the procfs health_check value.
4341 Frequency : occasional, depends on IO load
4343 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4344 after numerous create/delete cycles
4345 Details : The ext3 inode allocation policy would not allocate new inodes
4346 very well on the OSTs because there are no new directories
4347 being created. Instead we look for groups with free space if
4348 the parent directories are nearly full.
4353 Description: Network or server problems during mount may cause partially
4354 mounted clients instead of returning an error.
4355 Details : The config llog parsing code may overwrite the error return
4356 code during mount error handling, returning success instead
4362 Description: MDS can fail to allocate large reply buffers
4363 Details : After long uptimes the MDS can fail to allocate large reply
4364 buffers (e.g. zconf client mount config records) due to memory
4365 fragmentation or consumption by the buffer cache. Preallocate
4366 some large reply buffers so that these replies can be sent even
4367 under memory pressure.
4371 Frequency : rare (liblustre)
4372 Description: fsx running with liblustre complained that using truncate() to
4373 extend the file doesn't work. This patch corrects that issue.
4374 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4375 ATTR_SIZE and lsm use in llu_setattr_raw.
4379 Frequency : rare, only 2.6 kernels
4380 Description: Unusual file access patterns on the MDS may result in inode
4381 data being lost in very rare circumstances.
4382 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4383 code in the 2.6 kernel caused the inode and/or EA data not to
4384 be read from disk, causing single-file corruption.
4388 Frequency : rare, only 2.6 filesystems using extents
4389 Description: Heavy concurrent write and delete load may cause data corruption.
4390 Details : It was possible under high-load situations to have an extent
4391 metadata block in the block device cache from a just-unlinked
4392 file overwrite a newly-allocated data block. We now unmap any
4393 metadata buffers that alias just-allocated data blocks.
4397 Frequency : filesystems with default stripe_count larger than 77
4398 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4399 Details : lconf specifies an inode size of 4096 bytes when the default
4400 stripe_count is larger than 77. This conflicts with the default
4401 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4402 this case to avoid pinning too much memory for large EAs.
4404 ------------------------------------------------------------------------------
4406 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4411 Frequency : rare (extremely heavy IO load with hundreds of clients)
4413 Description: Client is evicted, gets IO error writing to file
4414 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4415 caused clients to be evicted for AST timeouts. The fixes in
4416 bug 5192 mean we no longer need to have such short AST timeouts
4417 so ldlm_timeout has been increased.
4420 Frequency : occasional during --force or --failover shutdown under load
4421 Bugzilla : 5949, 4834
4422 Description: Server oops/LBUG if stopped with --force or --failover under load
4423 Details : a collection of import/export refcount and cleanup ordering
4424 issues fixed for safer force cleanup
4427 Frequency : only filesystems larger than 120 OSTs
4428 Bugzilla : 5990, 6223
4429 Description: lfs getstripe would oops on a very large filesystem
4430 Details : lov_getconfig used kfree on vmalloc'd memory
4433 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4434 Bugzilla : 6242, 6243
4435 Description: reading from files that had been truncated to a non-zero size
4436 but never opened returned no data
4437 Details : ll_file_read() reads zeros from no-object files to EOF
4442 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4443 Details : There's little harm in aborting MDS/OSS recovery and letting it
4444 try again, so I removed the LASSERT and return an error instead.
4446 Severity : enhancement
4448 Description: New debugging infrastructure for tracking down data corruption
4449 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4450 (b) cover more of the client-side code path, and (c) try to narrow
4451 down where problems occurred
4455 Bugzilla : 3819, 4364, 4397, 6313
4456 Description: Racing close and eviction MDS could cause assertion in mds_close
4457 Details : It was possible to get multiple mfd references during close and
4458 client eviction, leading to one thread referencing a freed mfd.
4460 Severity: : enhancement
4461 Bugzilla : 3262, 6359
4462 Description: Attempts to reconnect to servers are now more aggressive.
4463 Details : This builds on the enhanced upcall-less recovery that was added
4464 in 1.4.2. When trying to reconnect to servers, clients will
4465 now try each server in the failover group every 10 seconds. By
4466 default, clients would previously try one server every 25 seconds.
4471 Description: After recovery, certain operations trigger a failed
4472 assertion on a client.
4473 Details : Failing over an mds, using lconf -d --failover, while a
4474 client was doing a readdir() call would cause the client to
4475 LBUG after recovery completed and the readdir() was resent.
4477 Severity : enhancement
4479 Description: Default groups are now added by lconf
4480 Details : You can now run lconf --group <servicename> without having to
4481 manually add groups with lmc.
4484 Frequency : occasional
4486 Description: Nodes with an elan id of 0 trigger a failed assertion
4489 Frequency : always when accessing e.g. tty/console device nodes
4491 Description: tty and some other devices nodes cannot be used on lustre
4492 Details : file's private_data field is used by device data and lustre
4493 values in there got lost. New field was added to struct file to
4494 store fs-specific private data.
4497 Frequency : when exporting Lustre via NFS
4499 Description: NFSD failed occasionally when looking up a path component
4500 Details : NFSD is looking up ".." which was broken in ext3 directories
4501 that had grown large enough to become hashed.
4504 Frequency : Clusters with multiple interfaces not on the same subnet
4506 Description: Nodes will repeatedly try to reconnect to an interface which it
4507 cannot reach and report an error to the log.
4508 Details : Extra peer list entries will be created by lconf with some peers
4509 unreachable. lconf now validates the peer before adding it.
4512 Frequency : Only if a default stripe is set on the filesystem root.
4514 Description: Setting a default stripe on the filesystem root prevented the
4515 filesystem from being remounted.
4516 Details : The client was sending extra request flags in the root getattr
4517 request and did not allocate a reply buffer for the dir EA.
4520 Frequency : occasional, higher if lots of files are accessed by one client
4521 Bugzilla : 6159, 6097
4522 Description: Client trips assertion regarding lsm mismatch/magic
4523 Details : While revalidating inodes the VFS looks up inodes with ifind()
4524 and in rare cases can find an inode that is being freed.
4525 The ll_test_inode() code will free the lsm during ifind()
4526 when it finds an existing inode and then the VFS later attaches
4527 this free lsm to a new inode.
4531 Bugzilla : 6422, 7030
4532 Description: MDS deadlock between mkdir and client eviction
4533 Details : Creating a new file via mkdir or mknod (starting a transaction
4534 and getting the ns lock) can deadlock with client eviction
4535 (gets ns lock and trying to finish a synchronous transaction).
4538 Frequency : occasional
4539 Description: While starting a server, the fsfilt_ext3 module could not be
4541 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4542 kernels. Previously, lconf would still use the ext3 name
4543 when trying to load modules. Now, it will correctly use
4544 ext3 on 2.4 and ldiskfs on 2.6.
4546 Severity : enhancement
4547 Description: The default stripe count has been changed to 1
4548 Details : The interpretation of the default stripe count (0, to lfs
4549 or lmc) has been changed to mean striping across a single
4550 OST, rather than all available. For general usage we have
4551 found a stripe count of 1 or 2 works best.
4553 Severity : enhancement
4554 Description: Add support for compiling against Cray portals.
4555 Details : Conditional compiling for some areas that are different
4559 Frequency : occasional
4560 Bugzilla : 6409, 6834
4561 Description: Creating files with an explicit stripe count may lead to
4562 a failed assertion on the MDS
4563 Details : If some OSTs are full or unavailable, creating files may
4564 trigger a failed assertion on the MDS. Now, Lustre will
4565 try to use other servers or return an error to the
4569 Frequency : occasional
4571 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4572 caused lock timeout during readahead (since 1.4.2).
4573 Details : Processes doing readahead might match a lock that hasn't been
4574 granted yet if there are overlapping and conflicting lock
4575 requests. The readahead process waits on ungranted lock
4576 (original lock is CBPENDING), while OST waits for that process
4577 to cancel CBPENDING read lock and eventually evicts client.
4579 Severity : enhancement
4581 Description: Initial enabling of flock support for clients
4582 Details : Implements fcntl advisory locking and file status functions.
4583 This feature is provided as an optional mount flag (default
4584 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4585 locking are implemented yet, and those that are are not guaranteed
4586 to be completely correct in production environments.
4587 mount -t lustre -o [flock|noflock] ...
4590 Frequency : occasional
4592 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4593 assertion in the ext3 JBD (journaling) layer.
4594 Details : The b_committed_data struct is protected by the big kernel lock
4595 in 2.4 kernels, serializing journal_commit_transaction() and
4596 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4597 access to this struct. In 2.6 kernels there is finer grained
4598 locking to improve SMP performance of the JBD layer.
4602 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4607 Description: merge and cleanup kernel patches.
4610 -----------------------------------------------------------------------------
4612 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4614 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4615 versions of Lustre. Please update all servers and clients to
4616 version 1.4.2 or later at the same time. You must also run
4617 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4618 to update the configuration logs.
4620 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4621 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4622 - swab llog records of type '0' so we get proper header size/idx (5861)
4623 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4624 - fix rename of one directory over another leaking an inode (5953)
4625 - avoid SetPageDirty on 2.6 (5981)
4626 - don't re-add just-being-destroyed locks to the waiting list (5653)
4627 - when creating new directories, inherit the parent's custom
4628 striping settings if present parent (3048)
4629 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4630 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4631 - don't hold client locks on temporary worklist from l_lru (5666)
4632 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4633 - automatically evict dead clients (5921)
4634 - Update file size properly in create+truncate+fstat case (6196)
4635 - Do not unhash mountpoint dentries, do not allow removal of
4637 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4638 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4639 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4640 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4641 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4642 - Correctly handle reads of files with no objects (6243)
4643 - lctl recover will also mark a device active if deactivate used (5933)
4645 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4646 - allow --write-conf on an MDS with different nettype than client (5619)
4647 - don't write config llogs to MDS for mounts not from that MDS (5617)
4648 - lconf should create multiple TCP connections from a client (5201)
4649 - init scripts are now turned off by default; run chkconfig --on
4650 lustre and chkconfig --on lustrefs to use them
4651 - upcalls are no longer needed for clients to recover to failover
4653 - add --abort-recovery option to lconf to abort recovery on device
4655 - add support for an arbitrary number of OSTs (3026)
4656 - Quota support protocol changes.
4657 - forward compatibility changes to wire structs (6007)
4658 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4659 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4660 - improved "lustre status" script
4661 - initialize blocksize for non-regular files (6062)
4662 - added --disable-server and --disable-client configure options (5782)
4663 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4664 - Vanilla 2.4.29 support
4665 - increase maximum number of obd devices to 520 (6242)
4666 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4667 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4668 SLES 9 2.6.5 kernels
4669 - limit stripes per file to 160 (the maximum EA size) (6093)
4671 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4674 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4675 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4676 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4677 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4678 - small ext3 extents cleanups and fixes (5733)
4679 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4680 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4681 - avoid needless client->OST connect, fix handle mismatch (5317)
4682 - fix DLM error path that led to out-of-sync client, long delays (5779)
4683 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4684 - fix several locking issues related to i_size (5492,5624,5654,5672)
4685 - don't move pending lock onto export if it is already evicted (5683)
4686 - fix kernel oops when creating .foo in unlinked directory (5548)
4687 - fix deadlock in obdfilter statistics vs. object create (5811)
4688 - use time_{before,after} to avoid timer jiffies wrap (5882)
4689 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4690 - Do not leak request if server was not able to process it (5154)
4691 - If mds_open unable to find parent dir, make that negative lookup(5154)
4692 - don't create new directories with extent-mapping (5909, 5936)
4694 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4695 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4696 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4697 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4698 - support for RHEL4 kernel on i686 (5773)
4699 - provide error messages when incompatible logs are encountered (5898)
4701 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4702 * version 1.4.0.10 (1.4.1 release candidate 1)
4704 - don't keep a lock reference when lock is not granted (4238)
4705 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4706 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4707 - drop import inflight refcount on signal_completed_replay error (5255)
4708 - unlock page after async write error during send (3677)
4709 - handle missing objects in filter_preprw_read properly (5265)
4710 - no transno return for symlink open, don't save no-trasno open (3440)
4711 - don't try to complete elan receive that already failed (4012)
4712 - free RPC server reply state on error (5406)
4713 - clean up thread from ptlrpc_start_thread() on error (5160)
4714 - readahead could read extra page into cache that wasn't ejected (5388)
4715 - prevent races in class_attach/setup/cleanup/detach (5260)
4716 - don't dereference de->d_inode after l_dput of de (5458)
4717 - use "int" for stripe value returned from lock_to_stripe (5544)
4718 - mballoc allocation and error-checking fixes in 2.6 (5504)
4719 - block device patches to fix I/O request sizes in 2.6 (5482)
4720 - look up hostnames for IB nals (5602)
4721 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4722 - don't start multiple acceptors for the same port (5277)
4723 - fix incorrect LASSERT in mds_getattr_name (5635)
4724 - export a proc file for general "ping" checking (5628)
4725 - fix "lfs check" to not block when the MDS is down (5628)
4727 - service request history (4965)
4728 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4729 - create an "evict_client" /proc entry on OSTs, like the MDS has
4730 - fix mount usage message, return errors per mount(8) (5168)
4731 - change grep [] to grep "[]" in tests so they work in more UMLs
4732 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4733 - remove ancient LOV_MAGIC_V0 EA support (5047)
4734 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4735 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4736 - allow readahead allocations to fail when low on memory (5383)
4737 - mmap locking landed again, after considerable improvement (2828)
4738 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4740 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4743 - send OST transaction number in read/write reply to free req (4966)
4744 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4745 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4746 - the watchdog thread now runs as interruptible (5246)
4747 - flock/lockf fixes (but it's still disabled, pending 5135)
4748 - don't use EXT3 constants in llite code (5094)
4749 - memory shortage at startup could cause assertion (5176)
4751 - reorganization of lov code
4752 - single portals codebase
4754 - add extents/mballoc support (5025)
4755 - direct I/O reads in the obdfilter (4048)
4756 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4758 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4761 - send OST transaction number in read/write reply to free req (4966)
4762 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4763 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4764 - the watchdog thread now runs as interruptible (5246)
4765 - handle missing objects in filter_preprw_read properly (5265)
4766 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4767 - drop import inflight refcount on signal_completed_replay error (5255)
4768 - unlock page after async write error during send (3677)
4769 - return original error code on reconstructed replies (3761)
4770 - no transno return for symlink open, don't save no-trasno open (3440)
4772 - add pid to ldlm debugging output (4922)
4773 - bump the watchdog timeouts -- we can't handle 30sec yet
4774 - extra debugging for orphan dentry/inode bug (5259)
4776 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4779 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4780 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4781 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4782 - fix NULL dereference in /proc/sys/portals/routes (4827)
4783 - allow failed mdc_close() operations to be interrupted (4561)
4784 - stop precreate on OST before MDS would time out on it (4778)
4785 - don't send partial-page writes before EOF from client (4410)
4786 - discard client grant for sub-page writes on large-page clients (4520)
4787 - don't free dentries not owned by NFS code, check generation (4806)
4788 - fix lsm leak if mds_create_objects() fails (4801)
4789 - limit debug_daemon file size, always print CERROR messages (4789)
4790 - use transno after validating reply (3892)
4791 - process timed out requests if import state changes (3754)
4792 - update mtime on OST during writes, return in glimpse (4829)
4793 - add mkfsoptions to LDAP (4679)
4794 - use ->max_readahead method instead of zapping global ra (5039)
4795 - don't interrupt __l_wait_event() during strace
4797 - add software watchdogs to catch hung threads quickly (4941)
4798 - make lustrefs init script start after nfs is mounted
4799 - fix CWARN/ERROR duplication (4930)
4800 - return async write errors to application if possible (2248)
4801 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4802 - print NAL number in %x format (4645)
4803 - update barely-supported suse-2.4.21-171 series (4842)
4804 - support for sles 9 %post scripts
4805 - support for building 2.6 kernel-source packages
4806 - support for sles km_* packages
4808 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4811 - ignore -ENOENT errors in osc_destroy (3639)
4812 - notify osc create thread that OSC is being cleaned up (4600)
4813 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4814 - reconstruct ost_handle() like mds_handle() (4657)
4815 - create a new thread to do import eviction to avoid deadlock (3969)
4816 - let lconf resolve symlinked-to devices (4629)
4817 - don't unlink "objects" from directory with default EA (4554)
4818 - hold socknal file ref over connect in case target is down (4394)
4819 - allow more than 32000 subdirectories in a single directory (3244)
4820 - fix blocks count for O_DIRECT writes (3751)
4821 - OST returns ENOSPC from object create when no space left (4539)
4822 - don't send truncate RPC if file size isn't changing (4410)
4823 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4824 - bind to privileged port in socknal and tcpnal (3689)
4826 - rate limit CERROR/CWARN console message to avoid overload (4519)
4827 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4828 - basic mmap support (3918)
4829 - kernel patch series update from b1_4 (4711)
4831 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4834 - avoid crash during MDS cleanup with OST shut down (2775)
4835 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4836 - don't use bad inodes on the MDS (3744)
4837 - dynamic object preallocation to improve recovery speed (4236)
4838 - don't hold spinlock over lock dumping or change debug flags (4401)
4839 - don't zero obd_dev when it is force cleaned (3651)
4840 - print grants to console if they go negative (4431)
4841 - "lctl deactivate" will stop automatic recovery attempts (3406)
4842 - look for existing locks in ldlm_handle_enqueue() (3764)
4843 - don't resolve lock handle twice in recovery avoiding race (4401)
4844 - revalidate should check working dir is a directory (4134)
4846 - don't always mark "slow" obdfilter messages as errors (4418)
4848 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4851 - don't close LustreDB during write_conf until it is done (3860)
4852 - fix typo in lconf for_each_profile (3821)
4853 - allow dumping logs from multiple threads at one time (3820)
4854 - don't allow multiple threads in OSC recovery (3812)
4855 - fix debug_size parameters (3864)
4856 - fix mds_postrecov to initialize import for llog ctxt (3121)
4857 - replace config semaphore with spinlock (3306)
4858 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4859 - don't allow enqueue to complete on a destroyed export (3822)
4860 - down write_lock before checking llog header bitmap (3825)
4861 - recover from lock replay timeout (3764)
4862 - up llog sem before sending rpc (3652)
4863 - reduce ns lock hold times when setting kms (3267)
4864 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4865 - fix NULL deref and obd_dev leak on setup error (3312)
4866 - replace some LBUG about llog ops with error handling (3841)
4867 - don't match INVALID dentries from d_lookup and spin (3784)
4868 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4869 - fix invalid assertion in ptlrpc_set_wait (3880)
4871 - add libwrap support for the TCP acceptor (3996)
4872 - add /proc/sys/portals/routes for non-root route listing (3994)
4873 - allow setting MDS UUID in .xml (2580)
4874 - print the stack of a process that LBUGs (4228)
4876 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4879 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4880 - make sure to unset replay flag from failed open requests (3440)
4881 - if default stripe count is 0, use OST count for inode size (3636)
4882 - update parent mtime/ctime on client for create/unlink (2611)
4883 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4884 - free recovery state on server during a forced cleanup (3571)
4885 - unregister_reply for resent reqs (3063)
4886 - loop back devices mounting and status check on 2.6 (3563)
4887 - fix resource-creation race that can provoke i_size == 0 (3513)
4888 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4889 - more debugging for page-accounting assertion (3746)
4890 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4891 - avoid deadlock after precreation failure (3758)
4892 - fix race and lock order deadlock in orphan handling (3450, 3750)
4893 - add validity checks when grabbing inodes from l_ast_data (3599)
4895 - add /proc/.../recovery_status to obdfilter (3428)
4896 - lightweight CDEBUG infrastructure, debug daemon (3668)
4897 - change default OSC RPC parameters to be better on small clusters
4898 - turn off OST read cache for files smaller than 32MB
4899 - install man pages and include them in rpms (3100)
4900 - add new init script for (un)mounting lustre filesystems (2593)
4901 - run chkconfig in %post for init scripts (3701)
4902 - drop scimac NAL (unmaintained)
4904 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4907 - clean kiobufs before and after use (3485)
4908 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4909 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4910 - change init script to fail more clearly if not run as root (1528)
4911 - allow clients to reconnect during replay (1742)
4912 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4913 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4914 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4915 - fix cond_resched (3554)
4916 - only evict unfinished clients after recovery (3515)
4917 - allow bulk resend, prevent data loss (3570)
4918 - dynamic ptlrpc request buffer allocation (2102)
4919 - don't allow unlinking open directory if it isn't empty (2904)
4920 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4921 - remove extraneous obd dereference causing LASSERT failure (3334)
4922 - don't use get_cycles() when creating temp. files on the mds (3156)
4923 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4924 - handle EEXIST for set-stripe, set proper directory name (3336)
4926 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4927 - fix ksocknal_fmb_callback() error messages (2918)
4929 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4932 - don't copy lvb into (possibly NULL) reply on error (2983)
4933 - don't deref dentry after dput, don't free lvb on error (2922)
4934 - use the kms to determine writeback rpc length (2947)
4935 - increment oti_logcookies when osc is inactive (2948)
4936 - update client's i_blocks count via lvb messages (2543)
4937 - handle intent open/close of special files properly (1557)
4938 - mount MDS with errors=remount-ro, like obdfilter (2009)
4939 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4940 - don't use cancelling-locks' kms values (2947)
4941 - use highest lock extent for kms, not last one (2925)
4942 - don't dereference ERR_PTR() dentry in error handling path (3107)
4943 - fix thread race in portals_debug_dumplog() (3122)
4944 - create lprocfs device entries at setup instead of at attach (1519)
4945 - common AST error handler, don't evict client on completion race (3145)
4946 - zero nameidata in detach_mnt in 2.6 (3118)
4947 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4948 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4949 - initialize RPC timeout timer earlier for 2.6 (3219)
4950 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4951 - print nal/nid for unknown nid (3258)
4952 - additional checks for oscc recovery before doing precreate (3284)
4953 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4954 - don't crash in mdc_close for bad permissions on open (3285)
4955 - zero i_rdev for non-device files (3147)
4956 - clear page->private before handing to FS, better assertion (3119)
4957 - tune the read pipeline (3236)
4958 - fix incorrect decref of invalidated dentry (2350)
4959 - provide read-ahead stats and refine rpc in flight stats (3328)
4960 - don't hold journal transaction open across create RPC (3313)
4961 - update atime on MDS at close time (3265)
4962 - close LDAP connection when recovering to avoid server load (3315)
4963 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4964 - don't leak open file on MDS after open resend (3325)
4965 - serialize filter_precreate and filter_destroy_precreated (3329)
4966 - loop device shouldn't call sync_dev() for nul device (3092)
4967 - clear page cache after eviction (2766)
4968 - resynchronize MDS->OST in background (2824)
4969 - refuse to mount the same filesystem twice on same mountpoint (3394)
4970 - allow llmount to create routes for mounting behind routers (3320)
4971 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4972 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4973 - fix rare problem with rename on htree directories (3417)
4975 - allow default OST striping configuration per directory (1414)
4976 - fix compilation for qswnal for 2.6 kernels (3125)
4977 - increase maximum number of MDS request buffers for large systems
4978 - change liblustreapi to be useful for external progs like lfsck (3098)
4979 - increase local configuration timeout for slow disks (3353)
4980 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4982 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4985 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4986 - don't overwrite extent policy data in reply if lock was blocked (2901)
4987 - drop filter export grants atomically with removal from device (2663)
4988 - del obd_self_export from work_list in class_disconnect_exports (2908)
4989 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4990 - swab reply message in mdc_close, other PPC fixes (2464)
4991 - fix destroying of named logs (2325)
4992 - overwrite old logs when running lconf --write_conf (2264)
4993 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4994 - fix race in target_handle_connect (2898)
4995 - mds_reint_create() should take same inode create lock (2926)
4996 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4997 - don't close files for self_export to avoid uninitialized obd (2936)
4998 - allow MDS with the same name as client node (2939)
4999 - hold dentry reference for closed log files for unlink (2325)
5000 - reserve space for all logs during transactions (2059)
5001 - don't evict page beyond end of stripe extent (2925)
5002 - don't oops on a deleted current working directory (2399)
5003 - handle hard links to targets without a parent properly (2517)
5004 - don't dereference NULL lock when racing during eviction (2867)
5005 - don't grow lock extents when lots of conflicting locks (2919)
5007 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5010 - account for cache space usage on clients to avoid data loss (974)
5011 - lfsck support in lustre kernel code (2349)
5012 - reduce journal credits needed for BRW writes (2370)
5013 - orphan handling to avoid losing space on client/server crashes
5014 - ptlrpcd can be blocked, stopping ALL progress (2477)
5015 - use lock value blocks to assist in proper KMS, faster stat (1021)
5016 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5017 - recovery for initial connections (2355)
5018 - fixes for mds_cleanup_orphans (1934)
5019 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5020 - block all file creations until orphan recovery completes (1901)
5021 - client remove rq_connection from request struct (2423)
5022 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5023 - recovery timer race (2670)
5024 - mdc_close recovey bug (2532)
5025 - ptlrpc cleanup bug (2710)
5026 - mds timeout on local locks (2588)
5027 - namespace lock held during RPCs (2431)
5028 - handle interrupted sync write properly (2503)
5029 - don't try to handle a message that hasn't been replied to (2699)
5030 - client assert failure during cleanup after abort recovery (2701)
5031 - leak mdc device after failed mount (2712)
5032 - ptlrpc_check_set allows timedout requests to complete (2714)
5033 - wait for inflight reqs when ptlrpcd finishes (2710)
5034 - make sure unregistered services are removed from the srv_list
5035 - reset bulk XID's when resending them (caught by 1138 test)
5036 - unregister_bulk after timeout
5037 - fix lconf error (2694)
5038 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5039 - readahead locks pages, leaves pending causing memory pressure (2673)
5040 - increase OST request buffers to 4096 on large machines (2729)
5041 - fix up permission of existing directories in simple_mkdir (2661)
5042 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5043 - don't assign transno to errored transactions (2742)
5044 - don't delete objects on OST if given a bogus objid from MDS (2751)
5045 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5046 - if rq_no_resend, then timeout request after recovery (2432)
5047 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5048 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5049 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5050 - MDS recovery shouldn't delete objects that it has given out (2730)
5051 - if enqueue arrives after completion, don't clobber LVB (2819)
5052 - don't unlock pages twice when trigger_group_io returns error (2814)
5053 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5054 - don't write pages to disk if there was an error (1450)
5055 - don't ping imports that have recovery disabled (2676)
5056 - take buffered bytes into account when balancing socknal conn (2817)
5057 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5058 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5059 - remove little-endian swabbing of llog records (1987)
5060 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5061 - retry reposting request buffers if they fail (1191)
5062 - grow extent at grant time to avoid granting a revoked lock (2809)
5063 - lock revoke doesn't evict page if covered by a second lock (2765)
5064 - disable VM readahead to avoid reading outside lock extents (2805)
5066 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5067 - updated kernel patches for hp-2.4.20 kernel (2681)
5069 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5072 - fix truncated write corruption (2366)
5073 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5075 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5076 - ensure len(uuid) < 37 in lmc (1171)
5077 - fix ia64 OOPS in llog_test (2255)
5078 - zero end of page at obdfilter for partial page writes (2648)
5079 - don't leave stale dentries around after renames (bug 2428)
5080 - fix timeouts when evicting a client with a single lock held (2642)
5081 - set deadline for the initial HELLO message to drain (2634)
5082 - print out dotted-quad IP addresses in the socknal (2302)
5084 - additional debugging for MDS client eviction problem (2443)
5085 - fix mkfsoptions support for osts (2603, 2604)
5087 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5090 - add series for the vanilla 2.6.0 kernel
5091 - add series for the vanilla 2.4.24 kernel
5092 - add series for a cray x86/64 UL kernel drop
5093 - fix xattr patches for the vanilla 2.4.19 series
5095 - generate true UUIDs in lmc (1171)
5096 - have portals stack dumping break in UML (2466)
5097 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5098 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5099 - wake callback waiting threads on client eviction (2460)
5100 - Add --ptldebug and --subsystem to lmc (1719)
5101 - update assertion to allow safe interrupt allocation
5102 - set rq_no_resend for cancel requests (2432)
5103 - recalculate ptlrpcd timeout after resend (2494)
5104 - call vfs_rmdir when removing pending directories (2368)
5105 - fix renaming a file to itself (2429)
5106 - lmc creates a default one-stripe lov (2454)
5107 - expand procfs space to handle large clusters (2326)
5108 - increase UML stack to avoid overflow
5109 - update lconf's list of debug and subsystem masks
5110 - fix lfs find --obd (2510)
5111 - /proc tunable for disabling filter read caching (2591)
5112 - stop rpm packages from altering slapd.conf (2301)
5113 - disable nagle in the socknal under 0conf (2578)
5114 - choose mds inode size based on stripe count (2572)
5115 - fix kernel-source rpm problems (2516)
5117 - add --disable-doc to avoid pdf generation (2421)
5118 - update documentation, tests, type-os, comments
5119 - avoid format warnings on ia64
5120 - remove the TOE NAL
5121 - tiny code cleanups by removing unused fields
5123 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5126 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5127 - remove the most blatant lies from BUILDING file (2371)
5128 - change default debug level to reasonable production setting
5129 - reduce client side cache size to reduce cache flush time
5130 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5131 - make TCP zerocopy and pinger support enabled by default (2476)
5132 - sync writes completed after process exits caused crashes (2319)
5133 - maintain correct mount count on the MDS (2356)
5134 - backout 1557, because 2316 wasn't really fixed
5135 - better file I/O statistics gathering in /proc
5136 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5137 - another kernel patch to fix zero-copy TCP function export
5138 - don't take duplicate lock when processing re-sent getattr (2420)
5139 - lctl uses obd_self_export instead of creating new conn (2353)
5140 - MDS/OST recovery case which requires object creation asserted (2425)
5141 - move lfs from /usr/sbin to /usr/bin in packages
5142 - fix race between mds_client_add and mds_client_free (2417)
5143 - use kmalloc instead of slabs in portals (2430)
5144 - don't create duplicate records when a failover MDS is present (2442)
5145 - remove unnecessary mount age check (2332)
5146 - don't remove directory inodes from locks prematurely (2451)
5147 - don't break if MDS service name is the same as hostname (2103)
5148 - fix races in client write RPC generation when cache full (2482)
5150 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5153 - remove now-unused request->rq_obd (278)
5154 - if an allocation fails, print out how much memory we've used (1933)
5155 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5156 - add the "configurable stack size" patch to most series files (1256)
5157 - ability to write large log records, for 100+ OST configs (2306)
5158 - fix NULL deref when filter_prep fails (2314)
5159 - fix operator precedence error in filter_sync
5160 - dynamic allocation of socknal TX descriptors (2315)
5161 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5162 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5163 - ensure that CWARN messages in Portals always get to the syslog
5164 - __init/__exit are not for prototype decls (ldlm_init/exit)
5165 - x86-64 compile warning fixes
5166 - fix gateway LMC keyword conflict (2318)
5167 - fix MDS lock inversions in getattr/reint paths (1844)
5168 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5169 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5170 - DLM race condition prevented some lock evictions (2328)
5171 - ENOMEM detection and retry on socknal sends (2230)
5172 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5173 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5174 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5175 - better error messages when a client is rejected during recovery (1505)
5176 - avoid cancelling locks which were never granted, after failure (2330)
5177 - fix i_sem/journal inversion in mds_client_add (2333)
5178 - fix truncate/getattr lock cycle deadlock (2334)
5179 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5180 - fix two rare exit paths which could leak an l_lock() ref (2321)
5181 - fencepost error in MDS/OST orphan recovery (2226)
5182 - make log record alignment 8 bytes (1988)
5183 - lstripe now fails when requested offset > ost_count (2237)
5184 - ensure that all kernel series have a complete list.h (1607)
5185 - fix crashes in special-file operations (2316)
5186 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5188 - allow configurable automake binary, for testing new versions
5189 - small update to the lfs documentation
5191 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5193 * fix negative export reference count in fsfilt_sync (2312)
5195 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5196 * release candidate 0.9.1
5198 - orphans are moved into the PENDING directory for possible recovery
5199 - replayed opens now open by fid for orphan/rename safety (1042)
5200 - last close of an orphan inode generates a transno (683)
5201 - chdir() and mount() now pin the directory entry (1020)
5202 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5203 - discard very old requests without processing them (1502)
5204 - remove some common, well-understood CERRORs (1505)
5205 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5206 - clear "grant" flags in OST replies until OST grant code lands (1644)
5207 - fix read performance by not clobbering i_blksize on client (1598)
5208 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5209 - make LRU size a /proc tunable, clears locks when reduced (707)
5210 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5211 - server locks take references on exports now (1558)
5212 - build fixes for 2.4.20-rh trees (1663)
5213 - return an error from lov_create if all OSCs are inactive (1751)
5214 - fix import levels when a reconnect happens without a timeout (1597)
5215 - exit early from mds_open if we get a lookup error (1749)
5216 - partial page read at EOF wouldn't wait for disk before sending (1642)
5217 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5218 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5219 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5220 - write/truncate lock inversion (1639)
5221 - Don't auto-load obdclass, portals modules during cleanup (1495)
5222 - fix timestamps from jumping to "now" (1763)
5223 - extra journal assertions (1648)
5224 - add an extra multiunlink test (1771)
5225 - fix read_record/write_record API (1776)
5226 - fix leak of offset_extent, possible incorrect i_size later (1772)
5227 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5228 - Debugging for the kqswnal_get_idle_tx problems (1820)
5229 - Allow recovery to be attempted multiple times (1536)
5230 - Write out MDS last_rcvd file after it is first created (1600)
5231 - Fix tx_descriptor leak in failed transmit situations (1827)
5232 - ext3 journaling fixes for assertion failure after IO error (1871)
5233 - class_export_put() on freed export after completion AST error (1896)
5234 - Fix revalidate looping in VFS (1322)
5235 - Don't access a freed export during MDS_REINT timeout (1521)
5236 - Add open-unlink recovery support on the MDS (1673,1764)
5237 - Return an error if no MDS data was read from last_rcvd (1946)
5238 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5239 - Don't setup a disk device that doesn't match exported UUID (317)
5240 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5241 - avoid committing NULL handle in force close
5242 - local.sh is now a one-stripe LOV configuration
5243 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5244 - fix ext3 htree duplicate directory entry corruption (1516)
5245 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5246 - update to new LOV EA format (2097)
5247 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5249 2003-06-15 Phil Schwan <phil@clusterfs.com>
5252 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5253 - per-import/export recovery handling (958, 931, 959)
5254 - multiple last-rcvd slots, for serving multiple FSes (949)
5255 - connections are again shared between multiple imp/exports (963, 964)
5256 - "umount -f" would hang if any requests needed to be sent (393, 978)
5257 - avoid pinning large req buffer by copying for queued messages (989)
5258 - add "uuid" to "lctl device" command to help upcalls (991)
5259 - "open" RPCs with transnos would confuse recovery counters (1037)
5260 - do proper endian conversion of all wire messages (288, 340, 891)
5261 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5262 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5263 - fix ptlrpc_connection leak in target_handle_connect (1174)
5264 - fix import refcounting bug in OST and MDS cleanup (1134)
5265 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5266 - fix very unlikely obd_types race condition (501)
5267 - remove osc_open hack for echo_client (1187)
5268 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5269 - a failure in read_inode2 leads to deadlock (1139)
5270 - cancel ack-locks as soon as transaction is committed (1072)
5271 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5272 - make sure to commitrw after any preprw to avoid deadlock (1162)
5273 - failing to execute a file in a lustre FS would lock inode (1203)
5274 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5275 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5276 - obd_uuid2tgt would walk off the end of the list (1255)
5277 - on IA64 the osc would give portals incorrect bulk size (1258)
5278 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5279 - fix lock inversion caused by new llite matching code (1282)
5280 - limit the number of dirty pages on a client to 10MB (1286)
5281 - timed out locks were not being corrected cancelled (1289)
5282 - fix O_DIRECT above 4GB on IA-32 (1292)
5283 * major user-visible changes
5284 - fail out/fail over policy now controlled by the upcall (993)
5286 - add OBD_PING to check server availability and failure (954)
5287 - lustre messages are now sent in sending host order (288, 340, 891)
5288 - add eadatalen to MDS getattr reply (340)
5289 - OST read replies may contain second buffer, with per-page status (593)
5291 2003-03-11 Phil Schwan <phil@clusterfs.com>
5294 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5295 - failed open()s could cause deadlock; fixed (867, 869)
5296 - stop cancelling OST locks when files are closed (481)
5297 - overlapping XID spaces caused network corruption (851, 853)
5298 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5299 - setattr_raw added, to do single-RPC, server-side setattrs
5300 - lmc/lconf syntax change for OST UUIDs
5301 - fix crashy race condition between ptlrpc_free_req and osc_close
5302 - don't use request in mdc_enqueue if we hit a timeout (889)
5303 - don't set the inode i_size for regular files from the MDS (896)
5304 - handle out of order completion AST (842)
5305 - don't LBUG if a lock request times out after receiving AST (913)
5306 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5307 - if a bad lock AST arrives, send an error instead of dropping entirely
5308 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5309 - fix leak in bulk IO when only partially completed (899, 900, 926)
5310 - fix O_DIRECT for ia64 (55)
5311 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5312 - C-z after timeout could hang a process forever; fixed (977)
5314 - client-side I/O cache (678, 924, 929, 941, 970)
5316 - READPAGE and SETATTRs which don't take server-side locks get
5319 2003-02-11 Phil Schwan <phil@clusterfs.com>
5322 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5323 - Fix fsfilt_extN_readpage() to read a full page of directory
5324 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5325 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5326 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5327 - Most intent operations were removed, in favour of a new RPC mode
5328 that does a single RPC to the server and bypasses most of the VFS
5329 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5330 - Aggressively cancel local locks on DLM servers
5331 - mds_reint_unlink sends EA to the client if it's the last nlink.
5332 client uses that EA to unlink OST objects.
5333 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5334 - recursive symlinks were fixed (439)
5335 - fixed NULL deref in DEBUG_REQ
5336 - filter_update_lastobjid no longer calls sync, which annoyed extN
5337 - fixed multi-client small-writes to a single file problem (445)
5338 - fixed mtime updates during file writes (607)
5339 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5340 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5341 - fixed Linux OST size reporting problem (444, 656)
5342 - OST now updates object mtime with writes or setattr (607, 619)
5343 - client verifies file size before zeroing page past EOF (445)
5344 - OST now writes last allocated objid to disk with allocation (108)
5345 - LOV on echo now works (409)
5347 - mds_reint_unlink sends a new buffer, with the EA included. this
5348 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5349 set if a regular file was being unlinked, and it was the last link
5350 - use PtlGet from the target for bulk writes (315)
5351 - OST now updates object mtime with writes or setattr (607, 619)
5352 - LDLM now has a grant-time callback to revalidate locked items, if
5354 - Many MDS operations were reorganized to combat race conditions
5356 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5357 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5359 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5362 - Fully reactivate OST imports after reconnection (512, others)
5363 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5364 - More graceful error handling for truncating on dead OST (515)
5365 - Don't error out unless we're actually accessing dead stripes (474)
5366 - Fix garbage sizes when stripes are missing (410)
5367 - LRU counters were broken, causing constant lock purge (433, 432)
5368 - garbage on read from stripes with failed OSTs (441)
5369 - mark OSCs as active before reconnecting during recovery (438)
5370 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5371 - lfind did not preserve OST order in output (443)
5372 - symlinks cause hung clients, incorrect data (439)
5373 - stop dereferencing request after dropping refcount (457)
5374 - don't LASSERT(spin_is_locked) on non-SMP (455)
5375 - fixes for many rename() bugs
5376 - fstat didn't correctly synchronize attributes (399)
5377 - server must handle lock cancellation during blocking AST prep (487)
5378 - bulk descriptors were free()d too soon (511)
5379 - fix paths in lconf, which would load incorrect modules (451, 507)
5380 - fix confusing lconf 'host not found' error message (386)
5381 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5382 - fix race condition in mdc_blocking_ast() for inode access (526)
5383 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5384 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5385 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5386 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5387 - fix lstripe in several ways (kernel side) (527)
5388 - fix request leak in ldlm_cli_enqueue (262)
5389 - incorrect OSC was marked inactive after OST failure
5390 - call mds_fs_cleanup before unmounting filesystem (524)
5391 - fix races between taking ns_lock and ldlm_lock_change_resource
5392 - fix races updating LOV export open file list
5393 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5394 - fix recovery NULL deref in ldlm_cli_cancel_unused
5395 - fix some DLM races by using new hash table for lock handles (419)
5396 - permit the client to specify desired inodes, at replay
5397 - duplicate requests when we queue them for replay reintegration
5398 - fix last_rcvd offset calculation
5399 - sync after each recovered transaction, so we always make progress
5400 - never, not always, ERESTART requests without transnos
5401 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5403 - skip replay if the MDS says that the client is already connected
5404 - don't check for a recovery-enabled export to match lctl's UUID
5405 - don't INC_USE_COUNT for phantom exports
5406 - don't crash when cleaning up phantom exports (567)
5407 - don't double-finish or set replay data for errored mdc_open requests
5408 - abort requests when they time out, so we don't get old replies
5409 - send/receive replies for AST messages again
5410 - if the client says that it doesn't have the lock, cancel it on the
5412 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5413 mark it as destroyed, it will all work out when decref is called
5414 - fix module use counts (22, 581)
5416 - ASTs now expect a reply (server cancels lock on error reply)
5418 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5421 - fix many simultaneous client startup (392)
5422 - fix dentry->d_it clobbering
5423 - credentials weren't being shipped for readdir/getattr operations
5424 - remove invalid assertions triggered during some concurrent MD
5426 - proper Lustre versions added (336, 389)
5427 - fix memory leak for create error case (398)
5428 - fix LOV locking bug that would get cli/srv out of sync
5429 - fix echo client over LOV (409)
5430 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5431 - fix double-O_EXCL intent crash (424)
5432 - avoid sending multiple lock CANCELs (352)
5434 - MDS can do multi-client recovery (modulo bugs in new code)
5436 - many updates, edits, cleanups
5438 2002-11-18 Phil Schwan <phil@clusterfs.com>
5441 - fix null d_it dereference (346)
5442 - fix full OST/dbench hang (333)
5443 - fix permission problem with file removal (286)
5444 - fix removal of OSCs from LOV when they fail
5445 - fix NULL deref during bulk timeout (214)
5446 - fix problems related to multiple filesystems on one MDS (241)
5447 - fixed serious subtle metadata locking bugs
5448 - free locks on clients when inodes are removed due to memory
5450 - fix inode pointer in lock data (285)
5451 - partial support for multiple MDS on a single host (241)
5452 - data locks weren't cancelled at clear_inode time (290, 311)
5453 - intent locks could lead to unbounded lock growth (205)
5454 - added a maximum lock count, an LRU list, and a flusher
5455 - fix multiple rename (365)
5456 - properly abstracted the echo client
5457 - OSC locked 1 byte too many; fixed
5458 - rewrote brw callback code:
5459 - fixed recovery bugs related to LOVs (306)
5460 - fixed too-many-pages-in-one-write crash (191)
5461 - fixed (again) crash in sync_io_timeout (214)
5462 - probably fixed callback-related race (385)
5464 - Add capability to MDS protocol
5465 - LDLM cancellations and callbacks on different portals
5467 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5470 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5471 - defer open object destruction to close time (601981, 138)
5472 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5473 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5474 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5475 - don't call obd_setattr in ll_file_release for destroyed objects
5476 * protocol change to lustre_msg: move |version| and add |flags|
5477 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5478 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5479 * added replay of create, unlink, link and rename operations during
5480 MDS failover; recovery should be much more robust now
5481 * remove failed OSCs from LOVs (only lov_create uses this so far)
5482 * the lustre-HOWTO was brought (more) up to date (582544)
5484 2002-10-23 Phil Schwan <phil@clusterfs.com>
5487 - in-use dentries weren't being reused properly (617851)
5488 - prevent multiple LDLM setup (599178)
5489 - fix LOV size calculations for truncate (617853)
5490 - fix client handling of MDS intent errors (POSIX)
5491 - fix permission bug in lovstripe.c test (624321)
5492 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5493 - truncate past end of file could corrupt data
5494 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5495 - a race in recovery could return ETIMEDOUT to apps (623947)
5496 - building outside the source directory was fixed
5497 * the lustre-HOWTO was brought (more) up to date (582544)
5498 * major progress was made on recovery functionality
5500 2002-10-10 Phil Schwan <phil@clusterfs.com>
5503 - recovery deadlock fix
5504 - rm -rf causes LBUG fix (617817)
5505 - file open by multiple tasks fix (618962)
5506 - directory permissions bugs (602707 and 620007)
5507 - journal_stop fixed with locking (611313)
5508 - O_APPEND failures resolved (618273, perhaps 614459)
5509 - lconf PATH fix (619770)
5510 - IA64 build fix (621450)
5511 - RPC buffer sizes scale with amount of memory
5513 2002-10-01 Phil Schwan <phil@clusterfs.com>
5516 - locks would be cancelled without throwing away data pages,
5517 resulting in inconsistent data (605627)
5518 - inode attributes were not always being refreshed (605627, 612449)
5519 - lconf now continues to cleanup after lctl reports an error
5520 - MDS now enforces user permissions (602707)
5521 - lprocfs cleanup fixed, but not yet enabled (614157)
5522 - fixed infinite server hang, should a client not respond to an AST
5523 - avoid going into recovery if user calls readlink() with a buffer
5524 that's too small (613941)
5525 - AST RPCs no longer require replies (614867) -- this may be changed
5526 - don't crash server if client sends an IOV that's too big (611336)
5527 - fixed lock conversion deadlock (611892)
5528 - fixed the following of symlinks (614622)
5529 * recovery: the server can remove locks from a client that dies, other
5530 clients can make progress
5531 * more extN patch fixes
5532 * compile-time configurable ptlrpc buffer allocations
5534 - collaborative read cache document
5535 - Lustre Lite Performance CDR document-in-progress
5537 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5540 - fix typo in patch-2.4.18
5542 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5545 - clear ptlrpc request each time in handle_incoming_request()
5546 - unlink of files now destroys the object on the OST
5548 2002-09-19 Peter Braam <braam@clusterfs.com>
5550 * add hard link support
5551 * change obdfile creation method
5552 * kernel patch changed
5554 2002-09-19 Peter Braam <braam@clusterfs.com>
5557 - stack overflow bug in extN fixed
5559 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5561 * documentation updates
5562 - add man pages for config tools
5563 - update tests/README to describe testing with new config tools
5564 - finish metadata API descriptions
5565 * bug fixes and cleanups
5566 - statfs workaround for 16TB limit
5567 - LOV stripe allocation improved, can stripe on subset of OSTs
5568 - LOV file size/IO offset was wrong for files > 4GB in size
5569 - object EA data was being dropped, caused files to be unreadable
5570 - memory overflow with non-LOV OST caused memory corruption
5571 - fixed regression tests to work with new config tools, obdfilter
5572 - fixed bug when directory size became larger than 1 block
5573 - fixed bug (for single client case) when PWD was deleted
5574 - invalidate local directory pages when doing intent-based ops
5575 - avoid LDLM oops when lock callback contained bad data
5577 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5579 * documentation updates
5580 * bug fixes and cleanups
5581 - configuration tools
5584 - 64-bit compile warnings
5585 - 64-bit internal statfs data
5587 * test_brw on persistent OST devices
5589 * lprocfs (disabled)
5591 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5593 * documentation updates
5594 * bug fixes and cleanups
5595 * configuration tools
5597 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5600 * many small fixes to 0_5_4
5601 * io/network handling
5602 * thinkos in MDS operations
5604 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5607 * crucial basic fixes to 0.5.3
5608 * IOR, Iozone work over Elan
5611 2002-08-07 Phil Schwan <phil@clusterfs.com>
5612 * version 0_5_3, our first alpha
5613 * we use the new Portals iovs
5614 * documentation updates
5615 * bug fixes and cleanups
5616 * small changes in the DLM wire protocol
5618 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5619 * version 0_5_1 with some initial stability,
5620 * locking on MD and file I/O.
5621 * documentation updates
5622 * several bug fixes since 0.5.0
5623 * small changes in wire protocol
5625 2002-07-18 Phil Schwan <phil@clusterfs.com>
5627 * delivered as Lustre Light Alpha
5628 * fixed a crash after handling invalid MDS requests
5629 * fixed directory pages for architectures with non-4k pages sizes
5631 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5632 * release version v0_4_4
5633 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5635 * Fixes a number of interruption problems with OST operations.
5636 * Update documentation for portals header changes
5637 * Move all wire protocol structs/defines to lustre_idl.h
5638 * Fixes symlink length bug.
5639 * Add tcpdump to repository.
5641 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5642 * release version v0_4_3
5643 * Fixes statfs for inodes on extN.
5644 * Fixes bug in runtests which would delete /etc/hosts.
5645 * Use 64-bit object IDs wherever possible (not into VFS though)
5646 Remove ost_get_info, which is unused by lustre, and out of date.
5648 2002-07-03 Peter Braam <braam@clusterfs.com>
5649 * release version v0_4_2 Fixes a lookup error (type not passed)
5650 * move forward to head of Portals
5651 * move forward to latest Lustre kernel
5653 2002-06-25 Peter Braam <braam@clusterfs.com>
5654 * release version v0_4_1. Hopefully stable on single node use.