1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.27 (SLES 10),
5 2.6.18-92.1.10.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
17 Frequency : Create a symlink file with a very long name
19 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
20 Details : If there is no extra space in the request for early cancels,
21 ldlm_req_handles_avail() returns 0 instead of a negative value.
23 Severity : enhancement
25 Description: Add /proc entry for import status
26 Details : The mdc, osc, and mgc import directories now have
27 an import directory that contains useful import data for debugging
30 Severity : enhancement
32 Description: Re-disable certain /proc logging
33 Details : Enable and disable client's offset_stats, extents_stats and
34 extents_stats_per_process stats logging on the fly.
37 Frequency : Only on FC kernels 2.6.22+
39 Description: oops in statahead
40 Details : Do not drop reference count for the dentry from VFS when lookup,
41 VFS will do that by itself.
43 Severity : enhancement
45 Description: Generic /proc file permissions
46 Details : Set /Proc file permissions in a more generic way to enable non-
47 root users operate on some /proc files.
51 Description: Hitting mdc_commit_close() ASSERTION
52 Details : Properly handle request reference release in
53 ll_release_openhandle().
57 Frequency : only patchless client
58 Description: add workaround for race between add/remove dentry from hash
60 Severity : enhancement
62 Description: Allow OST glimpses to return PW locks
66 Description: LBUG when llog conf file is full
67 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
71 Description: Prevent import from entering FULL state when server in recovery
75 Description: service mount cannot take device name with ":"
76 Details : Only when device name contains ":/" will mount treat it as
82 Description: replace ptlrpcd with the statahead thread to interpret the async
83 statahead RPC callback
87 Frequency : on recovery
88 Description: I/O failures after umount during fail back
89 Details : if client reconnected to restarted server we need join to recovery
90 instead of find server handler is changed and process self
91 eviction with cancel all locks.
93 Severity : enhancement
95 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
97 Severity : enhancement
99 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
101 Severity : enhancement
103 Description: Upcall on Lustre log has been dumped
104 Details : Allow for a user mode script to be called once a Lustre log has
105 been dumped. It passes the filename of the dumped log to the
106 script, the location of the script can be specified via
107 /proc/sys/lnet/debug_log_upcall.
112 Description: avoid idr_remove called for id which is not allocated.
113 Details : Move assigment s_dev for clustered nfs to end of initialization,
114 to avoid problem with error handling.
119 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
120 Details : When connection is reused this not moved from CONN_UNUSED_HASH
121 into CONN_USED_HASH and this prodice warning when put connection
122 again in unused hash.
124 Severity : enhancement
126 Description: File striping can now be set to use an arbitrary pool of OSTs.
128 Severity : enhancement
130 Description: Export bytes_read/bytes_write count on OSC/OST.
134 Description: Early reply size mismatch, MGC loses connection
135 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
136 the connect flags are properly negotiated.
140 Description: Properly propagate oinfo flags from lov to osc for statfs
141 Details : restore missing copy oi_flags to lov requests.
143 Severity : enhancement
145 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
146 : l_getgroups(8), lst(8), routerstat(8)
147 Details : included man pages for llobdstat(8), llstat(8),
148 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
150 Severity : enhancement
152 Description: configure's --enable-quota should check the
153 : kernel .config for CONFIG_QUOTA
154 Details : configure is terminated if --enable-quota is passed but
155 : no quota support is in kernel
159 Description: Remove portals compatibility
160 Details : Remove portals compatibility, not interoperable with releases
165 Description: Resolve device initialization race
166 Details : Prevent proc handler from accessing devices added to the
167 obd_devs array but yet be intialized.
169 Severity : enhancement
171 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
173 Severity : enhancement
175 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
180 Description: Using wrong pointer in osc_brw_prep_request
181 Details : Access to array[-1] can produce panic if kernel compiled with
182 CONFIG_PAGE_ALLOC enabled
186 Description: Client runs out of low memory
187 Details : Consider only lowmem when counting initial number of llap pages
191 Description: *optional* service tags registration
192 Details : if the "service tags" package is installed on a Lustre node
193 When the filesystem is mounted, a local-node service tag will
194 be created. See http://inventory.sun.com/ for more information
195 about the Service Tags asset management system.
199 Description: Kernel BUG tries to release flock
200 Details : Lustre does not destroy flock lock before last reference goes
201 away. So always drop flock locks when client is evicted and
202 perform unlock regardless of successfulness of speaking to MDS.
206 Description: add refcount protection for osc callbacks, avoid panic on shutdown
210 Description: sanity test 65a fails if stripecount of -1 is set
211 Details : handle -1 striping on filesystem in ll_dirstripe_verify
216 Description: ASSERTION(CheckWriteback(page,cmd)) failed
217 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
220 Severity : enhancement
222 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
226 Description: do not process already freed flock
227 Details : flock can possibly be freed by another thread before it reaches
228 to ldlm_flock_completion_ast.
232 Description: LBUG during stress test
233 Details : Need properly lock accesses the flock deadlock detection list.
237 Description: oops in page fault handler
238 Details : kernel page fault handler can return two special 'pages' in error
239 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
243 Description: timeout with invalidate import.
244 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
245 should be handled by ptlrpcd. This produce long age waiting and
246 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
248 Severity : enhancement
250 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
254 Description: enable MGS and MDT services start separately
255 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
256 a co-located MGS without starting the MGS, which is a complement
257 to 'nosvc' mount option.
261 Frequency : after recovery
262 Description: precreate to many object's after del orphan.
263 Details : del orphan st in oscc last_id == next_id and this triger growing
264 count of precreated objects. Set flag LOW to skip increase count
265 of precreated objects.
269 Frequency : rare, on clear nid stats
270 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
271 Details : when clean nid stats sometimes try destroy live entry,
272 and this produce panic in free.
276 Description: Stack overflow during MDS log replay
277 ease stack pressure by using a thread dealing llog_process.
281 Description: wait until IO finished before start new when do lock cancel.
282 Details : VM protocol want old IO finished before start new, in this case
283 need wait until PG_writeback is cleared until check dirty flag and
284 call writepages in lock cancel callback.
286 Severity : enhancement
288 Description: using special macro for print time and cleanup in includes.
292 Description: mds_mfd_close() ASSERTION(rc == 0)
293 Details : In mds_mfd_close(), we need protect inode's writecount change
294 within its orphan write semaphore to prevent possible races.
298 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
302 Frequency : rare, on shutdown ost
303 Description: don't hit live lock with umount ost.
304 Details : shrink_dcache_parent can be in long loop with destroy dentries,
305 use shrink_dcache_sb instead.
309 Description: don't panic with use echo client
310 Details : echo client pass NULL as client nid pointer and this produce null
315 Description: fix build on ppc32
316 Details : compile code with -m64 flag produce wrong object file for ppc32.
320 Description: add message levels for liblustreapi
324 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
325 Details : Move the 'good_osts' check before the 'total_bavail' check. This
326 will result in an -EAGAIN and in the exit call path we call
327 alloc_rr() which will with increasing aggressiveness attempt to
328 aquire precreated objects on the minimum number of required OSCs.
332 Description: Use old size assignment to avoid deadlock
333 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
334 the scheduling while holding a spinlock. We do not need locking
335 for size in ll_update_inode() because size is only updated from
336 the MDS for directories or files without objects, so there is no
337 other place to do the update, and concurrent access to such inodes
338 are protected by the inode lock.
342 Description: resolve "_IOWR redefined" build error on SLES10
346 Description: dump the memory debugging after all modules are unloaded to
347 suppress false negative in conf_sanity test 39
349 Severity : enhancement
351 Description: build kernel-ib packages for OFED 1.3 in our release cycle
356 Description: fix SLES kernel versioning
357 Details : the kernel version for our SLES 10 kernel did not include a "-"
358 before the "smp" at the end. while this was not a problem in
359 general, it did mean that software trying to use the kernel
360 version to try to detect a vendor specific kernel would fail.
361 this was most evident by the OFED build scripts.
365 Description: Don't update lov_desc members until making sure they are valid
366 Details : When updating lov_desc members via proc fs, need fix their
367 validities before doing the real update.
371 Description: don't put request into delay list while invalidate in flight.
372 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
373 import in flight. this produce timeout for invalidate and sometimes
374 can cause stale data.
378 Frequency : on ppc only
379 Description: not convert ost objects for directory because it's not exist.
380 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
384 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
386 Severity : enhancement
388 Description: Update to support 2.6.22.14 vanilla kernel.
392 Frequency : rare, on recovery
393 Description: read procfs can produce deadlock in some situation
394 Details : Holding lprocfs lock which send rpc can produce block for destroy
395 obd objects and this also block reconnect with -EALREADY. This
396 isn't fix all lprocfs bugs - but make it rare.
398 Severity : enhancement
400 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
403 Frequency : frequent on X2 node
405 Description: mdc_set_open_replay_data LBUG
406 Details : Set replay data for requests that are eligible for replay.
410 Description: lustre_mgs: operation 101 on unconnected MGS
411 Details : When MGC is disconnected from MGS long enough, MGS will evict the
412 MGC, and late on MGC cannot successfully connect to MGS and a lot
413 of the error messages complaining that MGS is not connected.
417 Frequency : on network error
418 Description: panic with double free request if network error
419 Details : mdc_finish_enqueue is finish request if any network error ocuring,
420 but it's true only for synchronus enqueue, for async enqueue
421 (via ptlrpcd) this incorrect and ptlrpcd want finish request
424 Severity : enhancement
426 Description: client-side metadata stat-ahead during readdir(directory readahead)
427 Details : perform client-side metadata stat-ahead when the client detects
428 readdir and sequential stat of dir entries therein
431 Frequency : on start mds
433 Description: Implement get_info(last_id) in obdfilter.
436 Frequency : occasional
438 Description: Correctly check stale fid, not start epoch if ost not support SOM
439 Details : open with flag O_CREATE need set old fid in op_fid3 because
440 op_fid2 was overwritten with new generated fid, but mds can answer
441 with one of these two fids and both is not stale. Setattr
442 incorrectly started an epoch and assume will be called
443 done_writing, but without SOM done_writing ever being called.
446 Frequency : rare, depends on device drivers and load
448 Description: MDS or OSS nodes crash due to stack overflow
449 Details : Code changes in 1.8.0 increased the stack usage of some functions.
450 In some cases, in conjunction with device drivers that use a lot
451 of stack the MDS (or possibly OSS) service threads could overflow
452 the stack. One change which was identified to consume additional
453 stack has been reworked to avoid the extra stack usage.
456 Frequency : occasional
458 Description: Do not fail import if osc_interpret_create gets -EAGAIN
459 Details : If osc_interpret_create got -EAGAIN it immediately exits and
460 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
461 oscc_has_objects and see OSC has no objests and call
462 oscc_internal_create to resend create request.
464 Severity : enhancement
466 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
468 Severity : enhancement
470 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
473 Frequency : very rare
475 Description: Fix replay if there is an un-replied request and open
476 Details : In some cases, older replay request will revert the
477 mcd->mcd_last_xid on MDS which is used to record the client's
480 Severity : enhancement
482 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
484 Severity : enhancement
486 Description: Add rhel5 support to HEAD.
488 Serverity : enhancement
490 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
495 Description: Don't allow skipping OSTs if index has been specified.
496 Details : Don't allow skipping OSTs if index has been specified, make
497 locking in internal create lots better.
501 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
502 Details : access to bitfield in structure is always rounded to long
503 and this produce problem with not atomic change any bit.
507 Description: Lustre make rpms failed.
508 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
509 builds Lustre rpms from tarball.
512 Frequency : rare on shutdown ost
514 Description: If llog cancel was not send before clean_exports phase, this can
515 produce deadlock in llog code.
516 Details : If llog thread has last reference to obd and call class_import_put
517 this produce deadlock because llog_cleanup_commit_master wait when
518 last llog_commit_thread exited, but this never success because was
519 called from llog_commit_thread.
523 Description: allow userland application know is lost one of stripes.
524 Details : fill lvb_blocks with error code on ost and return it to
525 application if error flag found.
529 Description: NULL lov_tgts causing MDS oops
530 Details : more safe checks for NULL lov_tgts for avoid oops.
532 Severity : enhancement
534 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
538 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
539 Details : If an OST is down the MDS will hang indefinitely in
540 obd_statfs_rqset() waiting for the statfs data. While for
541 MDS QOS usage of statfs, it should not stuck in waiting.
543 Severity : enhancement
545 Description: remote_acl support
546 Details : Support ACL-based permission check for remote user.
547 Support setfacl/getfacl for remote user with the utils
548 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
549 the system "{s,g}etfacl" utils.
551 Severity : enhancement
553 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
555 Severity : enhancement
557 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
561 Description: make mgs_setparam() handle fsname containing dash
562 Details : fsname containing a dash does not work with lctl conf_param
564 Severity : enhancement
566 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
568 Severity : enhancement
570 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
574 Frequency : rare, at shutdown
575 Description: access already free / zero obd_namespace.
576 Details : if client_disconnect_export was called without force flag set,
577 and exist connect request in flight, this can produce access to
578 NULL pointer (or already free pointer) when connect_interpret
579 store ocd flags in obd_namespace.
583 Frequency : only at startup
584 Description: not alloc memory with spinlock held.
585 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
586 if any spinlock held.
588 Severity : enhancement
590 Description: make lustre randomly fail allocating memory
591 Details : Make lustre randomly failed allocating memory for testing purpose.
593 Severity : enhancement
595 Description: lost problems with lov objid file
596 Details : Fixes some scability and access to not inited memory problems
597 in work with lov objdid file.
602 Description: lfs find does not continue on file error
603 Details : Continue other files processing when a file/dir is absent.
607 Description: Inconsistent usage of lustre_pack_reply()
608 Details : Standardize the usage of lustre_pack_reply() such that it
609 always generate a CERROR on failure.
612 Frequency : occasional
614 Description: MDS hang or stay in waiting lock
615 Details : If client receive lock with CBPENDING flag ldlm need send lock
616 cancel as separate rpc, to avoid situation when cancel request
617 can't processed due all i/o threads stay in wait lock.
621 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
625 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
629 Description: Update to SLES9 kernel-2.6.5-7.287.3.
633 Description: Update to RHEL4 latest kernel.
635 Severity : enhancement
637 Description: Build SLES10 patchless client fails
638 Details : The configure was broken by run ./configure with
639 --with-linux-obj=.... argument for patchless client. When the
640 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
641 can't search header adequately. Use absolute path such as
642 -I($LINUX)/include instead.
646 Description: interrupt oig_wait produce painc on resend.
647 Details : brw_redo_request can be used for resend requests from ptlrpcd and
648 private set, and this produce situation when rq_ptlrpcd_data not
649 copyed to new allocated request and triggered LBUG on assert
650 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
651 ptlrpcd set if request is changed and can be safety changed to use
656 Description: LASSERT_{REQ,REP}SWAB macros are buggy
657 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
658 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
659 these macros and replace them with their difinitions instead.
663 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
664 Details : Update which_patch & target file for SLES10 latest kernel.
667 Description: Remove client patches from SLES 10 kernel.
668 Details : This causes SLES 10 clients to behave as patchless clients
669 even on a Lustre-patched (server) kernel.
671 Severity : enhancement
673 Description: self-adjustable client's lru lists
674 Details : use adaptive algorithm for managing client cached locks lru
675 lists according to current server load, other client's work
676 pattern, memory activities, etc. Both, server and client
677 side namespaces provide number of proc tunables for controlling
680 Severity : enhancement
682 Description: light-weight GSS support
683 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
684 and reduce performance overhead.
686 Severity : enhancement
688 Description: Linux keyring support
689 Details : Support using service of Linux keyring for Lustre GSS internal
690 context refresh/cache mechanism.
694 Description: Fix errors in lfs documentation
695 Details : Fixes man pages
699 Description: don't use GFP_* in generic Lustre code.
700 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
704 Description: obdclass is limited by single OBD_ALLOC(idarray)
705 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
709 Description: warnings with build patchless client with vanila 2.6.19 and up
710 Details : change old ctl_table style and replace ctl_table/ctl_table_header
711 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
715 Frequency : Only for SLES
716 Description: sanity_quota fail test_1
717 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
718 file, which makes TSTID[2] inunique.
722 Frequency : Only for SLES9
723 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
724 Details : Update target/ChangeLog/which_patch .
728 Frequency : Only for test_10 in sanity-quota.sh
729 Description: a bug in quota test code
730 Details : checking a test flag in a wrong place
734 Frequency : Only for test_13 in sanity-quota.sh
735 Description: a bug in quota test script
736 Details : shouldn't check overfull allocated quota
740 Description: jbd statistics
741 Details : Port older jbd statistics patch for sles10
743 Severity : enhancement
745 Description: Add printing inode info into message about error in writepage.
750 Description: reply_lock_interpret crash due to race with it and lock cancel.
751 Details : Do not replay locks that are being cancelled. Do not reference
752 locks by their address during replay, just by their handle.
756 Frequency : When flocks are used.
757 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
758 Details : Flock locks might destroy just granted lock if it could be merged
759 with another existing flock, this is done in completion handler,
760 so teach ldlm_cli_enquque_fini that this is a valid case for
766 Description: Oops in read and write path when failing to allocate lock.
767 Details : Check if lock allocation failed and return error back.
771 Description: lstripe command fails for valid OST index
772 Details : The stripe offset is compared to lov->desc.ld_tgt_count
773 instead of lov->desc.ld_active_tgt_count.
777 Description: sanity.sh failed test 103
778 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
783 Description: df doesn't work properly if diskfs blocksize != 4K
784 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
788 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
789 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
791 Severity : enhancement
793 Description: Adaptive timeouts
794 Details : RPC timeouts adapt to changing server load and network
795 conditions to reduce resend attempts and improve recovery time.
799 Description: llapi_file_create() does not allow some changes
800 Details : add llapi_file_open() that allows specifying the mode and
801 open flags, and also returns an open file handle.
805 Description: merge and cleanup kernel patches.
806 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
810 Description: Add journal checksum support.(Kernel part)
811 Details : The journal checksum feature adds two new flags i.e
812 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
813 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
814 indicates that the commit block contains the checksum for
815 the blocks described by the descriptor blocks. Now commit
816 record can be sent to disk without waiting for descriptor
817 blocks to be written to disk. This behavior is controlled
818 using JBD2_FEATURE_ASYNC_COMMIT flag.
822 Description: OSS needs mutliple precreate threads
823 Details : Add ability to start more than one create thread per OSS.
827 Description: Sanity.sh test_65k failed: setstripe should have succeeded
828 Details : execute lfs setstripe on client
832 Description: mds_obd_create error creating tmp object
833 Details : When the user sets quota on root, llog will be affected and can't
834 create files and write files.
837 Frequency : Always on ia64 patchless client, and possibly others.
839 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
840 Details : This allows the patchless client to be loaded on architectures
845 Description: RedHat Update kernel for RHEL5
846 Details : Add the kernel config file for RHEL5.
850 Description: RedHat Update kernel for RHEL5
851 Details : Modify the kernel config file more closer RHEL5.
855 Description: Build failure against Centos5 (RHEL5)
856 Details : Define PAGE_SIZE when it isn't present.
860 Description: test_7 on sanity-quota.sh can't be used on separate machines
861 Details : add facet to handle it
865 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
866 Details : with NFS, the anon dentry's parent was set to itself in
867 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
868 corresponding dentry other than use rec->ur_name.
870 Severity : enhancement
872 Description: lfs setstripe enhancement
873 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
877 Description: enable data checksumming by default
878 Details : enable checksum by default, allow --disable-checksum
879 configure option and "-o nochecksum" mount option. Checksums
880 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
884 Description: lustre support for RHEL5
885 Details : Add support for RHEL5.
889 Description: Client eviction due to failover config
890 Details : after a connection loss, the lustre client should attempt to
891 reconnect to the last active server first before trying the
892 other potential connections.
896 Description: when mds and osts use different quota unit(32bit and 64bit),
897 quota will be released repeatly.
898 Details : void sending multiple quota reqs to mds, which will keep the
899 status between the reqs.
903 Description: osts not allocated evenly to files
904 Details : change the condition to increase offset_idx
907 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
909 Description: Data corruption for OSTs that are formatted larger than 2TB
911 Details : When generating the bio request for lustre file writes the
912 sector number would overflow a temporary variable before being
913 used for the IO. The data reads correctly from Lustre (which
914 will overflow in a similar manner) but other file data or
915 filesystem metadata may be corrupted in some cases.
919 Description: Tune the kernel for good SCSI performance.
920 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
921 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
926 Description: rewrite ext2-derived code in obdclass/uuid.c
927 Details : rewrite inherited code (uuid parsing code from ext2 utils)
928 from scratch preserving functionality.
932 Description: "lfs find -obd UUID" prints directories
933 Details : "lfs find -obd UUID" will return all directory names instead
934 of just file names. It is incorrect because the directories
935 do not reside on the OSTs.
939 Description: use i_size_read and i_size_write in 2.6 port
940 Details : replace inode->i_size access with i_size_read/write()
943 Frequency : when removing large files
945 Description: scheduling issue during removal of large Lustre files
946 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
947 It causes scheduling issues when removing large files (17TB in the
951 Frequency : only with liblustre clients on XT3
953 Description: evictions taking too long
954 Details : allow llrd to evict clients directly on OSTs
957 Frequency : only on ppc
959 Description: /proc/fs/lustre/devices broken on ppc
960 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
961 We should make sure the type of 'index' is loff_t and then cast
962 explicitly as needed below. Do not assign an explicitly cast
966 Frequency : only for rhel5
968 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
969 Details : Modify the target file & which_kernel.
971 Severity : enhancement
973 Description: omit set fsid for export NFS
974 Details : fix set/restore device id for avoid EMFILE error and mark lustre
975 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
979 Frequency : Always, for kernels after 2.6.16
980 Description: Fix warning idr_remove called for id=.. which is not allocated.
981 Details : Last kernels save old s_dev before kill super and not allow
982 to restore from callback - restore it before call kill_anon_super.
986 Description: buffer overruns could theoretically occur
987 Details : llapi_semantic_traverse() modifies the "path" argument by
988 appending values to the end of the origin string, and a
989 overrun may occur. Adding buffer overrun check in liblustreapi.
993 Description: Fix error on 'ls .' at the top of the Lustre mount.
994 Details : Don't revalidate dentry if it is a root dentry.
998 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
999 Details : Modify vm-tunables-rhel4.patch.
1003 Description: Kernel config for 2.6.18-vanilla.
1004 Details : Modify targets/2.6-vanilla.target.in.
1005 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1006 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1007 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1008 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1012 Description: improve handling recoverable errors
1013 Details : if request processig with error which can be recoverable on server
1014 request should be resend, otherwise page released from cache and
1019 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1020 Details : Modify target file & which_patch.
1021 A flaw was found in the IA32 system call emulation provided
1022 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1023 value could be stored in the %RAX register, which could trigger an
1024 out-of-bounds system call table access. An untrusted local user
1025 could exploit this flaw to run code in the kernel
1026 (ie a root privilege escalation). (CVE-2007-4573).
1030 Description: O_DIRECT bypasses client statistics.
1031 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1032 read_ahead_stats were updated. copied the stats section from
1033 osc_send_oap_rpc() into async_internal().
1037 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1041 Description: change order of libsysio includes
1042 Details : '#include sysio.h' should always come before '#include xtio.h'
1046 Description: handle "serious error: objid * already exists" more gracefully
1047 Details : If LAST_ID value on disk is smaller than the objects existing in
1048 the O/0/d* directories, it indicates disk corruption and causes an
1049 LBUG(). If the object is 0-length, then we should use the existing
1050 object. This will help to avoid a full fsck in most cases.
1052 Severity : enhancement
1054 Description: adapt the lustre_config script to support the upgrade case
1055 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1056 server targets to 1.6 in parallel.
1060 Description: To avoid grant space > avaible space when the disk is almost
1061 full. Without this patch you might see the error "grant XXXX >
1062 available" or some LBUG about grant, when the disk is almost
1064 Details : In filter_check_grant, for non_grant cache write, we should
1065 check the left space by if (*left > ungranted + bytes), instead
1066 of (*left > ungranted), because only we are sure the left space
1067 is enough for another "bytes", then the ungrant space should be
1068 increase. In client, we should update cl_avail_grant only there
1069 is OBD_MD_FLGRANT in the reply.
1073 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1074 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1075 described in CVE-2007-4573. This problem could allow untrusted
1076 local users to gain root access.
1079 Frequency : when using O_DIRECT and quotas
1081 Description: Incorrect file ownership on O_DIRECT output files
1082 Details : block usage reported by 'lfs quota' does not take into account
1083 files that have been written with O_DIRECT.
1088 Description: touch file failed when fs is not full
1089 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1090 otherwise we can get ENOSP while fs is not full.
1094 Description: parallel lock callbacks
1095 Details : Instead of sending blocking and completion callbacks as separated
1096 requests, adding them to a set and sending in parallel.
1099 Frequency : only for Cray XT3
1100 Bugzilla : 12829/13455
1101 Description: Changing primary group doesn't change the group lustre assigns to
1103 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1104 overridden with the primary group provided by the group upcall,
1105 whereas the supplied fsgid can be trusted if it is in the list of
1106 supplementary groups returned by the group upcall.
1108 Severity : enhancement
1110 Description: Allow masking D_WARNING, D_ERROR messages from console
1111 Details : Console messages can now be disabled via lnet.printk.
1115 Description: User code with malformed file open parameter crashes client node
1116 Details : Before packing join_file req, all the related reference should be
1117 checked carefully in case some malformed flags cause fake
1118 join_file req on client.
1122 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1123 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1124 mode is equal to the granted mode. However, it should also check
1125 that LDLM_FL_CP_REQD is not set.
1129 Description: Heavy nfs access might result in deadlocks
1130 Details : After ELC code landed, it is now improper to enqueue any mds
1131 locks under och_sem, because enqueue might want to decide to
1132 cancel open locks for same inode we are holding och_sem for.
1136 Description: Client eviction while running blogbench
1137 Details : A lot of unlink operations with concurrent I/O can lead to a
1138 deadlock causing evictions. To address the problem, the number of
1139 oustanding OST_DESTROY requests is now throttled to
1140 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1141 callbacks are processed in priority.
1145 Description: enable ACLs on MDS by default
1146 Details : ACLs must be enabled on MDS by default.
1149 Frequency : PPC/PPC64 only
1151 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1152 Details : fix duplicated definitions between asm-ppc64/types.h and
1153 lustre_types.h on PPC.
1156 Frequency : PPC/PPC64 only
1158 Description: asm-ppc/segment.h does not exist
1159 Details : fix compile issue on PPC.
1163 Description: better handle error messages in extents code
1166 Frequency : RHEL4 only
1168 Description: mkfs is very slow on IA64/RHEL4
1169 Details : A performance regression has been discovered in the MPT Fusion
1170 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1171 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1172 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1174 Severity : enhancement
1176 Description: SNMP support enhancement
1177 Details : Adding total number of sampled request for an MDS node in snmp
1180 Severity : enhancement
1182 Description: Optimize ldlm waiting list processing for PR extent locks
1183 Details : When processing waiting list for read extent lock and meeting read
1184 lock that is same or wider to it that is not contended, skip
1185 processing rest of the list and immediatelly return current
1186 status of conflictness, since we are guaranteed there are no
1187 conflicting locks in the rest of the list.
1191 Description: Time out and refuse to reconnect
1192 Details : When the failover node is the primary node, it is possible
1193 to have two identical connections in imp_conn_list. We must
1194 compare not conn's pointers but NIDs, otherwise we can defeat
1195 connection throttling.
1199 Description: port llog fixes from b1_6 into HEAD
1200 Details : Port llog reference couting and some llog cleanups from b1_6
1201 (bug 10800) into HEAD, for protect from panic and access to already
1202 free llog structures.
1206 Description: Detect stride IO mode in read-ahead
1207 Details : When a client does stride read, read-ahead should detect that and
1208 read-ahead pages according to the detected stride pattern.
1212 Description: data checksumming impacts single node performance
1213 Details : add support for several checksum algorithm. Currently, only CRC32
1214 and Adler-32 are supported. The checksum type can be changed on
1215 the fly via /proc/fs/lustre/osc/*/checksum_type.
1219 Description: use adler32 for page checksums
1220 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1225 Description: build for x2 fails
1226 Details : fix compile issue on Cray systems.
1230 Description: Properly match for duplicate locks
1231 Details : Due to different lock order from skiplists code, we need to
1232 traverse entire list for now
1235 Frequency : only on PPC/SLES10
1237 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1238 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1239 be defined. Add a hack in mkfs_lustre.c to work around this compile
1244 Description: LASSERT on MDS when client holding flock lock dies
1245 Details : ldlm pool logic depends on number of granted locks equal to
1246 number of released locks which is not true for flock locks, so
1247 just exclude such locks from consideration.
1251 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1252 Details : Use fsfilt_sync() for both the whole filesystem sync and
1253 individual file sync to eliminate dangerous inode locking
1254 with I_LOCK that can lead to a deadlock.
1258 Description: performance in 1.6.3
1259 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1260 sure that 1MB requests are not fragmented by the block layer.
1262 Severity : enhancement
1264 Description: organize the server-side client stats on per-nid basis
1265 Details : Change the structure of stats under obdfilter and mds to
1273 The "uuid"s file would list the uuids of _active_ exports.
1274 And the clear entry is to clear all stats and stale nids.
1276 Severity : enhancement
1278 Description: eliminate client locks in face of contention
1279 Details : file contention detection and lockless i/o implementation
1280 for contended files.
1284 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1287 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1289 Description: MDS may lose file striping (and hence file data) in some cases
1290 Details : If there are additional extended attributes stored on the MDS,
1291 in particular ACLs, SELinux, or user attributes (if user_xattr
1292 is specified for the client mount options) then there is a risk
1293 of attribute loss. Additionally, the Lustre file striping
1294 needs to be larger than default (e.g. striped over all OSTs),
1295 and an additional attribute must be stored initially in the
1296 inode and then increase in size enough to be moved to the
1297 external attribute block (e.g. ACL growing in size) for file
1302 Description: skiplist implementation simplification
1303 Details : skiplists are used to group compatible locks on granted list
1304 that was implemented as tracking first and last lock of each
1305 lock group the patch changes that to using doubly linked lists
1309 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1310 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1311 because LASSERT on both the data supplied by a client, and the
1312 data on disk is dangerous and incorrect.
1314 Severity : enhancement
1316 Description: Slow truncate/writes to huge files at high offsets.
1317 Details : Directly associate cached pages to lock that protect those pages,
1318 this allows us to quickly find what pages to write and remove
1319 once lock callback is received.
1323 Description: more ldlm soft lockups
1324 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1325 starve other threads from the resource lock for a long time in
1326 case of long waiting queue, so change the debug level from
1327 D_OTHER to the less frequently used D_INFO.
1329 Severity : enhancement
1331 Description: add -gid, -group, -uid, -user options to lfs find
1335 Description: Hung threads in invalidate_inode_pages2_range
1336 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1337 once one is completed. As a result, some RPCs are stuck in the
1338 queue and are never sent.
1342 Description: filter threads hungs on waiting journal commit
1343 Details : Cleanup filter group llog code, then only filter group llog will
1344 be only created in the MDS/OST syncing process.
1348 Description: Procfs and llog threads access destoryed import sometimes.
1349 Details : Sync the import destoryed process with procfs and llog threads by
1350 the import refcount and semaphore.
1352 Severity : enhancement
1354 Description: openlock cache of b1_6 port to HEAD
1359 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1360 Details : If insertion of an extent fails, then discard the inode
1361 preallocation and free data blocks else it can lead to duplicate
1366 Description: don't always update ctime in ext3_xattr_set_handle()
1367 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1368 In some cases the ctime should not be updated, for example for
1369 2.0->1.8 compatibility it is necessary to delete an xattr and it
1370 should not update the ctime.
1374 Bugzilla : 15713/16362
1375 Description: Assertion in iopen_connect_dentry in 1.6.3
1376 Details : looking up an inode via iopen with the wrong generation number can
1377 populate the dcache with a disconneced dentry while the inode
1378 number is in the process of being reallocated. This causes an
1379 assertion failure in iopen since the inode's dentry list contains
1380 both a connected and disconnected dentry.
1384 Description: assertion failure in ldlm_handle2lock()
1385 Details : fix a race between class_handle_unhash() and class_handle2object()
1386 introduced in lustre 1.6.5 by bug 13622.
1391 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1392 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1397 Description: oops in sd_iostats_seq_show()
1398 Details : unloading/reloading the scsi low level driver triggers a kernel
1399 bug when trying to access the sd iostat file.
1404 Description: Kernel panics during QLogic driver reload
1405 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1406 patch, causing memory corruption.
1411 Description: journal_dev option does not work in b1_6
1412 Details : pass mount option during pre-mount.
1414 Severity : enhancement
1416 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1417 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1418 extent information of a file. It can be used to map logical blocks
1419 in a file to physical blocks in the block device.
1423 Description: LDLM soft lockups - improvement
1424 Details : It is be possible to send the lock handle along with each read
1425 or write request because the client is already doing a lock match
1426 itself so there isn't any reason the OST should have to re-do that
1430 Frequency : only on Cray X2
1432 Description: X2 build failures
1433 Details : fix build failures on Cray X2.
1437 Description: xid & resent requests
1438 Details : Initialize RPC XID from clock at startup (randomly if clock is
1441 Severity : enhancement
1443 Description: Add lustre_start utility to start or stop multiple Lustre servers
1448 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1449 Details : In case of memory pressure, list_del() can be called twice on
1450 req->rq_history_list, causing a kernel oops.
1454 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1455 Details : fix stack overflow in the distributed lock manager by defering
1456 export eviction after a failed AST to the elt thread instead of
1457 handling it in the dlm interpret routine.
1461 Description: Convert some comments to new format.
1462 Details : Update documenting comments to match doxygen conventions.
1466 Description: Grammar fixes.
1467 Details : A couple of trivial spelling fixes.
1471 Description: OSD_COUNTERS-mandatory
1472 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1473 to simplify the code.
1477 Description: simplify cmm_device freeing logic.
1478 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1482 Description: Add lockdep support to dt_object_operations locking interface.
1483 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1484 parameter indicating lock ordering. Update mdd code to use new
1489 Description: Introduce failloc constants for lockless IO tests.
1490 Details : Add two new failloc constants to test lockless IO. Only one of
1491 them in implemented---another is checked in yet to be landed
1496 Description: Add lockdep support for inode mutex.
1497 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1498 in the situations where Lustre has to lock more than one inode
1503 Description: Add optional invariants checking support.
1504 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1505 configure switch. This macro is to be used for consistency and
1506 sanity checks that are too expensive to be left in `production'
1511 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1512 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1517 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1518 Details : Use spin_lock_nested() in (the only) situation where more than
1519 one ldlm_lock is locked simultaneously. Also, fix possible
1520 dead-lock in ldlm_lock_change_resource() by enforcing particular
1525 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1526 Details : Instead of specifying each ldlm_lock call-back through separate
1527 parameter, wrap them into struct ldlm_callback_suite.
1531 Description: Kill join_lru obd method and its callers.
1532 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1533 in memory---a requirement implemented through ->o_join_lru() obd
1534 method in HEAD. Remove this method and its users.
1538 Description: Add asynchronous ldlm ENQUEUE completion handler.
1539 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1540 case that stock ldlm_completion_ast() cannot handle as it waits
1541 until lock is granted. Introduce new ldlm_completion_ast_async()
1542 for this. Also comment ldlm_completion_ast().
1546 Description: ldlm_error <-> errno conversion.
1547 Details : Add functions to map (rather arbitrary) between LDLM error codes
1548 and standard errno values. CLIO needs this to prevent LDLM specific
1549 constants from escaping ldlm and osc.
1553 Description: Kill unused ldlm_handle2lock_ns() function.
1554 Details : Kill unused ldlm_handle2lock_ns() function.
1558 Description: Add lu_ref support to ldlm_lock
1559 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1560 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1561 and ldlm_resource. LDLM interface has to be changed, because code that
1562 releases a reference on a lock, has to "know" what reference this is.
1563 In the most frequent case
1565 lock = ldlm_handle2lock(handle);
1567 LDLM_LOCK_PUT(lock);
1569 no changes are required. When any other reference (received _not_ from
1570 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1571 instead of LDLM_LOCK_PUT().
1573 Arguably, changes are pervasive, and interface requires some discipline
1574 for proper use. On the other hand, it was very instrumental in finding
1575 a few leaked lock references.
1579 Description: Add ldlm_lock_addref_try().
1580 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1581 attempts to addref a lock that might be being canceled
1586 Description: Add ldlm_weigh_callback().
1587 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1588 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1589 hard-coded `number of pages' logic.
1593 Description: Add lockdep annotations to llog code.
1594 Details : Use appropriately tagged _nested() locking calls in the places
1595 where llog takes more than one ->lgh_lock lock.
1599 Description: Add loi_kms_set().
1600 Details : Wrap kms updates into a helper function.
1604 Description: Constify instances of struct lsm_operations.
1605 Details : Constify instances of struct lsm_operations.
1609 Description: lu_conf support.
1610 Details : On a server, a file system object is uniquely identified
1611 by a fid, which is sufficient to locate and load all object
1612 state (inode). On a client, on the other hand, more data are
1613 necessary instantiate an object. Change lu_object_find() and
1614 friends to take additional `lu_conf' argument describing object.
1615 Typically this includes layout information.
1619 Description: lu_context fixes.
1620 Details : Introduce new lu_context functions that are needed on the client
1621 side, where some system threads (ptlrpcd) are shared by multiple
1622 modules, and so cannot be stopped during module shutdown.
1626 Description: Add start and stop methods to lu_device_type_operations.
1627 Details : Introduce two new methods in lu_device_type_operations, that are
1628 invoked when first instance of a given type is created and last one
1629 is destroyed respectively. This is need by CLIO.
1633 Description: Add lu_ref support to struct lu_device.
1634 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1635 track leaked references.
1639 Description: Introduce lu_kmem_descr.
1640 Details : lu_kmem_descr and its companion interface allow to create
1641 and destroy a number of kmem caches at once.
1645 Description: Fix lu_object finalization race.
1646 Details : Fix a race between lu_object_find() finding an object and its
1647 concurrent finalization. This race is (most likely) not possible
1648 on the server, but might happen on the client.
1652 Description: Introduce lu_ref interface.
1653 Details : lu_ref is a debugging module allowing to track references to
1654 a given object. It is quite cpu expensive, and has to be
1655 explicitly enabled with --enable-lu_ref. See usage description
1660 Description: Factor lu_site procfs stats into a separate function.
1661 Details : Separate lu_site stats printing code into a separate function
1662 that can be reused on a client.
1666 Description: Constify instances of struct {lu,dt,md}_device_operations.
1667 Details : Constify instances of struct {lu,dt,md}_device_operations.
1671 Description: Introduce struct md_site and move meta-data specific parts of
1672 struct lu_site here.
1673 Details : Move md-specific fields out of struct lu_site into special struct
1674 md_site, so that lu_site can be used on a client.
1678 Description: Kill mdd_lov_destroy().
1679 Details : Remove unused mdd code.
1683 Description: Add st_block checking to multistat.c.
1684 Details : Add st_block checking to multistat.c.
1688 Description: Add lu_ref support to struct obd_device.
1689 Details : Add lu_ref tracking to obd_device.
1693 Description: Kill obd_set_fail_loc().
1694 Details : Remove unused code.
1698 Description: Add special type for ptlrpc_request interpret functions.
1699 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1700 there. Actual usage will be in CLIO.
1704 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1705 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1706 doesn't work with lockdep.
1710 Description: Add rwv.c test program.
1711 Details : New testing program exercising readv(2) and writev(2) (Qian).
1715 Description: Add sendfile.c test program.
1716 Details : New testing program exercising sendfile(2) (Jay).
1720 Description: Ratelimit a message that can be very frequent.
1721 Details : Ratelimit a memory allocation failure message that can
1726 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1727 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1731 Description: Liblustre build fixes.
1732 Details : Liblustre build fixes.
1736 Description: libcfs: add cfs_{need,cond}_resched() interface.
1737 Details : libcfs: add cfs_{need,cond}_resched() definition and
1738 implementations for Linux, NT, and liblustre.
1740 Severity : enhancement
1742 Description: More exported tunables for mballoc
1743 Details : Add support for tunable preallocation window and new tunables for
1744 large/small requests
1748 Description: Detect corruption of block bitmap and checking for preallocations
1749 Details : Checks validity of on-disk block bitmap. Also it does better
1750 checking of number of applied preallocations. When corruption is
1751 found, it turns filesystem readonly to prevent further corruptions.
1755 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1756 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1757 window to 0. If it does hit cache miss, and read-pattern is still
1758 stride-io mode, does not reset the stride window, but also does not
1759 increase the stride window length in this case.
1763 Frequency : only for big-endian servers
1764 Description: Check if system is big-endian while mounting fs with extents feature
1765 Details : Mounting a filesystem with extents feature will fail on big-endian
1766 systems since ext3-based ldiskfs is not supported on big-endian
1767 systems. This can be over-riden with "bigendian_extents" mount option.
1769 --------------------------------------------------------------------------------
1771 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1773 * Support for kernels:
1774 2.6.5-7.283 (SLES 9),
1775 2.6.9-55.EL (RHEL 4),
1776 2.6.16.46-0.14 (SLES 10),
1777 2.6.18.8 vanilla (kernel.org)
1778 * Client support for unpatched kernels:
1779 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1780 2.6.16 - 2.6.22 vanilla (kernel.org)
1781 * Due to recently discovered recovery problems, we do not recommend
1782 using patchless RHEL 4 clients with this or any earlier release.
1783 * Recommended e2fsprogs version: 1.39.cfs8
1784 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1785 * Starting with this release, the ldiskfs backing filesystem required
1786 by Lustre is now in its own package, lustre-ldiskfs. This package
1787 should be installed. It is versioned separately from Lustre and
1788 may be released separately in future.
1792 Description: block reactivating mgc import until all deactivates complete
1793 Details : Fix race when failing back MDT/MGS to itself (testing)
1796 Frequency : at statup only
1798 Description: mds_lov_synchronize race leads to various problems
1799 Details : simultaneous MDT->OST connections at startup can cause the
1800 sync to abort, leaving the OSC in a bad state.
1802 Severity : enhancement
1804 Description: add optional extra BUILD_VERSION info
1805 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1806 to override the lustre version.
1809 Frequency : 2.6.18 servers only
1811 Description: ll_kern_mount() doesn't release the module reference
1812 Details : The ldiskfs module reference count never drops down to 0
1813 because ll_kern_mount() doesn't release the module reference.
1818 Description: server LBUG when using old ost_num_threads parameter
1819 Details : Accept the old ost_num_threads parameter but warn that it
1820 is deprecated, and fix an off-by-one error that caused an LBUG.
1825 Description: Transient SCSI error results in persistent IO issue
1826 Details : iobuf->dr_error is not reinitialized to 0 between two
1830 Frequency : sometimes when underlying device returns I/O errors
1832 Description: OSTs not going read-only during write failures
1833 Details : OSTs are not remounted read-only when the journal commit threads
1834 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1835 instead of the ext3 wrappers.
1839 Description: poor connect scaling with increasing client count
1840 Details : Don't run filter_grant_sanity_check for more than 100 exports
1841 to improve scaling for large numbers of clients.
1844 Frequency : SLES10 only
1846 Description: sanity-quota.sh quotacheck failed: rc = -22
1847 Details : Quotas cannot be enabled on SLES10.
1850 Frequency : liblustre clients only
1852 Description: getdirentries does not give error when run on compute nodes
1853 Details : getdirentries does not fail when the size specified as an argument
1854 is too small to contain at least one entry
1856 Severity : enhancement
1858 Description: Add LNET router traceability for debug purposes
1859 Details : If a checksum failure occurs with a router as part of the
1860 IO path, the NID of the last router that forwarded the bulk data
1861 is printed so it can be identified.
1866 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1867 Details : Due to a race condition, liblustre clients were occasionally
1868 evicted incorrectly.
1870 Severity : enhancement
1872 Description: lfs setstripe use optional parameters instead of postional
1875 Severity : enhancement
1877 Description: Nanosecond timestamp support for ldiskfs
1878 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1879 resolution timestamps. There is not yet support for this at
1880 the Lustre filesystem level.
1883 Frequency : during server recovery
1885 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1886 Details : request with rq_no_resend flag not awake l_wait_event if they get
1890 Frequency : nfs export on patchless client
1892 Description: connectathon hang when test nfs export over patchless client
1893 Details : Disconnected dentry cannot be found with lookup, so we do not need
1894 to unhash it or make it invalid
1897 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1898 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1899 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1904 Description: open req refcounting wrong on reconnect
1905 Details : If reconnect happened between getting open reply from server and
1906 call to mdc_set_replay_data in ll_file_open, we will schedule
1907 replay for unreferenced request that we are about to free.
1908 Subsequent close will crash in variety of ways.
1909 Check that request is still eligible for replay in
1910 mdc_set_replay_data().
1915 Description: disable writes to filesystem when reading health_check file
1916 Details : the default for reading the health_check proc file has changed
1917 to NOT do a journal transaction and write to disk, because this
1918 can cause reads of the /proc file to hang and block HA state
1919 checking on a healthy but otherwise heavily loaded system. It
1920 is possible to return to the previous behaviour during configure
1921 with --enable-health-write.
1923 Severity : enhancement
1925 Description: 64-bit inode version
1926 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1927 to the inode. This will be required for version-based recovery.
1932 Description: MDS fails to start if a duplicate client export is detected
1933 Details : in some rare cases it was possible for a client to connect to
1934 an MDS multiple times. Upon recovery the MDS would detect this
1935 and fail during startup. Handle this more gracefully.
1937 Severity : enhancement
1939 Description: Add -o localflock option to simulate old noflock
1941 Details : This will achieve local-only flock/fcntl locks
1947 Description: log_commit_thread vs filter_destroy race leads to crash
1948 Details : Take import reference before releasing llog record semaphore
1953 Description: Wrong request locking in request set processing
1954 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1955 imp_delayed_list, in this place should be used imp_lock.
1958 Frequency : when reconnection
1960 Description: Grant Leak when osc reconnect to OST
1961 Details : When osc reconnect ost, OST(filter) should check whether it
1962 should grant more space to client by comparing fed_grant and
1963 cl_avail_grant, and return the granted space to client instead
1964 of "new granted" space, because client will call osc_init_grant
1965 to update the client grant space info.
1968 Frequency : when client reconnect to OST
1970 Description: Grant Leak when osc do resend and replay bulk write
1971 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1972 bulk write request, because the grant info while be sync between
1973 OSC and OST when reconnect, and we should ignore the grant info
1974 these of resend/replay write req.
1979 Description: Grant space more than avaiable left space sometimes.
1980 Details : When then OST is about to be full, if two bulk writing from
1981 different clients came to OST. Accord the avaliable space of the
1982 OST, the first req should be permitted, and the second one
1983 should be denied by ENOSPC. But if the seconde arrived before
1984 the first one is commited. The OST might wrongly permit second
1985 writing, which will cause grant space > avaiable space.
1988 Frequency : when client is evicted
1990 Description: Grant might be wrongly erased when osc is evicted by OST
1991 Details : when the import is evicted by server, it will fork another
1992 thread ptlrpc_invalidate_import_thread to invalidate the
1993 import, where the grant will be set to 0. While the original
1994 thread will update the grant it got when connecting. So if
1995 the former happened latter, the grant will be wrongly errased
1996 because of this race.
2001 Description: Checking Stale with correct fid
2002 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2003 to check whether it is stale, because sometimes, we want the
2004 enqueue happened anyway, and op_data.fid2 will not be initialized.
2006 Severity : enhancement
2008 Description: update patchless client
2009 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2012 Frequency : only with 2.4 kernel
2014 Description: random memory corruption
2015 Details : size of struct ll_inode_info is to big for union inode.u and this
2016 can be cause of random memory corruption.
2021 Description: Memory leak in recovery
2022 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2023 It should also check obd_fail before fsfilt_start, otherwise if
2024 fsfilt_start return -EROFS,(failover mds during mds recovery).
2025 then the req will return with repmsg->transno = 0 and rc = EROFS.
2026 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2027 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2028 be freed no matter whether fsfilt_commit success or not.
2031 Frequency : only with huge count clients
2033 Description: Prevents from taking the superblock lock in llap_from_page for
2035 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2036 for llap_from_page call in ll_removepage prevents from taking the
2037 superblock lock for a soon died page.
2042 Description: Not check open intent error before release open handle
2043 Details : in some rare cases, the open intent error is not checked before
2044 release open handle, which may cause
2045 ASSERTION(open_req->rq_transno != 0), because it tries to release
2046 the failed open handle.
2051 Description: Set cat log bitmap only after create log success.
2052 Details : in some rare cases, the cat log bitmap is set too early. and it
2053 should be set only after create log success.
2057 Description: Accessing a block bevice can re-enable I/O when Lustre is
2058 tearing down a device.
2059 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2064 Description: service threads may hog cpus when there are a lot of requests
2065 Details : Insert cond_resched to give other threads a chance to use some CPU
2070 Description: the cat log was not initialized in recovery
2071 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2072 unlink log on mds will not be initialized until mds post
2073 recovery. And also in mds post recovery, the unlink log will
2074 initialization will be done asynchronausly, so there will be race
2075 between add unlink log and unlink log initialization.
2079 Description: brw_stats were being printed incorrectly
2080 Details : brw_stats were being printed as log2 but all of them were not
2081 recorded as log2. Also remove some code duplication arising from
2082 filter_tally_{read,write}.
2086 Frequency : rare, only in recovery.
2087 Description: ASSERTION(req->rq_type != LI_POISON) failed
2088 Details : imp_lock should be held while iterating over imp_sending_list for
2089 prevent destroy request after get timeout in ptlrpc_queue_wait.
2093 Description: replay-single.sh test 52 fails
2094 Details : A lock's skiplist need to be cleanup when it being unlinked
2095 from its resource list.
2099 Description: Short directio read returns full requested size rather than
2101 Details : Direct I/O operations should return actual amount of bytes
2102 transferred rather than requested size.
2104 Severity : enhancement
2106 Description: metadata RPC reduction (e.g. for rm performance)
2107 Details : decrease the amount of synchronous RPC between clients and servers
2108 by canceling conflicing lock before the operation on the client
2109 and packing thier handles into the main operation RPC to server.
2111 Severity : enhancement
2113 Description: Async OSC create to avoid the blocking unnecessarily.
2114 Details : If a OST has no remain object, system will block on the creating
2115 when need to create a new object on this OST. Now, ways use
2116 pre-created objects when available, instead of blocking on an
2117 empty osc while others are not empty. If we must block, we block
2118 for the shortest possible period of time.
2122 Description: improve handling recoverable errors
2123 Details : if request processig with error which can be recoverable on server
2124 request should be resend, otherwise page released from cache and
2127 Severity : enhancement
2129 Description: refine locking for avoid write wrong info into lov_objid file
2130 Details : fix possible races with add new target and write/update data in
2133 --------------------------------------------------------------------------------
2135 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2140 Frequency : on some architectures
2142 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2143 Details : Uninitialized flags sometimes cause configuration commands to
2147 Frequency : patchless clients only
2149 Description: missing __iget() symbol export
2150 Details : The __iget() symbol export is missing. To avoid the need for
2151 this on patchless clients the deathrow inode reaper is turned
2152 off, and we depend on the VM to clean up old inodes. This
2153 dependency was during via the fix for bug 12181.
2155 --------------------------------------------------------------------------------
2157 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2159 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2160 INTEROPERATE with older versions automatically. In many cases a
2161 special upgrade step is needed. Please read the
2162 user documentation before upgrading any part of a live system.
2163 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2164 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2165 * WARNING: Lustre configuration and startup changes are required with
2166 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2170 Severity : enhancement
2172 Description: MountConf
2173 Details : Lustre configuration is now managed via mkfs and mount
2174 commands instead of lmc and lconf. New obd types (MGS, MGC)
2175 are added for dynamic configuration management. See
2176 https://mail.clusterfs.com/wikis/lustre/MountConf for
2179 Severity : enhancement
2181 Description: dynamic OST addition
2182 Details : OSTs can now be added to a live filesystem
2184 Severity : enhancement
2186 Description: startup order invariance
2187 Details : MDTs and OSTs can be started in any order. Clients only
2188 require the MDT to complete startup.
2190 Severity : enhancement
2192 Description: parallel, asynchronous orphan cleanup
2193 Details : orphan cleanup is now performed in separate threads for each
2194 OST, allowing parallel non-blocking operation.
2196 Severity : enhancement
2198 Description: optimized stripe assignment
2199 Details : stripe assignments are now made based on ost space available,
2200 ost previous usage, and OSS previous usage, in order to try
2201 to optimize storage space and networking resources.
2203 Severity : enhancement
2205 Description: Permanently set tunables
2206 Details : All writable /proc/fs/lustre tunables can now be permanently
2207 set on a per-server basis, at mkfs time or on a live system.
2209 Severity : enhancement
2211 Description: Lustre message v2
2212 Details : Add lustre message format v2.
2214 Severity : enhancement
2216 Description: client OST exclusion list
2217 Details : Clients can be started with a list of OSTs that should be
2218 declared "inactive" for known non-responsive OSTs.
2222 Description: SPEC SFS validation failure on NFS v2 over lustre.
2223 Details : Changes the blocksize for regular files to be 2x RPC size,
2224 and not depend on stripe size.
2226 Severity : enhancement
2228 Description: Multiple MD RPCs in flight.
2229 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2230 To avoid overly-overloading MDS, introduce a limit on number of
2231 MDS RPCs in flight for a single client and add /proc controls
2232 to adjust this limit.
2234 Severity : enhancement
2236 Description: client read/write statistics
2237 Details : Add client read/write call usage stats for performance
2238 analysis of user processes.
2239 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2240 file access. extents_stats shows chunk size distribution.
2241 extents_stats_per_process show chunk size distribution per
2244 Severity : enhancement
2246 Description: mds statistics
2247 Details : Add detailed mds operations statistics in
2248 /proc/fs/lustre/mds/*/stats.
2252 Description: Failure of copying files with lustre special EAs.
2253 Details : Client side always return success for setxattr call for lustre
2254 special xattr (currently only "trusted.lov").
2259 Description: Refcount LNET uuids
2260 Details : The global LNET uuid list grew linearly with every startup;
2261 refcount repeated list entries instead of always adding to
2264 Severity : enhancement
2266 Description: Dynamic service threads
2267 Details : Within a small range, start extra service threads
2268 automatically when the request queue builds up.
2271 Frequency : mixed-endian client/server environments
2273 Description: mixed-endian crashes
2274 Details : The new msg_v2 system had some failures in mixed-endian
2277 Severity : enhancement
2279 Description: Easy OST removal
2280 Details : OSTs can be permanently deactivated with e.g. 'lctl
2281 conf_param lustre-OST0001.osc.active=0'
2283 Severity : enhancement
2285 Description: MGS proc entries
2286 Details : Added basic proc entries for the MGS showing what filesystems
2289 Severity : enhancement
2291 Description: provide MGS failover
2292 Details : Added config lock reacquisition after MGS server failover.
2294 Severity : enhancement
2296 Description: add Linux 2.4 support
2297 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2301 Description: a large application tries to do I/O to the same resource and dies
2302 in the middle of it.
2303 Details : Check the req->rq_arrival time after the call to
2304 ost_brw_lock_get(), but before we do anything about
2305 processing it & sending the BULK transfer request. This
2306 should help move old stale pending locks off the queue as
2307 quickly as obd_timeout.
2310 Frequency : when an incorrect nid is specified during startup
2312 Description: ptlrpc connect to non-existant node causes kernel crash
2313 Details : LNET can't be re-entered from an event callback, which
2314 happened when we expire a message after the export has been
2315 cleaned up. Instead, hand the zombie cleanup off to another
2318 Severity : enhancement
2320 Description: plain/inodebits lock performance improvement
2321 Details : Grouping plain/inodebits in granted list by their request modes
2322 and bits policy, thus improving the performance of search through
2326 Frequency : only if OST filesystem is corrupted
2328 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2329 Details : for a short time RPCs with bulk IO are in the replay list,
2330 but replay of bulk IOs is unimplemented. If the OST filesystem
2331 is corrupted due to disk cache incoherency and then replay is
2332 started it is possible to trip an assertion. Avoid putting
2333 committed RPCs into the replay list at all to avoid this issue.
2336 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2338 Description: System hang on startup
2339 Details : This bug allowed the liblustre (e.g. catamount) client to
2340 return to the app before handling all startup RPCs. This
2341 could leave the node unresponsive to lustre network traffic
2342 and manifested as a server ptllnd timeout.
2344 Severity : enhancement
2346 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2347 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2348 boolean to control whether to print peer debug info when a
2349 client's RPC times out.
2352 Frequency : only for kernels with patches from Lustre below 1.4.3
2354 Description: Remove old rdonly API
2355 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2358 Frequency : only for devices with external journals
2360 Description: Set external device read-only also
2361 Details : During a commanded failover stop, we set the disk device
2362 read-only while the server shuts down. We now also set any
2363 external journal device read-only at the same time.
2366 Frequency : when upgrading from 1.4 while trying to change parameters
2368 Description: The wrong (new) MDC name was used when setting parameters for
2369 upgraded MDT's. Also allows changing of OSC (and MDC)
2370 parameters if --writeconf is specified at tunefs upgrade time.
2373 Frequency : when setting specific ost indicies
2375 Description: QOS code breaks on skipped indicies
2376 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2377 created with --index need not be sequential.
2379 Severity : enhancement
2381 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2382 Details : The uninit_groups feature works in conjunction with the kernel
2383 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2384 up the pass1 processing of e2fsck. This is a read-only feature
2385 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2386 mount filesystems that have had this feature enabled.
2388 Severity : enhancement
2390 Description: Improve multi-block allocation algorithm to avoid fragmentation
2391 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2392 allocation locality and avoid filesystem fragmentation.
2394 ------------------------------------------------------------------------------
2396 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2398 * Support for kernels:
2399 2.4.21-47.0.1.EL (RHEL 3)
2400 2.6.5-7.283 (SLES 9)
2401 2.6.9-42.0.10.EL (RHEL 4)
2402 2.6.12.6 vanilla (kernel.org)
2403 2.6.16.27-0.9 (SLES 10)
2404 * Recommended e2fsprogs version: 1.39.cfs5
2406 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2410 Frequency : occasional, depends on client load and configuration
2411 Bugzilla : 12181, 12203
2412 Description: data loss for recently-modified files
2414 Details : In some cases it is possible that recently written or created
2415 files may not be written to disk in a timely manner (this should
2416 normally be within 30s unless client IO load is very high).
2417 The problem appears as zero-length files or files that are a
2418 multiple of 1MB in size after a client crash or client eviction
2419 that are missing data at the end of the file.
2421 This problem is more likely to be hit on clients where files are
2422 repeatedly created and unlinked in the same directory, clients
2423 have a large amount of RAM, have many CPUs, the filesystem has
2424 many OSTs, the clients are rebooted frequently, and/or the files
2425 are not accessed by other nodes after being written.
2427 The presence of the problem can be detected by looking at
2428 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2429 smaller than the second (nr_unused) then dirty files will not
2430 be flushed automatically to disk. "sync; sleep 10" should be
2431 run several times on the node before unmounting it to update
2432 Lustre (this is also safe to run on nodes without this problem).
2434 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2435 that can cause this same problem, so customers using that kernel
2436 also need to update the kernel in addition to Lustre. In order
2437 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2439 It is normal that files written just before a client crash (less
2440 than 30s) may not yet have been flushed to disk, even for local
2444 Frequency : frequent on thin XT3 nodes
2446 Description: UUID collision on thin XT3 Linux nodes
2447 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2448 randomly, since we relied on an insufficiently-seeded PRNG.
2453 Description: OSS hangs after "All ost request buffers busy"
2454 Details : A deadlock between quota and journal operations caused OSS
2455 hangs after printing "All ost request buffers busy."
2458 Frequency : always on liblustre builds
2460 Description: Cleanup compiler warnings on liblustre
2463 Frequency : always on liblustre builds on XT3
2465 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2469 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2470 > 32000 subdirectories
2471 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2472 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2473 subdirectory count crosses 32000. This will aid e2fsck to
2474 correctly handle more than 32000 subdirectories.
2477 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2479 Description: System hang on startup
2480 Details : This bug allowed the liblustre (e.g. catamount) client to
2481 return to the app before handling all startup RPCs. This
2482 could leave the node unresponsive to lustre network traffic
2483 and manifested as a server ptllnd timeout.
2485 Severity : enhancement
2487 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2488 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2489 boolean to control whether to print peer debug info when a
2490 client's RPC times out.
2495 Description: make O_SYNC working on 2.6 kernels
2496 Details : 2.6 kernels use different method for mark pages for write,
2497 so need add a code to lustre for O_SYNC work.
2502 Description: Failure to close file and release space on NFS
2503 Details : Put inode details into lock acquired in ll_intent_file_open.
2504 Use mdc_intent_lock in ll_intent_open to properly
2505 detect all kind of errors unhandled by mdc_enqueue.
2510 Description: proc file read during shutdown sometimes raced obd removal,
2512 Details : Add lock to prevent obd access after proc file removal.
2515 Frequency : Only for files larger than 4GB on 32-bit clients.
2517 Description: improperly doing page alignment of locks
2518 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2519 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2524 Description: RPCs being resent when they shouldn't be
2525 Details : Some RPCs that should not be resent are being resent. This
2526 can cause inconsistencies in the RPC state machine. Do not
2527 resend such requests.
2530 Frequency : rare, only with NFS export
2532 Description: Crash on NFS re-export node
2533 Details : under very unusual load conditions an assertion is hit in
2534 ll_intent_file_open()
2537 Frequency : only if OST filesystem is corrupted
2539 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2540 Details : for a short time RPCs with bulk IO are in the replay list,
2541 but replay of bulk IOs is unimplemented. If the OST filesystem
2542 is corrupted due to disk cache incoherency and then replay is
2543 started it is possible to trip an assertion. Avoid putting
2544 committed RPCs into the replay list at all to avoid this issue.
2549 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2550 Details : Large single O_DIRECT read and write calls can fail to allocate
2551 a sufficiently large buffer to process the request. In case of
2552 allocation failure the allocation is retried with a smaller
2553 buffer and broken into smaller requests.
2555 Severity : enhancement
2557 Description: Add -o localflock option to simulate old noflock behaviour.
2558 Details : This will achieve local-only flock/fcntl locks coherentness.
2563 Description: versioning check is incomplete
2564 Details : Checking the version difference of client vs. server, report
2565 error if the gap is too big.
2570 Description: add support PG_writeback bit
2571 Details : add support for PG_writeback bit for Lustre, for more carefull
2572 work with page cache in 2.6 kernel. This also fix some deadlocks
2573 and remove hack for work O_SYNC with 2.6 kernel.
2575 Severity : enhancement
2577 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2578 Details : The uninit_groups feature works in conjunction with the kernel
2579 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2580 up the pass1 processing of e2fsck. This is a read-only feature
2581 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2582 mount filesystems that have had this feature enabled.
2584 Severity : enhancement
2586 Description: Improve multi-block allocation algorithm to avoid fragmentation
2587 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2588 allocation locality and avoid filesystem fragmentation.
2590 ------------------------------------------------------------------------------
2592 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2594 * Support for kernels:
2595 2.6.9-42.0.3.EL (RHEL 4)
2596 2.6.5-7.276 (SLES 9)
2597 2.4.21-47.0.1.EL (RHEL 3)
2598 2.6.12.6 vanilla (kernel.org)
2599 2.6.16.21-0.8 (SLES10)
2600 * Recommended e2fsprogs version: 1.39.cfs2-0
2602 * The backwards-compatible /proc/sys/portals symlink has been removed
2603 in this release. Before upgrading, please ensure that you change
2604 any configuration scripts or /etc/sysctl.conf files that access
2605 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2606 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2607 in advance of the upgrade on any system running Lustre 1.4.6 or
2608 newer, since /proc/sys/lnet was added in that version.
2609 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2613 Frequency : only when quota is used
2615 Description: avoid scanning export list for quota master
2616 Details : Change the algorithms to avoid scanning export list in order
2617 to improve the efficiency.
2620 Frequency : MDS failover only, very rarely
2622 Description: "went back in time" messages on mds failover
2623 Details : The greatest transno may be lost when the current operation
2624 finishes with an error (transno==0) and the client's last_rcvd
2625 record is over-written. Save the greatest transno in the
2626 mds_last_transno for this case.
2629 Frequency : always for specific kernels and striping counts
2631 Description: client may get "Matching packet too big" without ACL support
2632 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2633 when trying to access files in certain configurations. The
2634 clients should in fact be denied when mounting because they do
2635 not understand ACLs.
2638 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2640 Description: many clients connecting with IO in progress causes connect timeouts
2641 Details : Avoid synchronous journal commits to avoid delays caused by many
2642 clients connecting/disconnecting when bulk IO is in progress.
2643 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2644 OST_IO_PORTAL to avoid delays behind potentially many pending
2648 Frequency : occasionally with multiple writers to a single file
2650 Description: shared writes to file may result in wrong size reported by stat()
2651 Details : Allow growing of kms when extent lock is cancelled
2654 Frequency : always with random mmap IO to multi-striped file
2656 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2657 Details : Only if the hole is at the end of OST object so that kms is too
2658 small. Fix is to increase kms accordingly in ll_nopage.
2661 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2663 Description: writes to a missing object would leak memory on the OST
2664 Details : If there is an inconsistency between the MDS and OST filesystems,
2665 such that the MDS references an object that doesn't exist, writes
2666 to that object will leak memory due to incorrect cleanup in the
2667 error handling path, eventually running out of memory on the OST.
2672 Description: Creating too long symlink causes lustre errors
2673 Details : Check symlink and name lengths before sending requests to MDS.
2676 Frequency : only if flock is enabled (not on by default)
2678 Description: posix locks not released on fd closure on 2.6.9+
2679 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2680 kernels, this caused such locks not to be released on fd close and
2681 then assertions on fs unmount about still used locks.
2684 Frequency : MDS failover only, very rarely
2686 Description: clients may get ASSERTION(granted_lock != NULL)
2687 Details : When request was taking a long time, and a client was resending
2688 a getattr by name lock request. The were multiple lock requests
2689 with the same client lock handle and
2690 mds_getattr_name->fixup_handle_for_resent_request found one of the
2691 lock handles but later failed with ASSERTION(granted_lock != NULL).
2696 Description: handle->h_buffer_credits > 0, assertion failure
2697 Details : h_buffer_credits is zero after truncate, causing assertion
2698 failure. This patch extends the transaction or creates a new
2702 Frequency : NFS re-export or patchless client
2703 Bugzilla : 11179, 10796
2704 Description: Crash on NFS re-export node (__d_move)
2705 Details : We do not want to hash the dentry if we don't have a lock.
2706 But if this dentry is later used in d_move, we'd hit uninitialised
2707 list head d_hash, so we just do this to init d_hash field but
2708 leave dentry unhashed.
2711 Frequency : NFS re-export or patchless client
2713 Description: NFS exports has problem with symbolic link
2714 Details : lustre client didn't properly install dentry when re-exported
2715 to NFS or running patchless client.
2718 Frequency : NFS re-export or patchless client
2720 Description: Various nfs/patchless fixes.
2721 Details : fixes reuse disconected alias for lookup process - this fixes
2722 warning "find_exported_dentry: npd != pd",
2723 fix permission error with open files at nfs.
2724 fix apply umask when do revalidate.
2727 Frequency : occasional
2729 Description: Crash on NFS re-export node
2730 Details : calling clear_page() on the wrong pointer triggered oops in
2731 generic_mapping_read().
2734 Frequency : rarely, using O_DIRECT IO
2736 Description: unaligned directio crashes client with LASSERT
2737 Details : check for unaligned buffers before trying any requests.
2740 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2742 Description: stale data returned from RAID cache
2743 Details : If only a small amount of IO is done to the RAID device before
2744 reading it again it is possible to get stale data from the RAID
2745 cache instead of reading it from disk.
2748 Frequency : always for sles10 kernel
2750 Description: sles10 support
2751 Details : ll_follow_link: compile fixes and using of nd_set_link
2752 under newer kernels.
2755 Frequency : depends on arch, kernel and compiler version, always on sles10
2758 Description: recursive or deep enough symlinks cause stack overflow
2759 Details : getting rid of large stack-allocated variable in
2763 Frequency : depends on hardware
2765 Description: lustre write performance loss in the SLES10 kernel
2766 Details : the performance loss is caused by using of write barriers in the
2767 ext3 code. The SLES10 kernel turns barrier support on by
2768 default. The fix is to undo that change for ldiskfs.
2770 ------------------------------------------------------------------------------
2772 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2774 * Support for kernels:
2775 2.6.9-42.0.3EL (RHEL 4)
2776 2.6.5-7.276 (SLES 9)
2777 2.4.21-47.0.1.EL (RHEL 3)
2778 2.6.12.6 vanilla (kernel.org)
2782 Frequency : quota enabled and large files being deleted
2784 Description: releasing more than 4GB of quota at once hangs OST
2785 Details : If a user deletes more than 4GB of files on a single OST it
2786 will cause the OST to spin in an infinite loop. Release
2787 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2792 Description: statfs data retrieved from /proc may be stale or zero
2793 Details : When reading per-device statfs data from /proc, in the
2794 {kbytes,files}_{total,free,avail} files, it may appear
2795 as zero or be out of date.
2798 Frequency : systems with MD RAID1 external journal devices
2800 Description: lconf's call to blkid is confused by RAID1 journal devices
2801 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2802 of returning all block devices that match the journal UUID.
2805 Frequency : always, for aggregate stripe size over 4GB
2807 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2808 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2809 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2812 Frequency : always if "lfs find" used on a local file/directory
2814 Description: "lfs find" segfaults if used on a local file/directory
2815 Details : The case where a directory component was not specified wasn't
2816 handled correctly. Handle this properly.
2819 Frequency : always on ppc64
2821 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2822 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2823 perform endianness conversion.
2826 Frequency : rarely (truncate to non-zero file size after write under load)
2827 Bugzilla : 10730, 10687
2828 Description: Files padded with zeros to next 4K multiple
2829 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2830 it is possible that files that are truncated to a non-zero size
2831 immediately after being written are filled with zero bytes beyond
2832 the truncated size. No file data is lost.
2834 Severity : enhancement
2836 Description: Allow recovery/failover for liblustre clients.
2837 Details : liblustre clients were unaware of failover configurations until
2840 Severity : enhancement
2842 Description: user file locks should fail when not mounting with flock option
2843 Details : Set up an error-returning stub in ll_file_operations.lock field
2844 to prevent incorrect behaviour when client is mounted without
2845 flock option. Also, set up properly f_op->flock field for
2849 Frequency : always on ia64
2851 Description: "lfs df" loops on printing out MDS statfs information
2852 Details : The obd_ioctl_data was not initialized and in some systems
2853 this caused a failure during the ioctl that did not return
2854 an error. Initialize the struct and return an error on failure.
2857 Frequency : SLES 9 only
2859 Description: Error of copying files with lustre special EAs as root
2860 Details : Client side always return success for setxattr call for lustre
2861 special xattr (currently only "trusted.lov").
2864 Frequency : rarely on clusters with both ia64+i386 clients
2866 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2867 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2868 clients are doing concurrent writes to the same file it is
2869 possible that the ia64 clients may LASSERT because the OST
2870 extent locks are not PAGE_SIZE aligned. Ensure that grown
2871 locks are always aligned on the request boundary.
2874 Frequency : specific use, occasional
2876 Description: Overwriting in use executable truncates on-disk binary image
2877 Details : If one node attempts to overwrite an executable in use by
2878 another node, we now correctly return ETXTBSY instead of
2879 truncating the file.
2881 Severity : enhancement
2883 Description: Async OSC create to avoid the blocking unnecessarily.
2884 Details : If a OST has no remain object, system will block on the creating
2885 when need to create a new object on this OST. Now, ways use
2886 pre-created objects when available, instead of blocking on an
2887 empty osc while others are not empty. If we must block, we block
2888 for the shortest possible period of time.
2893 Description: chmod on Lustre root is propagated to other clients
2894 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2895 invalid by the follow_mount time.
2898 Frequency : liblustre clients only
2900 Description: Race in 'instant cancel' lock handling could lead to such locks
2901 never to be granted in case of SMP MDS
2902 Details : Do not destroy not yet granted but cbpending locks in
2906 Frequency : replay/resend of open
2908 Description: non null lock assetion failure in mds_intent_policy
2909 Details : Trying to replay/resend lockless open requests resulted in
2910 mds_open() returning 0 with no lock. Now it sets a flag if
2911 a lock is going to be returned.
2913 Severity : enhancement
2915 Description: Checksum enhancements
2916 Details : New checksum enhancements allow for resending RPCs that failed
2919 Severity : enhancement
2921 Description: Tunables on number of dirty pages in cacche
2922 Details : Allow to set limit on number of dirty pages cached.
2927 Description: client crash on unmount - lock still has references
2928 Details : In some error handling cases it was possible to leak a lock
2929 reference on a client while accessing a file. This was not
2930 harmful to the client during operation, but would cause the
2931 client to crash when the filesystem is unmounted.
2934 Frequency : specific case, rare
2936 Description: ETXTBSY on mds though file not in use by client
2937 Details : ETXTBSY is no longer incorrectly returned when attempting to
2938 chmod or chown a directory that the user previously tried to
2939 execute or a currently-executing binary.
2942 Frequency : extremely rare except on liblustre-based clients
2944 Description: Lustre space not freed when files are deleted
2945 Details : Clean up open-unlinked files after client eviction. Previously
2946 the unlink was skipped and the files remained as orphans.
2951 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2952 Details : In some cases it was possible to send a blocking callback to
2953 a client doing a glimpse, even though that client didn't get
2954 a lock granted. When the glimpse lock is cancelled on the OST
2955 the freed lock is left on the waiting list and corrupted the list.
2958 Frequency : all core dumps
2960 Description: Broke core dumps to lustre
2961 Details : Negative dentry may be unhashed if parent does not have UPDATE
2962 lock, but some callers, e.g. do_coredump, expect dentry to be
2963 hashed after successful create, hash it in ll_create_it.
2965 ------------------------------------------------------------------------------
2967 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2969 * Support for kernels:
2970 2.6.9-42.0.2.EL (RHEL 4)
2971 2.6.5-7.276 (SLES 9)
2972 2.4.21-40.EL (RHEL 3)
2973 2.6.12.6 vanilla (kernel.org)
2977 Frequency : always on RHEL 3
2979 Description: Number of open files grows over time
2980 Details : The number of open files grows over time, whether or not
2981 Lustre is started. This was due to a filp leak introduced
2982 by one of our kernel patches.
2984 ------------------------------------------------------------------------------
2986 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2988 * Support for kernels:
2989 2.6.9-42.EL (RHEL 4)
2990 2.6.5-7.276 (SLES 9)
2991 2.4.21-40.EL (RHEL 3)
2992 2.6.12.6 vanilla (kernel.org)
2997 Bugzilla : 5719, 9635, 9792, 9684
2998 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2999 Details : If a server is under heavy load and cannot reply to new
3000 connection requests before the client resends the (re)connect,
3001 the connection handling code can behave badly if two service
3002 threads are concurrently handing separate (re)connections from
3003 the same client. Add better locking to the connection handling
3004 code, and ensure that only a single connection will be processed
3005 for a given client UUID, even if the lock is dropped.
3007 Severity : enhancement
3009 Description: add TCP zero-copy support to kernel
3010 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3011 sends if the hardware supports scatter-gather and checksumming.
3012 This allows socklnd to do client-write and server-read more
3013 efficiently and reduce CPU utilization from skbuf copying.
3016 Frequency : only if NFS exporting from client
3018 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3019 Details : If setting attributes on a file created under NFS that had
3020 never been opened it would be possible to oops the client
3021 if the file had no objects.
3024 Frequency : always for liblustre
3026 Description: liblustre client does MDS+OSTs setattr RPC for each write
3027 Details : When doing a write from a liblustre client, the client
3028 incorrectly issued an RPC to the MDS and each OST the file was
3029 striped over in order to update the timestamps. When writing
3030 with small chunks and many clients this could overwhelm the MDS
3031 with RPCs. In all cases it would slow down the write because
3032 these RPCs are unnecessary.
3034 Severity : enhancement
3036 Description: allow number of MDS service threads to be changed at module load
3037 Details : It is now possible to change the number of MDS service threads
3038 running. Adding "options mds mds_num_threads={N}" to the MDS's
3039 /etc/modprobe.conf will set the number of threads for the next
3040 time Lustre is restarted (assuming the "mds" module is also
3041 reloaded at that time). The default number of threads will
3042 stay the same, 32 for most systems.
3047 Description: OST crash if filesystem is unformatted or corrupt
3048 Details : If an OST is started on a device that has never been formatted
3049 or if the filesystem is corrupt and cannot even mount then the
3050 error handling cleanup routines would dereference a NULL pointer.
3055 Description: NULL pointer deref in llap_from_page.
3056 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3057 incorrect) mapping if the page was truncated/reclaimed while it was
3058 searched for. Check for this condition and skip such pages when
3059 doing readahead. Introduce extra check to llap_from_page() to
3060 verify page->mapping->host is non-NULL (so page is not anonymous).
3063 Frequency : Sometimes when using sys_sendfile
3065 Description: "page not covered by a lock" warnings from ll_readpage
3066 Details : sendfile called ll_readpage without right page locks present.
3067 Now we introduced ll_file_sendfile that does necessary locking
3068 around call to generic_file_sendfile() much like we do in
3072 Frequency : with certain MDS communication failures at client mount time
3074 Description: NULL pointer deref after failed client mount
3075 Details : a client connection request may delayed by the network layer
3076 and not be sent until after the PTLRPC layer has timed out the
3077 request. If the client fails the mount immediately it will try
3078 to clean up before the network times out the request. Add a
3079 reference from the request import to the obd device and delay
3080 the cleanup until the network drops the request.
3083 Frequency : occasionally during client (re)connect
3085 Description: assertion failure during client (re)connect
3086 Details : processing a client connection request may be delayed by the
3087 client or server longer than the client connect timeout. This
3088 causes the client to resend the connection request. If the
3089 original connection request is replied in this interval, the
3090 client may trip an assertion failure in ptlrpc_connect_interpret()
3091 which thought it would be the only running connect process.
3094 Frequency : only with obd_echo servers and clients that are rebooted
3096 Description: kernel BUG accessing uninitialized data structure
3097 Details : When running an obd_echo server it did not start the ping_evictor
3098 thread, and when a client was evicted an uninitialized data
3099 structure was accessed. Start the ping_evictor in the RPC
3100 service startup instead of the OBD startup.
3102 Severity : enhancement
3103 Bugzilla : 10193 (patchless)
3104 Description: Remove dependency on various unexported kernel interfaces.
3105 Details : No longer need reparent_to_init, exit_mm, exit_files,
3106 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3109 Frequency : rare (only users of deprecated and unsupported LDAP config)
3111 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3112 Details : LDAP apparently contains 'lustreName' attributes instead of
3113 'name'. A simple remapping of the name is sufficient.
3116 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3118 Description: waiting_locks_callback trips kernel BUG if client is evicted
3119 Details : Running with the dump_on_timeout debug flag turned on makes
3120 it possible that the waiting_locks_callback() can try to dump
3121 the Lustre kernel debug logs from an interrupt handler. Defer
3122 this log dumping to the expired_lock_main() thread.
3124 Severity : enhancement
3126 Description: Support NFS exporting on 2.6 kernels.
3127 Details : Implement non-rawops metadata methods for NFS server to use without
3128 changing NFS server code.
3131 Frequency : very rare (synthetic metadata workload only)
3133 Description: two racing renames might cause an MDS thread to deadlock
3134 Details : Running the "racer" program may cause one MDS thread to rename
3135 a file from being the source of a rename to being the target of
3136 a rename at exactly the same time that another thread is doing
3137 so, and the second thread has already enqueued these locks after
3138 doing a lookup of the target and is trying to relock them in
3139 order. Ensure that we don't try to re-lock the same resource.
3142 Frequency : only very large systems with liblustre clients
3144 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3145 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3146 all OSTs in parallel. This allows the eviction of stale liblustre
3147 clients to proceed much faster than if they were done in series,
3148 and also offers similar improvements for other set_info RPCs.
3153 Description: excessive CPU usage during initial read phase on client
3154 Details : During the initial read phase on a client, it would agressively
3155 retry readahead on the file, consuming too much CPU and impacting
3156 performance (since 1.4.5.8). Improve the readahead algorithm
3157 to avoid this, and also improve some other common cases (read
3158 of small files in particular, where "small" is files smaller than
3159 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3164 Description: MDS crash when receiving packet with unknown intent.
3165 Details : Do not LBUG in unknown intent case, just return -EFAULT
3167 Severity : enhancement
3168 Bugzilla : 9293, 9385
3169 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3170 Details : Do not serialize getattr (non-intent version) and statfs.
3173 Frequency : occasional, when OST network is overloaded/intermittent
3175 Description: client evicted by OST after bulk IO timeout
3176 Details : If a client sends a bulk IO request (read or write) the OST
3177 may evict the client if it is unresposive to its data GET/PUT
3178 request. This is incorrect if the network is overloaded (takes
3179 too long to transfer the RPC data) or dropped the OST GET/PUT
3180 request. There is no need to evict the client at all, since
3181 the pinger and/or lock callbacks will handle this, and the
3182 client can restart the bulk request.
3185 Frequency : Always when mmapping file with no objects
3187 Description: client crashes when mmapping file with no objects
3188 Details : Check that we actually have objects in a file before doing any
3189 operations on objects in ll_vm_open, ll_vm_close and
3195 Description: Request leak when working with deleted CWD
3196 Details : Introduce advanced request refcount tracking for requests
3197 referenced from lustre intent.
3199 Severity : Enhancement
3201 Description: Cache open file handles on client.
3202 Details : MDS now will return special lock along with openhandle, if
3203 requested and client is allowed to hold openhandle, even if unused,
3204 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3205 closing files for every read/write openration.
3207 Severity : Enhancement
3209 Description: Cache open negative dentries on client when possible.
3210 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3211 negative dentries on lock revocation.
3216 Description: Remounting a client read-only wasn't possible with a zconf mount
3217 Details : It wasn't possible to remount a client read-only with llmount.
3219 Severity : enhancement
3220 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3221 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3222 MPICH is included to add Lustre-specific ADIO interfaces.
3223 This is based closely on the UFS ADIO layer and only differs
3224 in file creation, in order to allow the OST striping to be set.
3225 This is user-contributed code and not supported by CFS.
3230 Description: extended inode attributes (immutable, append-only) work improperly
3231 when 2.4 and 2.6 kernels are used on client/server or vice versa
3232 Details : Introduce kernel-independent values for these flags.
3234 Severity : enhancement
3237 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3238 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3239 granularity. Now, Fractional MB granularity is be supported,
3240 this is very useful for low memory system.
3242 Severity : enhancement
3244 Description: Getattr by fid
3245 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3246 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3247 have locks and avoids getattr rpc requests for it on every stat.
3250 Frequency : Always, for filesystems larger than 2TB
3252 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3253 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3254 filesystems larger than 2TB to be mounted with mballoc enabled.
3257 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3259 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3260 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3261 filesystems larger than 2TB is mounted then the kernel will
3262 silently corrupt the start of the filesystem. CONFIG_LBD is
3263 enabled for all CFS-supported kernels, but the possibility of
3264 this happening with a modified kernel config exists.
3266 Severity : enhancement
3268 Description: add client O_DIRECT support for 2.6 kernels
3269 Details : It is now possible to do O_DIRECT reads and writes to files
3270 in the Lustre client mountpoint on 2.6 kernel clients.
3272 Severity : enhancement
3274 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3275 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3276 parallel, not waiting for response from every OST before sending
3277 a rpc to the next OST.
3282 Description: setattr vs write race when updating file timestamps
3283 Details : Client processes that update a file timestamp into the past
3284 right after writing to the file (e.g. tar) it is possible that
3285 the updated file modification time can be reset to the current
3286 time due to a race between processing the setattr and write RPC.
3288 Severity : enhancement
3290 Description: Bring 'lfs find' closer in line with regular Linux find.
3291 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3292 -print0 options and obtains all the needed info through the lustre
3295 Severity : enhancement
3297 Description: support up to 1024 configured devices on one node
3298 Details : change obd_dev array from statically allocated to dynamically
3299 allocated structs as they are first used to reduce memory usage
3304 Description: Flush dirty partially truncated pages during truncate
3305 Details : Immediatelly flush partially truncated pages in filter_setattr,
3306 this way we completely avoid having any pages in page cache on OST
3307 and can retire ugly workarounds during writes to flush such pages.
3312 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3313 Details : protect inode from truncation within vfs_unlink() context
3314 just take a reference before calling vfs_unlink() and release it
3315 when parent's i_sem is free.
3320 Description: last_id value checked outside lock on OST caused LASSERT failure
3321 Details : If there were multiple MDS->OST object precreate requests in
3322 flight, it was possible that the OST's last object id was checked
3323 outside a lock and incorrectly tripped an assertion. Move checks
3324 inside locks, and discard old precreate requests.
3327 Frequency : always, if extents are used on OSTs
3329 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3330 Details : OSTs using the extents format would not zero the high 16 bits of
3331 the index physical block number. This is not a problem for any
3332 OST filesystems smaller than 16TB, and no kernels support ext3
3333 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3334 new/modified files) and can be fixed for existing filesystems
3335 with e2fsprogs-1.39-cfs1.
3340 Description: import connection selection may be incorrect if timer wraps
3341 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3342 import connections to be ignored if the 32-bit jiffies counter
3343 wraps. Use a 64-bit jiffies counter.
3346 Frequency : very large clusters immediately after boot
3348 Description: LNET request buffers exhausted under heavy short-term load
3349 Details : If a large number of client requests are generated on a service
3350 that has previously never seen so many requests it is possible
3351 that the request buffer growth cannot keep up with the spike in
3352 demand. Instead of dropping incoming requests, they are held in
3353 the LND until the RPC service can accept more requests.
3356 Frequency : Sometimes during replay
3358 Description: Assertion failure in ll_local_open after replay.
3359 Details : If replay happened on an open request reply before we were able
3360 to set replay handler, reply will become not swabbed tripping the
3361 assertion in ll_local_open. Now we set the handler right after
3362 recognising of open request
3365 Frequency : very rare
3367 Description: kernel reports "badness in vsnprintf"
3368 Details : Reading from the "recovery_status" /proc file in small chunks
3369 may cause a negative length in lprocfs_obd_rd_recovery_status()
3370 call to vsnprintf() (which is otherwise harmless). Exit early
3371 if there is no more space in the output buffer.
3373 Severity : enhancement
3375 Description: clear OBD RPC statistics by writing to them
3376 Details : It is now possible to clear the OBD RPC statistics by writing
3377 to the "stats" file.
3382 Description: Client mtime is not the same on different clients after utimes
3383 Details : In some cases, the client was using the utimes() syscall on
3384 a file cached on another node. The clients now validate the
3385 ctime from the MDS + OSTs to determine which one is right.
3390 Description: Inability to activate failout mode
3391 Details : lconf script incorrectly assumed that in pythong string's numeric
3392 value is used in comparisons.
3395 Frequency : always with multiple stripes per file
3397 Description: Inefficient object allocation for mutli-stripe files
3398 Details : When selecting which OSTs to stripe files over, for files with
3399 a stripe count that divides evenly into the number of OSTs,
3400 the MDS is always picking the same starting OST for each file.
3401 Return the OST selection heuristic to the original design.
3406 Description: mount failures may take full timeout to return an error
3407 Details : Under some heavy load conditions it is possible that a
3408 failed mount can wait for the full obd_timeout interval,
3409 possibly several minutes, before reporting an error.
3410 Instead return an error as soon as the status is known.
3412 Frequency : quota enabled and large files being deleted
3414 Description: releasing more than 4GB of quota at once hangs OST
3415 Details : If a user deletes more than 4GB of files on a single OST it
3416 will cause the OST to spin in an infinite loop. Release
3417 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3422 Description: statfs data retrieved from /proc may be stale or zero
3423 Details : When reading per-device statfs data from /proc, in the
3424 {kbytes,files}_{total,free,avail} files, it may appear
3425 as zero or be out of date.
3428 Frequency : systems with MD RAID1 external journal devices
3430 Description: lconf's call to blkid is confused by RAID1 journal devices
3431 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3432 of returning all block devices that match the journal UUID.
3435 Frequency : always, for aggregate stripe size over 4GB
3437 Description: assertion fails when trying to use 4GB stripe size
3438 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3439 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3442 Frequency : always on ppc64
3444 Description: the first write on an ext3 filesystem with mballoc got stuck
3445 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3446 perform endianness conversion.
3448 ------------------------------------------------------------------------------
3450 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3452 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3453 INTEROPERATE with older versions automatically. Please read the
3454 user documentation before upgrading any part of a live system.
3455 * WARNING: Lustre networking configuration changes are required with
3456 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3459 * Support for kernels:
3460 2.6.9-22.0.2.EL (RHEL 4)
3461 2.6.5-7.244 (SLES 9)
3462 2.6.12.6 vanilla (kernel.org)
3465 Severity : enhancement
3466 Bugzilla : 7981/8208
3467 Description: Introduced Lustre Networking (LNET)
3468 Details : LNET is new networking infrastructure for Lustre, it includes
3469 a reorganized network configuration mode (see the user
3470 documentation for full details) as well as support for routing
3471 between different network fabrics. Lustre Networking Devices
3472 (LNDs) for the supported network fabrics have also been
3473 created for this new infrastructure.
3475 Severity : enhancement
3476 Description: Introduced Access control lists
3477 Details : clients can set ACLs on files and directories in order to have
3478 more fine-grained permissions than the standard Unix UGO+RWX.
3479 The MDS must be started with the "-o acl" mount option.
3481 Severity : enhancement
3482 Description: Introduced filesystem quotas
3483 Details : Administrators may now establish per-user quotas on the
3486 Severity : enhancement
3488 Description: Configuration change for the XT3
3489 The PTLLND is now used to run Lustre over Portals on the XT3
3490 The configure option(s) --with-cray-portals are no longer used.
3491 Rather --with-portals=<path-to-portals-includes> is used to
3492 enable building on the XT3. In addition to enable XT3 specific
3493 features the option --enable-cray-xt3 must be used.
3498 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3499 Details : A race between allocating a new llcd and re-getting the llcd_lock
3500 allowed another thread to grab newly-allocated llcd.
3502 Severity : enhancement
3504 Description: 2.6 OST async journal commit and locking fix to improve performance
3505 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3506 2.6 kernels are now async as they already were in 2.4 kernels so
3507 that they can commit concurrently with the network bulk transfer.
3508 For block-allocated files the filter allocation semaphore is held
3509 to avoid filesystem fragmentation during allocation. BKL lock
3510 removed for 2.6 xattr operations where it is no longer needed.
3515 Description: lconf incorrectly determined whether two IP networks could talk
3516 Details : In some more complicated routing and multiple-network
3517 configurations, lconf will avoid trying to make a network
3518 connection to a disjoint part of the IP space. It was doing the
3519 math incorrectly for one set of cases.
3524 Description: Fix for potential infinite loop processing records in an llog.
3525 Details : If an llog record is corrupted/zeroed, it is possible to loop
3526 forever in llog_process(). Validate the llog record length
3527 and skip the remainder of the block on error.
3530 Frequency : occasional (liblustre only)
3532 Description: liblustre could not open files whose last component is a symlink
3533 Details : sysio_path_walk() would incorrectly pass the open intent to
3534 intermediate path components.
3537 Frequency : rare (liblustre only with non-standard tuning)
3538 Bugzilla : 7201 (7350)
3539 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3540 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3541 after it has been released, only do this for referenced locks
3543 Severity : enhancement
3545 Description: specifying an (invalid) directory default stripe_size of -1
3546 would reset the directory default striping
3547 Details : stripe_size -1 was used internally to signal directory stripe
3548 removal, now use "all default" to signal dir stripe removal
3549 as a directory striping of "all default" is not useful
3552 Frequency : common for large clusters running liblustre clients
3554 Description: doing an ls when liblustre clients are running is slow
3555 Details : sending a glimpse AST to a liblustre client waits for every AST
3556 to time out, as liblustre clients will not respond. Since they
3557 cannot cache data we refresh the OST lock LVB from disk instead.
3559 Severity : enhancement
3561 Description: doing an ls at the same time as file IO can be slow
3562 Details : enqueue and other "small" requests can be blocked behind many
3563 large IO requests. Create a new OST IO portal for non-IO
3564 requests so they can be processed faster.
3567 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3569 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3570 created file with mode 000
3571 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3572 allow an ftruncate() as a non-root user to a file with mode 000.
3573 Lustre now respects this flag to disable mode checks when
3574 truncating a file owned by the user
3577 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3580 Description: Revoking locks from clients that went dead or catatonic might take
3582 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3583 cancellation of such locks instant on servers without waiting for
3584 any reply from clients. Clients drops these locks when cancel
3585 notification from server is received without replying.
3588 Frequency : liblustre-only, when liblustre client dies or becomes busy
3590 Description: Doing ls on Linux clients can take a long time with active
3592 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3593 granting such locks to it in the first place if possible. Locks
3594 are taken by proxy on the OST during the read or write and
3595 dropped immediately afterward. Add connect flags handling, do
3596 not grant locks to liblustre clients for glimpse ASTs.
3598 Severity : enhancement
3600 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3601 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3602 sequential reads and behaves badly with applications doing random
3603 reads. Improve it by reading ahead at least read region, and
3604 avoiding excessive large RPC for small reads.
3606 Severity : enhancement
3608 Description: Creating more than 1000 files for a single job may cause a load
3609 imbalance on the OSTs if there are also a large number of OSTs.
3610 Details : qos_prep_create() uses an OST index reseed value that is an
3611 even multiple of the number of available OSTs so that if the
3612 reseed happens in the middle of the object allocation it will
3613 still utilize the OSTs as uniformly as possible.
3618 Description: OST or MDS may oops in ping_evictor_main()
3619 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3620 but doesn't restart at beginning of obd_exports_timed list
3623 Severity : enhancement
3625 Description: improve by-nid export eviction on the MDS and OST
3626 Details : allow multiple exports with the same NID to be evicted at one
3627 time without re-searching the exports list.
3630 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3632 Description: MDS may oops in groups_free()
3633 Details : in rare race conditions a newly allocated group_info struct is
3634 freed again, and this can be NULL. The 2.4 compatibility code
3635 for groups_free() checked for a NULL pointer, but 2.6 did not.
3638 Frequency : common for liblustre clients doing little filesystem IO
3639 Bugzilla : 9352, 7313
3640 Description: server may evict liblustre clients accessing contended locks
3641 Details : if a client is granted a lock or receives a completion AST
3642 with a blocking AST already set it would not reply to the AST
3643 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3646 Frequency : lfs setstripe, only systems with more than 160 OSTs
3648 Description: unable to set striping with a starting offset beyond OST 160
3649 Details : llapi_create_file() incorrectly limited the starting stripe
3650 index to the maximum single-file stripe count.
3653 Frequency : LDAP users only
3655 Description: lconf did not handle in-kernel recovery with LDAP properly
3656 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3658 Severity : enhancement
3660 Description: bind OST threads to NUMA nodes to improve performance
3661 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3662 node and do their allocations there to localize memory access
3664 Severity : enhancement
3666 Description: llmount can determine client NID directly from Myrinet (GM)
3667 Details : the client NID code from gmnalnid was moved directly into
3668 llmount, removing the need to use this or specifying the
3669 client NID explicitly when mounting GM clients with zeroconf
3672 Frequency : if client is started with down MDS
3674 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3675 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3676 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3677 ptlrpc_import_delayed_req() because we didn't check if the
3678 request was interrupted, and we also didn't break out of the
3679 event loop if there was no timeout
3684 Description: data loss during non-page-aligned writes to a single file from
3685 both multiple nodes and multiple threads on one node at same time
3686 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3687 inconsistency led to false short-reads, that were cached and later
3688 used by ->prepare_write() to fill in partially written page,
3689 leading to data loss.
3692 Frequency : always, if lconf --abort_recovery used
3694 Description: lconf --abort_recovery fails with 'Operation not supported'
3695 Details : lconf was attempting to abort recovery on the MDT device and not
3698 Severity : enhancement
3700 Description: remove cleanup logs
3701 Details : replace lconf-generated cleanup logs with lustre internal
3702 cleanup routines. Eliminates the need for client-cleanup and
3705 Severity : enhancement
3707 Description: add support for EAs (user and system) on lustre filesystems
3708 Details : it is now possible to store extended attributes in the Lustre
3709 client filesystem, and with the user_xattr mount option it
3710 is possible to allow users to store EAs on their files also
3712 Severity : enhancement
3714 Description: Add possibility (config option) to show minimal available OST free
3716 Details : When compiled with --enable-mindf configure option, statfs(2)
3717 (and so, df) will return least minimal free space available from
3718 all OSTs as amount of free space on FS, instead of summary of
3719 free spaces of all OSTs.
3721 Severity : enhancement
3723 Description: do not expand extent locks acquired on OST-side
3724 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3725 by server: they are not cached anyway.
3728 Frequency : when mmap is used/binaries executed from Lustre
3730 Description: Unmmap pages before throwing them away from read cache.
3731 Details : llap_shrink cache now attempts to unmap pages before discarding
3732 them (if unmapping failed - do not discard). SLES9 kernel has
3733 extra checks that trigger if this unmapping is not done first.
3738 Description: lconf didn't resolve symlinks before checking to see whether a
3739 given mountpoint was already in use
3742 Frequency : when migrating failover services
3743 Bugzilla : 6395, 9514
3744 Description: When migrating a subset of services from a node (e.g. failback
3745 from a failover service node) the remaining services would
3746 time out and evict clients.
3747 Details : lconf --force (implied by --failover) sets the global obd_timeout
3748 to 5 seconds in order to quickly disconnect, but this caused
3749 other RPCs to time out too quickly. Do not change the global
3750 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3752 Severity : enhancement
3753 Frequency : if MDS is started with down OST
3754 Bugzilla : 9439,5706
3755 Description: Allow startup/shutdown of an MDS without depending on the
3756 availability of the OSTs.
3757 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3758 Add appropriate locking and lov-osc refcounts for safe
3759 cleaning. Add osc abort_inflight calls in case the
3760 synchronize never started.
3763 Frequency : occasional (Cray XT3 only)
3765 Description: root not authorized to access files in CRAY_PORTALS environment
3766 Details : The client process capabilities were not honoured on the MDS in
3767 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3768 been accessed by an authorized user then root was able to access
3769 the file on the local client also. The root user capabilities
3770 are now allowed on the MDS, as this environment has secure UID.
3773 Frequency : occasional
3775 Description: ldiskfs "too long searching" message happens too often
3776 Details : A debugging message (otherwise harmless) prints too often on
3777 the OST console. This has been reduced to only happen when
3778 there are fragmentation problems on the filesystem.
3783 Description: Division by zero in statfs when all OSCs are inactive
3784 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3785 lov_statfs divides by value returned by lov_get_stripecnt().
3789 Bugzilla : 9489, 3273
3790 Description: First write from each client to each OST was only 4kB in size,
3791 to initialize client writeback cache, which caused sub-optimal
3792 RPCs and poor layout on disk for the first writen file.
3793 Details : Clients now request an initial cache grant at (re)connect time
3794 and so that they can start streaming writes to the cache right
3795 away and always do full-sized RPCs if there is enough data.
3796 If the OST is rebooted the client also re-establishes its grant
3797 so that client cached writes will be honoured under the grant.
3802 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3803 Details : Now I/O RPCs go to different portal number and (presumably) fast
3804 lock requests (and glimses) and other RPCs get their own service
3805 threads pool that should be able to service those RPCs
3808 Severity : enhancement
3810 Description: Ability to exchange lustre version between client and servers and
3811 issue warnings at client side if client is too old. Also for
3812 liblustre clients there is ability to refuse connection of too old
3814 Details : New 'version' field is added to connect data structure that is
3815 filled with version info. That info is later checked by server and
3819 Frequency : rare, liblustre only.
3820 Bugzilla : 9296, 9581
3821 Description: Two simultaneous writes from liblustre at offset within same page
3822 might proceed at the same time overwriting eachother with stale
3824 Details : I/O lock withing llu_file_prwv was released too early, before data
3825 actually was hitting the wire. Extended lock-holding time until
3826 server acknowledges receiving data.
3829 Frequency : extremely rare. Never observed in practice.
3831 Description: avoid generating lustre_handle cookie of 0.
3832 Details : class_handle_hash() generates handle cookies by incrementing
3833 global counter, and can hit 0 occasionaly (this is unlikely, but
3834 not impossible, because initial value of cookie counter is
3835 selected randonly). Value of 0 is used as a sentinel meaning
3836 "unassigned handle" --- avoid it. Also coalesce two critical
3837 sections in this function into one.
3839 Severity : enhancement
3841 Description: allow liblustre clients to delegate truncate locking to OST
3842 Details : To avoid overhead of locking, liblustre client instructs OST to
3843 take extent lock in ost_punch() on client's behalf. New connection
3844 flag is added to handle backward compatibility.
3846 Severity : enhancement
3847 Bugzilla : 4928, 7341, 9758
3848 Description: allow number of OST service threads to be specified
3849 Details : a module parameter allows the number of OST service threads
3850 to be specified via "options ost ost_num_threads={N}" in the
3851 OSS's /etc/modules.conf or /etc/modprobe.conf.
3855 Bugzilla : 6146, 9635, 9895
3856 Description: servers crash with bad pointer in target_handle_connect()
3857 Details : In rare cases when a client is reconnecting it was possible that
3858 the connection request was the last reference for that export.
3859 We would temporarily drop the export reference and get a new
3860 one, but this may have been the last reference and the export
3861 was just destroyed. Get new reference before dropping old one.
3863 Severity : enhancement
3864 Frequency : if client is started with failover MDS
3866 Description: Allow multiple MDS hostnames in the mount command
3867 Details : Try to read the configuration from all specified MDS
3868 hostnames during a client mount in case the "primary"
3871 Severity : enhancement
3873 Description: Stop sending data to evicted clients as soon as possible.
3874 Details : Check if the client we are about to send or are sending data to
3875 was evicted already. (Check is done every second of waiting,
3876 for which l_wait_event interface was extended to allow checking
3877 of exit condition at specified intervals).
3880 Frequency : rare, normally only when NFS exporting is done from client
3882 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3884 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3887 Frequency : always, if srand() is called before liblustre initialization
3889 Description: Liblustre uses system PRNG disturbing its usage by user application
3890 Details : Introduce internal to lustre fast and high-quality PRNG for
3891 lustre usage and make liblustre and some other places in generic
3892 lustre code to use it.
3894 Severity : enhancement
3895 Bugzilla : 9477, 9557, 9870
3896 Description: Verify that the MDS configuration logs are updated when xml is
3897 Details : Check if the .xml configuration logs are newer than the config
3898 logs stored on the MDS and report an error if this is the case.
3899 Request --write-conf, or allow starting with --old_conf.
3901 Severity : enhancement
3903 Description: Handle symlinks in the path when checking if Lustre is mounted.
3904 Details : Resolve intermediate symlinks when checking if a client has
3905 mounted a filesystem to avoid duplicate client mounts.
3910 Description: lconf can hit an error exception but still return success.
3911 Details : The lconf command catches the Command error exception at the top
3912 level script context and will exit with the associated exit
3913 status, but doesn't ensure that this exit status is non-zero.
3918 Description: failure of ptlrpc thread startup can cause oops
3919 Details : Starting a ptlrpc service thread can fail if there are a large
3920 number of threads or the server memory is very fragmented.
3921 Handle this without oopsing.
3924 Frequency : always, only if liblustre and non-default acceptor port was used
3926 Description: liblustre cannot connect to servers with non-default acceptor port
3927 Details : tcpnal_set_default_params() was not called and was therefore
3928 ignoring the environment varaible TCPNAL_PORT, as well as other
3929 TCPNAL_ environment variables
3934 Description: two objects could be created on the same OST for a single file
3935 Details : If an OST is down, in some cases it was possible to create two
3936 objects on a single OST for a single file. No problems other
3937 than potential performance impact and spurious error messages.
3941 Bugzilla : 5681, 9562
3942 Description: Client may oops in ll_unhash_aliases
3943 Details : Client dcache may become inconsistent in race condition.
3944 In some cases "getcwd" can fail if the current directory is
3950 Description: Inode refcounting problems in NFS export code
3951 Details : link_raw functions used to call d_instantiate without obtaining
3952 extra inode reference first.
3956 Bugzilla : 9942, 9903
3957 Description: Referencing freed requests leading to crash, memleaks with NFS.
3958 Details : We used to require that call to ll_revalidate_it was always
3959 followed by ll_lookup_it. Also with revalidate_special() it is
3960 possible to call ll_revalidate_it() twice for the same dentry
3961 even if first occurence returned success. This fix changes semantic
3962 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3963 reference on a request referred from the intent.
3964 ll_intent_release() then releases such a request.
3967 Frequency : rare, normally benchmark loads only
3969 Description: unlinked inodes were kept in memory on the client
3970 Details : If a client is repeatedly creating and unlinking files it
3971 can accumulate a lot of stale inodes in the inode slab cache.
3972 If there is no other client load running this can cause the
3973 client node to run out of memory. Instead flush old inodes
3974 from client cache that have the same inode number as a new inode.
3977 Frequency : SLES9 2.6.5 kernel and long filenames only
3978 Bugzilla : 9969, 10379
3979 Description: utime reports stale NFS file handle
3980 Details : SLES9 uses out-of-dentry names in some cases, which confused
3981 the lustre dentry revalidation. Change it to always use the
3985 Frequency : rare, unless heavy write-truncate concurrency is continuous
3986 Bugzilla : 4180, 6984, 7171, 9963, 9331
3987 Description: OST becomes very slow and/or deadlocked during object unlink
3988 Details : filter_destroy() was holding onto the parent directory lock
3989 while truncating+unlinking objects. For very large objects this
3990 may block other threads for a long time and slow overall OST
3991 responsiveness. It may also be possible to get a lock ordering
3992 deadlock in this case, or run out of journal credits because of
3993 the combined truncate+unlink. Solution is to do object truncate
3994 first in one transaction without parent lock, and then do the
3995 final unlink in a new transaction with the parent lock. This
3996 reduces the lock hold time dramatically.
3999 Frequency : rare, 2.4 kernels only
4001 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4002 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4003 before unmounting their filesystem in order to avoid deadlock.
4004 The kernel_locked() function in 2.4 kernels only checks whether
4005 the kernel lock is held, not whether it is this process that is
4006 holding it as 2.6 kernels do.
4011 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4012 Details : The client ptlrpc code may be trying to reconnect to a down
4013 server before a previous connection attempt has timed out.
4014 Increase the reconnect interval to be longer than the connection
4015 timeout interval to avoid sending duplicate connections to
4019 Frequency : echo_client brw_test command
4021 Description: fix echo_client to work with OST preallocated code
4022 Details : OST preallocation code (5137) didn't take echo_client IO path
4023 into account: echo_client calls filter methods outside of any
4024 OST thread and, hence, there is no per-thread preallocated
4025 pages and buffers to use. Solution: hijack pga pages for IO. As
4026 a byproduct, this avoids unnecessary data copying.
4030 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4031 Description: Client can oops in mdc_commit_close() after open replay
4032 Details : It was possible for the MDS to return an open request with no
4033 transaction number in mds_finish_transno() if the client was
4034 evicted, but without actually returning an error. Clients
4035 would later try to replay that open and may trip an assertion
4036 Simplify the client close codepath, and always return an error
4037 from the MDS in case the open is not successful.
4040 Frequency : rare, 2.6 OSTs only
4042 Description: OST may deadlock under high load on fragmented files
4043 Details : If there was a heavy load and highly-fragmented OST filesystems
4044 it was possible to have all the OST threads deadlock waiting on
4045 allocation of biovecs, because the biovecs were not released
4046 until the entire RPC IO was completed. Instead, release biovecs
4047 as soon as they are complete to ensure forward IO progress.
4049 Severity : enhancement
4051 Description: Support for specifying external journal device at mount
4052 Details : If an OST or MDS device is formatted with an external journal
4053 device, this device major/minor is stored in the ext3 superblock
4054 and may not be valid for failover. Allow detecting and
4055 specifying the external journal at mount time.
4060 Description: Mounting an MDS with pending unlinked files may cause oops
4061 Details : target_finish_recovery() calls mds_postrecov() which returned
4062 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4063 considers this an error and immediately begins cleaning up the
4064 lov, just after starting the mds_lov process
4066 Severity : enhancement
4068 Description: Implement 'lfs df' to report actual free space on per-OST basis
4069 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4070 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4071 usage of objects; '-h' to report in human readable format.
4073 ------------------------------------------------------------------------------
4075 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4082 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4083 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4084 at startup during orphan unlinking. Instead, skip buddy update
4085 before setup, as it will be regenerated after recovery is complete.
4090 Description: If an OST is inactive, its locks might reference stale inodes.
4091 Details : lov_change_cbdata() must iterate over all namespaces, even if
4092 they are inactive to clear inode references from the lock.
4094 Severity : enhancement
4095 Frequency : occasional, if non-standard max_dirty_mb used
4097 Description: Client will block write RPCs if not enough grant
4098 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4099 then the client will block writes while waiting for another RPC
4100 to complete instead of consuming its dirty limit. With change
4101 we get improved performance when max_dirty_mb is small.
4103 Severity : enhancement
4104 Bugzilla : 3389, 6253
4105 Description: Add support for supplementary groups on the MDS.
4106 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4107 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4108 lookups for user supplementary groups into a cache.
4112 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4113 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4114 for liblustre O_EXCL is enforced by MDS.
4118 Description: readdir never returns NULL in liblustre.
4119 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4120 dentry in struct dirent.
4124 Frequency : liblustre only, depends on application IO pattern
4125 Description: liblustre clients evicted if not contacting servers
4126 Details : Don't put liblustre clients into the ping_evictor list, so
4127 they will not be evicted by the pinger ever.
4129 Severity : enhancement
4131 Description: Add ability to evict clients by NID from MDS.
4132 Details : By echoing "nid:$NID" string into
4133 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4134 $NID would be instantly evicted from this MDS and from all active
4135 OSTs connected to it.
4139 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4140 Details : lookup_it_finish() used to query file size from OSTs that was not
4145 Description: service threads change working directory to that of init
4146 Details : Starting lustre service threads may pin the working directory
4147 of the parent thread, making that filesystem busy. Threads
4148 now change to the working directory of init to avoid this.
4152 Frequency : during shutdown only
4153 Description: shutdown with a failed MDS or OST can cause unmount to hang
4154 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4159 Frequency : 2.6 only
4160 Description: chmod/chown may include an extra supplementary group
4161 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4162 supplementary group and if none is specified this is used.
4165 Bugzilla : 5479 (6816)
4167 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4168 Details : If lookup is in progress on a file that is unlinked we might try
4169 to revalidate the inode and fail in revalidate after lookup is
4170 complete and ll_file_open() enqueues the open again but
4171 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4174 Frequency : always, if lconf --abort_recovery used
4176 Description: lconf --abort_recovery fails with 'Operation not supported'
4177 Details : lconf was attempting to abort recovery on the MDT device and not
4180 ------------------------------------------------------------------------------
4182 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4187 Frequency : rare (only unsupported configurations with a node running as an
4189 Bugzilla : 6514, 5137
4190 Description: Mounting a Lustre file system on a node running as an OST could
4192 Details : OSTs now preallocates memory needed to write out data at
4193 startup, instead of when needed, to avoid having to
4194 allocate memory in possibly low memory situations.
4195 Specifically, if the file system is mounted on on OST,
4196 memory pressure could force it to try to write out data,
4197 which it needed to allocate memory to do. Due to the low
4198 memory, it would be unable to do so and the node would
4199 become unresponsive.
4201 Severity : enhancement
4203 Description: Addition of lconf --service command line option
4204 Details : lconf now accepts a '--service <arg>' option, which is
4205 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4207 Severity : enhancement
4209 Description: Failover mode is now the default for OSTs.
4210 Details : By default, OSTs will now run in failover mode. To return to
4211 the old behaviour, add '--failout' to the lmc line for OSTs.
4213 Severity : enhancement
4215 Description: Health checks are now provided for MDS and OSTs
4216 Details : Additional detailed health check information on MSD and OSTs
4217 is now provided through the procfs health_check value.
4220 Frequency : occasional, depends on IO load
4222 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4223 after numerous create/delete cycles
4224 Details : The ext3 inode allocation policy would not allocate new inodes
4225 very well on the OSTs because there are no new directories
4226 being created. Instead we look for groups with free space if
4227 the parent directories are nearly full.
4232 Description: Network or server problems during mount may cause partially
4233 mounted clients instead of returning an error.
4234 Details : The config llog parsing code may overwrite the error return
4235 code during mount error handling, returning success instead
4241 Description: MDS can fail to allocate large reply buffers
4242 Details : After long uptimes the MDS can fail to allocate large reply
4243 buffers (e.g. zconf client mount config records) due to memory
4244 fragmentation or consumption by the buffer cache. Preallocate
4245 some large reply buffers so that these replies can be sent even
4246 under memory pressure.
4250 Frequency : rare (liblustre)
4251 Description: fsx running with liblustre complained that using truncate() to
4252 extend the file doesn't work. This patch corrects that issue.
4253 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4254 ATTR_SIZE and lsm use in llu_setattr_raw.
4258 Frequency : rare, only 2.6 kernels
4259 Description: Unusual file access patterns on the MDS may result in inode
4260 data being lost in very rare circumstances.
4261 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4262 code in the 2.6 kernel caused the inode and/or EA data not to
4263 be read from disk, causing single-file corruption.
4267 Frequency : rare, only 2.6 filesystems using extents
4268 Description: Heavy concurrent write and delete load may cause data corruption.
4269 Details : It was possible under high-load situations to have an extent
4270 metadata block in the block device cache from a just-unlinked
4271 file overwrite a newly-allocated data block. We now unmap any
4272 metadata buffers that alias just-allocated data blocks.
4276 Frequency : filesystems with default stripe_count larger than 77
4277 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4278 Details : lconf specifies an inode size of 4096 bytes when the default
4279 stripe_count is larger than 77. This conflicts with the default
4280 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4281 this case to avoid pinning too much memory for large EAs.
4283 ------------------------------------------------------------------------------
4285 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4290 Frequency : rare (extremely heavy IO load with hundreds of clients)
4292 Description: Client is evicted, gets IO error writing to file
4293 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4294 caused clients to be evicted for AST timeouts. The fixes in
4295 bug 5192 mean we no longer need to have such short AST timeouts
4296 so ldlm_timeout has been increased.
4299 Frequency : occasional during --force or --failover shutdown under load
4300 Bugzilla : 5949, 4834
4301 Description: Server oops/LBUG if stopped with --force or --failover under load
4302 Details : a collection of import/export refcount and cleanup ordering
4303 issues fixed for safer force cleanup
4306 Frequency : only filesystems larger than 120 OSTs
4307 Bugzilla : 5990, 6223
4308 Description: lfs getstripe would oops on a very large filesystem
4309 Details : lov_getconfig used kfree on vmalloc'd memory
4312 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4313 Bugzilla : 6242, 6243
4314 Description: reading from files that had been truncated to a non-zero size
4315 but never opened returned no data
4316 Details : ll_file_read() reads zeros from no-object files to EOF
4321 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4322 Details : There's little harm in aborting MDS/OSS recovery and letting it
4323 try again, so I removed the LASSERT and return an error instead.
4325 Severity : enhancement
4327 Description: New debugging infrastructure for tracking down data corruption
4328 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4329 (b) cover more of the client-side code path, and (c) try to narrow
4330 down where problems occurred
4334 Bugzilla : 3819, 4364, 4397, 6313
4335 Description: Racing close and eviction MDS could cause assertion in mds_close
4336 Details : It was possible to get multiple mfd references during close and
4337 client eviction, leading to one thread referencing a freed mfd.
4339 Severity: : enhancement
4340 Bugzilla : 3262, 6359
4341 Description: Attempts to reconnect to servers are now more aggressive.
4342 Details : This builds on the enhanced upcall-less recovery that was added
4343 in 1.4.2. When trying to reconnect to servers, clients will
4344 now try each server in the failover group every 10 seconds. By
4345 default, clients would previously try one server every 25 seconds.
4350 Description: After recovery, certain operations trigger a failed
4351 assertion on a client.
4352 Details : Failing over an mds, using lconf -d --failover, while a
4353 client was doing a readdir() call would cause the client to
4354 LBUG after recovery completed and the readdir() was resent.
4356 Severity : enhancement
4358 Description: Default groups are now added by lconf
4359 Details : You can now run lconf --group <servicename> without having to
4360 manually add groups with lmc.
4363 Frequency : occasional
4365 Description: Nodes with an elan id of 0 trigger a failed assertion
4368 Frequency : always when accessing e.g. tty/console device nodes
4370 Description: tty and some other devices nodes cannot be used on lustre
4371 Details : file's private_data field is used by device data and lustre
4372 values in there got lost. New field was added to struct file to
4373 store fs-specific private data.
4376 Frequency : when exporting Lustre via NFS
4378 Description: NFSD failed occasionally when looking up a path component
4379 Details : NFSD is looking up ".." which was broken in ext3 directories
4380 that had grown large enough to become hashed.
4383 Frequency : Clusters with multiple interfaces not on the same subnet
4385 Description: Nodes will repeatedly try to reconnect to an interface which it
4386 cannot reach and report an error to the log.
4387 Details : Extra peer list entries will be created by lconf with some peers
4388 unreachable. lconf now validates the peer before adding it.
4391 Frequency : Only if a default stripe is set on the filesystem root.
4393 Description: Setting a default stripe on the filesystem root prevented the
4394 filesystem from being remounted.
4395 Details : The client was sending extra request flags in the root getattr
4396 request and did not allocate a reply buffer for the dir EA.
4399 Frequency : occasional, higher if lots of files are accessed by one client
4400 Bugzilla : 6159, 6097
4401 Description: Client trips assertion regarding lsm mismatch/magic
4402 Details : While revalidating inodes the VFS looks up inodes with ifind()
4403 and in rare cases can find an inode that is being freed.
4404 The ll_test_inode() code will free the lsm during ifind()
4405 when it finds an existing inode and then the VFS later attaches
4406 this free lsm to a new inode.
4410 Bugzilla : 6422, 7030
4411 Description: MDS deadlock between mkdir and client eviction
4412 Details : Creating a new file via mkdir or mknod (starting a transaction
4413 and getting the ns lock) can deadlock with client eviction
4414 (gets ns lock and trying to finish a synchronous transaction).
4417 Frequency : occasional
4418 Description: While starting a server, the fsfilt_ext3 module could not be
4420 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4421 kernels. Previously, lconf would still use the ext3 name
4422 when trying to load modules. Now, it will correctly use
4423 ext3 on 2.4 and ldiskfs on 2.6.
4425 Severity : enhancement
4426 Description: The default stripe count has been changed to 1
4427 Details : The interpretation of the default stripe count (0, to lfs
4428 or lmc) has been changed to mean striping across a single
4429 OST, rather than all available. For general usage we have
4430 found a stripe count of 1 or 2 works best.
4432 Severity : enhancement
4433 Description: Add support for compiling against Cray portals.
4434 Details : Conditional compiling for some areas that are different
4438 Frequency : occasional
4439 Bugzilla : 6409, 6834
4440 Description: Creating files with an explicit stripe count may lead to
4441 a failed assertion on the MDS
4442 Details : If some OSTs are full or unavailable, creating files may
4443 trigger a failed assertion on the MDS. Now, Lustre will
4444 try to use other servers or return an error to the
4448 Frequency : occasional
4450 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4451 caused lock timeout during readahead (since 1.4.2).
4452 Details : Processes doing readahead might match a lock that hasn't been
4453 granted yet if there are overlapping and conflicting lock
4454 requests. The readahead process waits on ungranted lock
4455 (original lock is CBPENDING), while OST waits for that process
4456 to cancel CBPENDING read lock and eventually evicts client.
4458 Severity : enhancement
4460 Description: Initial enabling of flock support for clients
4461 Details : Implements fcntl advisory locking and file status functions.
4462 This feature is provided as an optional mount flag (default
4463 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4464 locking are implemented yet, and those that are are not guaranteed
4465 to be completely correct in production environments.
4466 mount -t lustre -o [flock|noflock] ...
4469 Frequency : occasional
4471 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4472 assertion in the ext3 JBD (journaling) layer.
4473 Details : The b_committed_data struct is protected by the big kernel lock
4474 in 2.4 kernels, serializing journal_commit_transaction() and
4475 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4476 access to this struct. In 2.6 kernels there is finer grained
4477 locking to improve SMP performance of the JBD layer.
4481 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4486 Description: merge and cleanup kernel patches.
4489 -----------------------------------------------------------------------------
4491 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4493 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4494 versions of Lustre. Please update all servers and clients to
4495 version 1.4.2 or later at the same time. You must also run
4496 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4497 to update the configuration logs.
4499 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4500 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4501 - swab llog records of type '0' so we get proper header size/idx (5861)
4502 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4503 - fix rename of one directory over another leaking an inode (5953)
4504 - avoid SetPageDirty on 2.6 (5981)
4505 - don't re-add just-being-destroyed locks to the waiting list (5653)
4506 - when creating new directories, inherit the parent's custom
4507 striping settings if present parent (3048)
4508 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4509 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4510 - don't hold client locks on temporary worklist from l_lru (5666)
4511 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4512 - automatically evict dead clients (5921)
4513 - Update file size properly in create+truncate+fstat case (6196)
4514 - Do not unhash mountpoint dentries, do not allow removal of
4516 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4517 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4518 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4519 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4520 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4521 - Correctly handle reads of files with no objects (6243)
4522 - lctl recover will also mark a device active if deactivate used (5933)
4524 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4525 - allow --write-conf on an MDS with different nettype than client (5619)
4526 - don't write config llogs to MDS for mounts not from that MDS (5617)
4527 - lconf should create multiple TCP connections from a client (5201)
4528 - init scripts are now turned off by default; run chkconfig --on
4529 lustre and chkconfig --on lustrefs to use them
4530 - upcalls are no longer needed for clients to recover to failover
4532 - add --abort-recovery option to lconf to abort recovery on device
4534 - add support for an arbitrary number of OSTs (3026)
4535 - Quota support protocol changes.
4536 - forward compatibility changes to wire structs (6007)
4537 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4538 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4539 - improved "lustre status" script
4540 - initialize blocksize for non-regular files (6062)
4541 - added --disable-server and --disable-client configure options (5782)
4542 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4543 - Vanilla 2.4.29 support
4544 - increase maximum number of obd devices to 520 (6242)
4545 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4546 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4547 SLES 9 2.6.5 kernels
4548 - limit stripes per file to 160 (the maximum EA size) (6093)
4550 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4553 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4554 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4555 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4556 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4557 - small ext3 extents cleanups and fixes (5733)
4558 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4559 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4560 - avoid needless client->OST connect, fix handle mismatch (5317)
4561 - fix DLM error path that led to out-of-sync client, long delays (5779)
4562 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4563 - fix several locking issues related to i_size (5492,5624,5654,5672)
4564 - don't move pending lock onto export if it is already evicted (5683)
4565 - fix kernel oops when creating .foo in unlinked directory (5548)
4566 - fix deadlock in obdfilter statistics vs. object create (5811)
4567 - use time_{before,after} to avoid timer jiffies wrap (5882)
4568 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4569 - Do not leak request if server was not able to process it (5154)
4570 - If mds_open unable to find parent dir, make that negative lookup(5154)
4571 - don't create new directories with extent-mapping (5909, 5936)
4573 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4574 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4575 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4576 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4577 - support for RHEL4 kernel on i686 (5773)
4578 - provide error messages when incompatible logs are encountered (5898)
4580 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4581 * version 1.4.0.10 (1.4.1 release candidate 1)
4583 - don't keep a lock reference when lock is not granted (4238)
4584 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4585 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4586 - drop import inflight refcount on signal_completed_replay error (5255)
4587 - unlock page after async write error during send (3677)
4588 - handle missing objects in filter_preprw_read properly (5265)
4589 - no transno return for symlink open, don't save no-trasno open (3440)
4590 - don't try to complete elan receive that already failed (4012)
4591 - free RPC server reply state on error (5406)
4592 - clean up thread from ptlrpc_start_thread() on error (5160)
4593 - readahead could read extra page into cache that wasn't ejected (5388)
4594 - prevent races in class_attach/setup/cleanup/detach (5260)
4595 - don't dereference de->d_inode after l_dput of de (5458)
4596 - use "int" for stripe value returned from lock_to_stripe (5544)
4597 - mballoc allocation and error-checking fixes in 2.6 (5504)
4598 - block device patches to fix I/O request sizes in 2.6 (5482)
4599 - look up hostnames for IB nals (5602)
4600 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4601 - don't start multiple acceptors for the same port (5277)
4602 - fix incorrect LASSERT in mds_getattr_name (5635)
4603 - export a proc file for general "ping" checking (5628)
4604 - fix "lfs check" to not block when the MDS is down (5628)
4606 - service request history (4965)
4607 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4608 - create an "evict_client" /proc entry on OSTs, like the MDS has
4609 - fix mount usage message, return errors per mount(8) (5168)
4610 - change grep [] to grep "[]" in tests so they work in more UMLs
4611 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4612 - remove ancient LOV_MAGIC_V0 EA support (5047)
4613 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4614 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4615 - allow readahead allocations to fail when low on memory (5383)
4616 - mmap locking landed again, after considerable improvement (2828)
4617 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4619 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4622 - send OST transaction number in read/write reply to free req (4966)
4623 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4624 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4625 - the watchdog thread now runs as interruptible (5246)
4626 - flock/lockf fixes (but it's still disabled, pending 5135)
4627 - don't use EXT3 constants in llite code (5094)
4628 - memory shortage at startup could cause assertion (5176)
4630 - reorganization of lov code
4631 - single portals codebase
4633 - add extents/mballoc support (5025)
4634 - direct I/O reads in the obdfilter (4048)
4635 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4637 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4640 - send OST transaction number in read/write reply to free req (4966)
4641 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4642 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4643 - the watchdog thread now runs as interruptible (5246)
4644 - handle missing objects in filter_preprw_read properly (5265)
4645 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4646 - drop import inflight refcount on signal_completed_replay error (5255)
4647 - unlock page after async write error during send (3677)
4648 - return original error code on reconstructed replies (3761)
4649 - no transno return for symlink open, don't save no-trasno open (3440)
4651 - add pid to ldlm debugging output (4922)
4652 - bump the watchdog timeouts -- we can't handle 30sec yet
4653 - extra debugging for orphan dentry/inode bug (5259)
4655 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4658 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4659 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4660 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4661 - fix NULL dereference in /proc/sys/portals/routes (4827)
4662 - allow failed mdc_close() operations to be interrupted (4561)
4663 - stop precreate on OST before MDS would time out on it (4778)
4664 - don't send partial-page writes before EOF from client (4410)
4665 - discard client grant for sub-page writes on large-page clients (4520)
4666 - don't free dentries not owned by NFS code, check generation (4806)
4667 - fix lsm leak if mds_create_objects() fails (4801)
4668 - limit debug_daemon file size, always print CERROR messages (4789)
4669 - use transno after validating reply (3892)
4670 - process timed out requests if import state changes (3754)
4671 - update mtime on OST during writes, return in glimpse (4829)
4672 - add mkfsoptions to LDAP (4679)
4673 - use ->max_readahead method instead of zapping global ra (5039)
4674 - don't interrupt __l_wait_event() during strace
4676 - add software watchdogs to catch hung threads quickly (4941)
4677 - make lustrefs init script start after nfs is mounted
4678 - fix CWARN/ERROR duplication (4930)
4679 - return async write errors to application if possible (2248)
4680 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4681 - print NAL number in %x format (4645)
4682 - update barely-supported suse-2.4.21-171 series (4842)
4683 - support for sles 9 %post scripts
4684 - support for building 2.6 kernel-source packages
4685 - support for sles km_* packages
4687 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4690 - ignore -ENOENT errors in osc_destroy (3639)
4691 - notify osc create thread that OSC is being cleaned up (4600)
4692 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4693 - reconstruct ost_handle() like mds_handle() (4657)
4694 - create a new thread to do import eviction to avoid deadlock (3969)
4695 - let lconf resolve symlinked-to devices (4629)
4696 - don't unlink "objects" from directory with default EA (4554)
4697 - hold socknal file ref over connect in case target is down (4394)
4698 - allow more than 32000 subdirectories in a single directory (3244)
4699 - fix blocks count for O_DIRECT writes (3751)
4700 - OST returns ENOSPC from object create when no space left (4539)
4701 - don't send truncate RPC if file size isn't changing (4410)
4702 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4703 - bind to privileged port in socknal and tcpnal (3689)
4705 - rate limit CERROR/CWARN console message to avoid overload (4519)
4706 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4707 - basic mmap support (3918)
4708 - kernel patch series update from b1_4 (4711)
4710 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4713 - avoid crash during MDS cleanup with OST shut down (2775)
4714 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4715 - don't use bad inodes on the MDS (3744)
4716 - dynamic object preallocation to improve recovery speed (4236)
4717 - don't hold spinlock over lock dumping or change debug flags (4401)
4718 - don't zero obd_dev when it is force cleaned (3651)
4719 - print grants to console if they go negative (4431)
4720 - "lctl deactivate" will stop automatic recovery attempts (3406)
4721 - look for existing locks in ldlm_handle_enqueue() (3764)
4722 - don't resolve lock handle twice in recovery avoiding race (4401)
4723 - revalidate should check working dir is a directory (4134)
4725 - don't always mark "slow" obdfilter messages as errors (4418)
4727 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4730 - don't close LustreDB during write_conf until it is done (3860)
4731 - fix typo in lconf for_each_profile (3821)
4732 - allow dumping logs from multiple threads at one time (3820)
4733 - don't allow multiple threads in OSC recovery (3812)
4734 - fix debug_size parameters (3864)
4735 - fix mds_postrecov to initialize import for llog ctxt (3121)
4736 - replace config semaphore with spinlock (3306)
4737 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4738 - don't allow enqueue to complete on a destroyed export (3822)
4739 - down write_lock before checking llog header bitmap (3825)
4740 - recover from lock replay timeout (3764)
4741 - up llog sem before sending rpc (3652)
4742 - reduce ns lock hold times when setting kms (3267)
4743 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4744 - fix NULL deref and obd_dev leak on setup error (3312)
4745 - replace some LBUG about llog ops with error handling (3841)
4746 - don't match INVALID dentries from d_lookup and spin (3784)
4747 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4748 - fix invalid assertion in ptlrpc_set_wait (3880)
4750 - add libwrap support for the TCP acceptor (3996)
4751 - add /proc/sys/portals/routes for non-root route listing (3994)
4752 - allow setting MDS UUID in .xml (2580)
4753 - print the stack of a process that LBUGs (4228)
4755 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4758 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4759 - make sure to unset replay flag from failed open requests (3440)
4760 - if default stripe count is 0, use OST count for inode size (3636)
4761 - update parent mtime/ctime on client for create/unlink (2611)
4762 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4763 - free recovery state on server during a forced cleanup (3571)
4764 - unregister_reply for resent reqs (3063)
4765 - loop back devices mounting and status check on 2.6 (3563)
4766 - fix resource-creation race that can provoke i_size == 0 (3513)
4767 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4768 - more debugging for page-accounting assertion (3746)
4769 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4770 - avoid deadlock after precreation failure (3758)
4771 - fix race and lock order deadlock in orphan handling (3450, 3750)
4772 - add validity checks when grabbing inodes from l_ast_data (3599)
4774 - add /proc/.../recovery_status to obdfilter (3428)
4775 - lightweight CDEBUG infrastructure, debug daemon (3668)
4776 - change default OSC RPC parameters to be better on small clusters
4777 - turn off OST read cache for files smaller than 32MB
4778 - install man pages and include them in rpms (3100)
4779 - add new init script for (un)mounting lustre filesystems (2593)
4780 - run chkconfig in %post for init scripts (3701)
4781 - drop scimac NAL (unmaintained)
4783 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4786 - clean kiobufs before and after use (3485)
4787 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4788 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4789 - change init script to fail more clearly if not run as root (1528)
4790 - allow clients to reconnect during replay (1742)
4791 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4792 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4793 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4794 - fix cond_resched (3554)
4795 - only evict unfinished clients after recovery (3515)
4796 - allow bulk resend, prevent data loss (3570)
4797 - dynamic ptlrpc request buffer allocation (2102)
4798 - don't allow unlinking open directory if it isn't empty (2904)
4799 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4800 - remove extraneous obd dereference causing LASSERT failure (3334)
4801 - don't use get_cycles() when creating temp. files on the mds (3156)
4802 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4803 - handle EEXIST for set-stripe, set proper directory name (3336)
4805 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4806 - fix ksocknal_fmb_callback() error messages (2918)
4808 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4811 - don't copy lvb into (possibly NULL) reply on error (2983)
4812 - don't deref dentry after dput, don't free lvb on error (2922)
4813 - use the kms to determine writeback rpc length (2947)
4814 - increment oti_logcookies when osc is inactive (2948)
4815 - update client's i_blocks count via lvb messages (2543)
4816 - handle intent open/close of special files properly (1557)
4817 - mount MDS with errors=remount-ro, like obdfilter (2009)
4818 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4819 - don't use cancelling-locks' kms values (2947)
4820 - use highest lock extent for kms, not last one (2925)
4821 - don't dereference ERR_PTR() dentry in error handling path (3107)
4822 - fix thread race in portals_debug_dumplog() (3122)
4823 - create lprocfs device entries at setup instead of at attach (1519)
4824 - common AST error handler, don't evict client on completion race (3145)
4825 - zero nameidata in detach_mnt in 2.6 (3118)
4826 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4827 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4828 - initialize RPC timeout timer earlier for 2.6 (3219)
4829 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4830 - print nal/nid for unknown nid (3258)
4831 - additional checks for oscc recovery before doing precreate (3284)
4832 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4833 - don't crash in mdc_close for bad permissions on open (3285)
4834 - zero i_rdev for non-device files (3147)
4835 - clear page->private before handing to FS, better assertion (3119)
4836 - tune the read pipeline (3236)
4837 - fix incorrect decref of invalidated dentry (2350)
4838 - provide read-ahead stats and refine rpc in flight stats (3328)
4839 - don't hold journal transaction open across create RPC (3313)
4840 - update atime on MDS at close time (3265)
4841 - close LDAP connection when recovering to avoid server load (3315)
4842 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4843 - don't leak open file on MDS after open resend (3325)
4844 - serialize filter_precreate and filter_destroy_precreated (3329)
4845 - loop device shouldn't call sync_dev() for nul device (3092)
4846 - clear page cache after eviction (2766)
4847 - resynchronize MDS->OST in background (2824)
4848 - refuse to mount the same filesystem twice on same mountpoint (3394)
4849 - allow llmount to create routes for mounting behind routers (3320)
4850 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4851 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4852 - fix rare problem with rename on htree directories (3417)
4854 - allow default OST striping configuration per directory (1414)
4855 - fix compilation for qswnal for 2.6 kernels (3125)
4856 - increase maximum number of MDS request buffers for large systems
4857 - change liblustreapi to be useful for external progs like lfsck (3098)
4858 - increase local configuration timeout for slow disks (3353)
4859 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4861 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4864 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4865 - don't overwrite extent policy data in reply if lock was blocked (2901)
4866 - drop filter export grants atomically with removal from device (2663)
4867 - del obd_self_export from work_list in class_disconnect_exports (2908)
4868 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4869 - swab reply message in mdc_close, other PPC fixes (2464)
4870 - fix destroying of named logs (2325)
4871 - overwrite old logs when running lconf --write_conf (2264)
4872 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4873 - fix race in target_handle_connect (2898)
4874 - mds_reint_create() should take same inode create lock (2926)
4875 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4876 - don't close files for self_export to avoid uninitialized obd (2936)
4877 - allow MDS with the same name as client node (2939)
4878 - hold dentry reference for closed log files for unlink (2325)
4879 - reserve space for all logs during transactions (2059)
4880 - don't evict page beyond end of stripe extent (2925)
4881 - don't oops on a deleted current working directory (2399)
4882 - handle hard links to targets without a parent properly (2517)
4883 - don't dereference NULL lock when racing during eviction (2867)
4884 - don't grow lock extents when lots of conflicting locks (2919)
4886 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4889 - account for cache space usage on clients to avoid data loss (974)
4890 - lfsck support in lustre kernel code (2349)
4891 - reduce journal credits needed for BRW writes (2370)
4892 - orphan handling to avoid losing space on client/server crashes
4893 - ptlrpcd can be blocked, stopping ALL progress (2477)
4894 - use lock value blocks to assist in proper KMS, faster stat (1021)
4895 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4896 - recovery for initial connections (2355)
4897 - fixes for mds_cleanup_orphans (1934)
4898 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4899 - block all file creations until orphan recovery completes (1901)
4900 - client remove rq_connection from request struct (2423)
4901 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4902 - recovery timer race (2670)
4903 - mdc_close recovey bug (2532)
4904 - ptlrpc cleanup bug (2710)
4905 - mds timeout on local locks (2588)
4906 - namespace lock held during RPCs (2431)
4907 - handle interrupted sync write properly (2503)
4908 - don't try to handle a message that hasn't been replied to (2699)
4909 - client assert failure during cleanup after abort recovery (2701)
4910 - leak mdc device after failed mount (2712)
4911 - ptlrpc_check_set allows timedout requests to complete (2714)
4912 - wait for inflight reqs when ptlrpcd finishes (2710)
4913 - make sure unregistered services are removed from the srv_list
4914 - reset bulk XID's when resending them (caught by 1138 test)
4915 - unregister_bulk after timeout
4916 - fix lconf error (2694)
4917 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4918 - readahead locks pages, leaves pending causing memory pressure (2673)
4919 - increase OST request buffers to 4096 on large machines (2729)
4920 - fix up permission of existing directories in simple_mkdir (2661)
4921 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4922 - don't assign transno to errored transactions (2742)
4923 - don't delete objects on OST if given a bogus objid from MDS (2751)
4924 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4925 - if rq_no_resend, then timeout request after recovery (2432)
4926 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4927 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4928 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4929 - MDS recovery shouldn't delete objects that it has given out (2730)
4930 - if enqueue arrives after completion, don't clobber LVB (2819)
4931 - don't unlock pages twice when trigger_group_io returns error (2814)
4932 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4933 - don't write pages to disk if there was an error (1450)
4934 - don't ping imports that have recovery disabled (2676)
4935 - take buffered bytes into account when balancing socknal conn (2817)
4936 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4937 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4938 - remove little-endian swabbing of llog records (1987)
4939 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4940 - retry reposting request buffers if they fail (1191)
4941 - grow extent at grant time to avoid granting a revoked lock (2809)
4942 - lock revoke doesn't evict page if covered by a second lock (2765)
4943 - disable VM readahead to avoid reading outside lock extents (2805)
4945 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4946 - updated kernel patches for hp-2.4.20 kernel (2681)
4948 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4951 - fix truncated write corruption (2366)
4952 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4954 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4955 - ensure len(uuid) < 37 in lmc (1171)
4956 - fix ia64 OOPS in llog_test (2255)
4957 - zero end of page at obdfilter for partial page writes (2648)
4958 - don't leave stale dentries around after renames (bug 2428)
4959 - fix timeouts when evicting a client with a single lock held (2642)
4960 - set deadline for the initial HELLO message to drain (2634)
4961 - print out dotted-quad IP addresses in the socknal (2302)
4963 - additional debugging for MDS client eviction problem (2443)
4964 - fix mkfsoptions support for osts (2603, 2604)
4966 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4969 - add series for the vanilla 2.6.0 kernel
4970 - add series for the vanilla 2.4.24 kernel
4971 - add series for a cray x86/64 UL kernel drop
4972 - fix xattr patches for the vanilla 2.4.19 series
4974 - generate true UUIDs in lmc (1171)
4975 - have portals stack dumping break in UML (2466)
4976 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4977 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4978 - wake callback waiting threads on client eviction (2460)
4979 - Add --ptldebug and --subsystem to lmc (1719)
4980 - update assertion to allow safe interrupt allocation
4981 - set rq_no_resend for cancel requests (2432)
4982 - recalculate ptlrpcd timeout after resend (2494)
4983 - call vfs_rmdir when removing pending directories (2368)
4984 - fix renaming a file to itself (2429)
4985 - lmc creates a default one-stripe lov (2454)
4986 - expand procfs space to handle large clusters (2326)
4987 - increase UML stack to avoid overflow
4988 - update lconf's list of debug and subsystem masks
4989 - fix lfs find --obd (2510)
4990 - /proc tunable for disabling filter read caching (2591)
4991 - stop rpm packages from altering slapd.conf (2301)
4992 - disable nagle in the socknal under 0conf (2578)
4993 - choose mds inode size based on stripe count (2572)
4994 - fix kernel-source rpm problems (2516)
4996 - add --disable-doc to avoid pdf generation (2421)
4997 - update documentation, tests, type-os, comments
4998 - avoid format warnings on ia64
4999 - remove the TOE NAL
5000 - tiny code cleanups by removing unused fields
5002 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5005 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5006 - remove the most blatant lies from BUILDING file (2371)
5007 - change default debug level to reasonable production setting
5008 - reduce client side cache size to reduce cache flush time
5009 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5010 - make TCP zerocopy and pinger support enabled by default (2476)
5011 - sync writes completed after process exits caused crashes (2319)
5012 - maintain correct mount count on the MDS (2356)
5013 - backout 1557, because 2316 wasn't really fixed
5014 - better file I/O statistics gathering in /proc
5015 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5016 - another kernel patch to fix zero-copy TCP function export
5017 - don't take duplicate lock when processing re-sent getattr (2420)
5018 - lctl uses obd_self_export instead of creating new conn (2353)
5019 - MDS/OST recovery case which requires object creation asserted (2425)
5020 - move lfs from /usr/sbin to /usr/bin in packages
5021 - fix race between mds_client_add and mds_client_free (2417)
5022 - use kmalloc instead of slabs in portals (2430)
5023 - don't create duplicate records when a failover MDS is present (2442)
5024 - remove unnecessary mount age check (2332)
5025 - don't remove directory inodes from locks prematurely (2451)
5026 - don't break if MDS service name is the same as hostname (2103)
5027 - fix races in client write RPC generation when cache full (2482)
5029 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5032 - remove now-unused request->rq_obd (278)
5033 - if an allocation fails, print out how much memory we've used (1933)
5034 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5035 - add the "configurable stack size" patch to most series files (1256)
5036 - ability to write large log records, for 100+ OST configs (2306)
5037 - fix NULL deref when filter_prep fails (2314)
5038 - fix operator precedence error in filter_sync
5039 - dynamic allocation of socknal TX descriptors (2315)
5040 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5041 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5042 - ensure that CWARN messages in Portals always get to the syslog
5043 - __init/__exit are not for prototype decls (ldlm_init/exit)
5044 - x86-64 compile warning fixes
5045 - fix gateway LMC keyword conflict (2318)
5046 - fix MDS lock inversions in getattr/reint paths (1844)
5047 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5048 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5049 - DLM race condition prevented some lock evictions (2328)
5050 - ENOMEM detection and retry on socknal sends (2230)
5051 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5052 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5053 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5054 - better error messages when a client is rejected during recovery (1505)
5055 - avoid cancelling locks which were never granted, after failure (2330)
5056 - fix i_sem/journal inversion in mds_client_add (2333)
5057 - fix truncate/getattr lock cycle deadlock (2334)
5058 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5059 - fix two rare exit paths which could leak an l_lock() ref (2321)
5060 - fencepost error in MDS/OST orphan recovery (2226)
5061 - make log record alignment 8 bytes (1988)
5062 - lstripe now fails when requested offset > ost_count (2237)
5063 - ensure that all kernel series have a complete list.h (1607)
5064 - fix crashes in special-file operations (2316)
5065 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5067 - allow configurable automake binary, for testing new versions
5068 - small update to the lfs documentation
5070 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5072 * fix negative export reference count in fsfilt_sync (2312)
5074 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5075 * release candidate 0.9.1
5077 - orphans are moved into the PENDING directory for possible recovery
5078 - replayed opens now open by fid for orphan/rename safety (1042)
5079 - last close of an orphan inode generates a transno (683)
5080 - chdir() and mount() now pin the directory entry (1020)
5081 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5082 - discard very old requests without processing them (1502)
5083 - remove some common, well-understood CERRORs (1505)
5084 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5085 - clear "grant" flags in OST replies until OST grant code lands (1644)
5086 - fix read performance by not clobbering i_blksize on client (1598)
5087 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5088 - make LRU size a /proc tunable, clears locks when reduced (707)
5089 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5090 - server locks take references on exports now (1558)
5091 - build fixes for 2.4.20-rh trees (1663)
5092 - return an error from lov_create if all OSCs are inactive (1751)
5093 - fix import levels when a reconnect happens without a timeout (1597)
5094 - exit early from mds_open if we get a lookup error (1749)
5095 - partial page read at EOF wouldn't wait for disk before sending (1642)
5096 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5097 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5098 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5099 - write/truncate lock inversion (1639)
5100 - Don't auto-load obdclass, portals modules during cleanup (1495)
5101 - fix timestamps from jumping to "now" (1763)
5102 - extra journal assertions (1648)
5103 - add an extra multiunlink test (1771)
5104 - fix read_record/write_record API (1776)
5105 - fix leak of offset_extent, possible incorrect i_size later (1772)
5106 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5107 - Debugging for the kqswnal_get_idle_tx problems (1820)
5108 - Allow recovery to be attempted multiple times (1536)
5109 - Write out MDS last_rcvd file after it is first created (1600)
5110 - Fix tx_descriptor leak in failed transmit situations (1827)
5111 - ext3 journaling fixes for assertion failure after IO error (1871)
5112 - class_export_put() on freed export after completion AST error (1896)
5113 - Fix revalidate looping in VFS (1322)
5114 - Don't access a freed export during MDS_REINT timeout (1521)
5115 - Add open-unlink recovery support on the MDS (1673,1764)
5116 - Return an error if no MDS data was read from last_rcvd (1946)
5117 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5118 - Don't setup a disk device that doesn't match exported UUID (317)
5119 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5120 - avoid committing NULL handle in force close
5121 - local.sh is now a one-stripe LOV configuration
5122 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5123 - fix ext3 htree duplicate directory entry corruption (1516)
5124 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5125 - update to new LOV EA format (2097)
5126 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5128 2003-06-15 Phil Schwan <phil@clusterfs.com>
5131 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5132 - per-import/export recovery handling (958, 931, 959)
5133 - multiple last-rcvd slots, for serving multiple FSes (949)
5134 - connections are again shared between multiple imp/exports (963, 964)
5135 - "umount -f" would hang if any requests needed to be sent (393, 978)
5136 - avoid pinning large req buffer by copying for queued messages (989)
5137 - add "uuid" to "lctl device" command to help upcalls (991)
5138 - "open" RPCs with transnos would confuse recovery counters (1037)
5139 - do proper endian conversion of all wire messages (288, 340, 891)
5140 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5141 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5142 - fix ptlrpc_connection leak in target_handle_connect (1174)
5143 - fix import refcounting bug in OST and MDS cleanup (1134)
5144 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5145 - fix very unlikely obd_types race condition (501)
5146 - remove osc_open hack for echo_client (1187)
5147 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5148 - a failure in read_inode2 leads to deadlock (1139)
5149 - cancel ack-locks as soon as transaction is committed (1072)
5150 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5151 - make sure to commitrw after any preprw to avoid deadlock (1162)
5152 - failing to execute a file in a lustre FS would lock inode (1203)
5153 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5154 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5155 - obd_uuid2tgt would walk off the end of the list (1255)
5156 - on IA64 the osc would give portals incorrect bulk size (1258)
5157 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5158 - fix lock inversion caused by new llite matching code (1282)
5159 - limit the number of dirty pages on a client to 10MB (1286)
5160 - timed out locks were not being corrected cancelled (1289)
5161 - fix O_DIRECT above 4GB on IA-32 (1292)
5162 * major user-visible changes
5163 - fail out/fail over policy now controlled by the upcall (993)
5165 - add OBD_PING to check server availability and failure (954)
5166 - lustre messages are now sent in sending host order (288, 340, 891)
5167 - add eadatalen to MDS getattr reply (340)
5168 - OST read replies may contain second buffer, with per-page status (593)
5170 2003-03-11 Phil Schwan <phil@clusterfs.com>
5173 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5174 - failed open()s could cause deadlock; fixed (867, 869)
5175 - stop cancelling OST locks when files are closed (481)
5176 - overlapping XID spaces caused network corruption (851, 853)
5177 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5178 - setattr_raw added, to do single-RPC, server-side setattrs
5179 - lmc/lconf syntax change for OST UUIDs
5180 - fix crashy race condition between ptlrpc_free_req and osc_close
5181 - don't use request in mdc_enqueue if we hit a timeout (889)
5182 - don't set the inode i_size for regular files from the MDS (896)
5183 - handle out of order completion AST (842)
5184 - don't LBUG if a lock request times out after receiving AST (913)
5185 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5186 - if a bad lock AST arrives, send an error instead of dropping entirely
5187 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5188 - fix leak in bulk IO when only partially completed (899, 900, 926)
5189 - fix O_DIRECT for ia64 (55)
5190 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5191 - C-z after timeout could hang a process forever; fixed (977)
5193 - client-side I/O cache (678, 924, 929, 941, 970)
5195 - READPAGE and SETATTRs which don't take server-side locks get
5198 2003-02-11 Phil Schwan <phil@clusterfs.com>
5201 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5202 - Fix fsfilt_extN_readpage() to read a full page of directory
5203 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5204 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5205 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5206 - Most intent operations were removed, in favour of a new RPC mode
5207 that does a single RPC to the server and bypasses most of the VFS
5208 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5209 - Aggressively cancel local locks on DLM servers
5210 - mds_reint_unlink sends EA to the client if it's the last nlink.
5211 client uses that EA to unlink OST objects.
5212 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5213 - recursive symlinks were fixed (439)
5214 - fixed NULL deref in DEBUG_REQ
5215 - filter_update_lastobjid no longer calls sync, which annoyed extN
5216 - fixed multi-client small-writes to a single file problem (445)
5217 - fixed mtime updates during file writes (607)
5218 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5219 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5220 - fixed Linux OST size reporting problem (444, 656)
5221 - OST now updates object mtime with writes or setattr (607, 619)
5222 - client verifies file size before zeroing page past EOF (445)
5223 - OST now writes last allocated objid to disk with allocation (108)
5224 - LOV on echo now works (409)
5226 - mds_reint_unlink sends a new buffer, with the EA included. this
5227 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5228 set if a regular file was being unlinked, and it was the last link
5229 - use PtlGet from the target for bulk writes (315)
5230 - OST now updates object mtime with writes or setattr (607, 619)
5231 - LDLM now has a grant-time callback to revalidate locked items, if
5233 - Many MDS operations were reorganized to combat race conditions
5235 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5236 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5238 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5241 - Fully reactivate OST imports after reconnection (512, others)
5242 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5243 - More graceful error handling for truncating on dead OST (515)
5244 - Don't error out unless we're actually accessing dead stripes (474)
5245 - Fix garbage sizes when stripes are missing (410)
5246 - LRU counters were broken, causing constant lock purge (433, 432)
5247 - garbage on read from stripes with failed OSTs (441)
5248 - mark OSCs as active before reconnecting during recovery (438)
5249 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5250 - lfind did not preserve OST order in output (443)
5251 - symlinks cause hung clients, incorrect data (439)
5252 - stop dereferencing request after dropping refcount (457)
5253 - don't LASSERT(spin_is_locked) on non-SMP (455)
5254 - fixes for many rename() bugs
5255 - fstat didn't correctly synchronize attributes (399)
5256 - server must handle lock cancellation during blocking AST prep (487)
5257 - bulk descriptors were free()d too soon (511)
5258 - fix paths in lconf, which would load incorrect modules (451, 507)
5259 - fix confusing lconf 'host not found' error message (386)
5260 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5261 - fix race condition in mdc_blocking_ast() for inode access (526)
5262 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5263 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5264 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5265 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5266 - fix lstripe in several ways (kernel side) (527)
5267 - fix request leak in ldlm_cli_enqueue (262)
5268 - incorrect OSC was marked inactive after OST failure
5269 - call mds_fs_cleanup before unmounting filesystem (524)
5270 - fix races between taking ns_lock and ldlm_lock_change_resource
5271 - fix races updating LOV export open file list
5272 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5273 - fix recovery NULL deref in ldlm_cli_cancel_unused
5274 - fix some DLM races by using new hash table for lock handles (419)
5275 - permit the client to specify desired inodes, at replay
5276 - duplicate requests when we queue them for replay reintegration
5277 - fix last_rcvd offset calculation
5278 - sync after each recovered transaction, so we always make progress
5279 - never, not always, ERESTART requests without transnos
5280 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5282 - skip replay if the MDS says that the client is already connected
5283 - don't check for a recovery-enabled export to match lctl's UUID
5284 - don't INC_USE_COUNT for phantom exports
5285 - don't crash when cleaning up phantom exports (567)
5286 - don't double-finish or set replay data for errored mdc_open requests
5287 - abort requests when they time out, so we don't get old replies
5288 - send/receive replies for AST messages again
5289 - if the client says that it doesn't have the lock, cancel it on the
5291 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5292 mark it as destroyed, it will all work out when decref is called
5293 - fix module use counts (22, 581)
5295 - ASTs now expect a reply (server cancels lock on error reply)
5297 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5300 - fix many simultaneous client startup (392)
5301 - fix dentry->d_it clobbering
5302 - credentials weren't being shipped for readdir/getattr operations
5303 - remove invalid assertions triggered during some concurrent MD
5305 - proper Lustre versions added (336, 389)
5306 - fix memory leak for create error case (398)
5307 - fix LOV locking bug that would get cli/srv out of sync
5308 - fix echo client over LOV (409)
5309 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5310 - fix double-O_EXCL intent crash (424)
5311 - avoid sending multiple lock CANCELs (352)
5313 - MDS can do multi-client recovery (modulo bugs in new code)
5315 - many updates, edits, cleanups
5317 2002-11-18 Phil Schwan <phil@clusterfs.com>
5320 - fix null d_it dereference (346)
5321 - fix full OST/dbench hang (333)
5322 - fix permission problem with file removal (286)
5323 - fix removal of OSCs from LOV when they fail
5324 - fix NULL deref during bulk timeout (214)
5325 - fix problems related to multiple filesystems on one MDS (241)
5326 - fixed serious subtle metadata locking bugs
5327 - free locks on clients when inodes are removed due to memory
5329 - fix inode pointer in lock data (285)
5330 - partial support for multiple MDS on a single host (241)
5331 - data locks weren't cancelled at clear_inode time (290, 311)
5332 - intent locks could lead to unbounded lock growth (205)
5333 - added a maximum lock count, an LRU list, and a flusher
5334 - fix multiple rename (365)
5335 - properly abstracted the echo client
5336 - OSC locked 1 byte too many; fixed
5337 - rewrote brw callback code:
5338 - fixed recovery bugs related to LOVs (306)
5339 - fixed too-many-pages-in-one-write crash (191)
5340 - fixed (again) crash in sync_io_timeout (214)
5341 - probably fixed callback-related race (385)
5343 - Add capability to MDS protocol
5344 - LDLM cancellations and callbacks on different portals
5346 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5349 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5350 - defer open object destruction to close time (601981, 138)
5351 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5352 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5353 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5354 - don't call obd_setattr in ll_file_release for destroyed objects
5355 * protocol change to lustre_msg: move |version| and add |flags|
5356 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5357 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5358 * added replay of create, unlink, link and rename operations during
5359 MDS failover; recovery should be much more robust now
5360 * remove failed OSCs from LOVs (only lov_create uses this so far)
5361 * the lustre-HOWTO was brought (more) up to date (582544)
5363 2002-10-23 Phil Schwan <phil@clusterfs.com>
5366 - in-use dentries weren't being reused properly (617851)
5367 - prevent multiple LDLM setup (599178)
5368 - fix LOV size calculations for truncate (617853)
5369 - fix client handling of MDS intent errors (POSIX)
5370 - fix permission bug in lovstripe.c test (624321)
5371 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5372 - truncate past end of file could corrupt data
5373 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5374 - a race in recovery could return ETIMEDOUT to apps (623947)
5375 - building outside the source directory was fixed
5376 * the lustre-HOWTO was brought (more) up to date (582544)
5377 * major progress was made on recovery functionality
5379 2002-10-10 Phil Schwan <phil@clusterfs.com>
5382 - recovery deadlock fix
5383 - rm -rf causes LBUG fix (617817)
5384 - file open by multiple tasks fix (618962)
5385 - directory permissions bugs (602707 and 620007)
5386 - journal_stop fixed with locking (611313)
5387 - O_APPEND failures resolved (618273, perhaps 614459)
5388 - lconf PATH fix (619770)
5389 - IA64 build fix (621450)
5390 - RPC buffer sizes scale with amount of memory
5392 2002-10-01 Phil Schwan <phil@clusterfs.com>
5395 - locks would be cancelled without throwing away data pages,
5396 resulting in inconsistent data (605627)
5397 - inode attributes were not always being refreshed (605627, 612449)
5398 - lconf now continues to cleanup after lctl reports an error
5399 - MDS now enforces user permissions (602707)
5400 - lprocfs cleanup fixed, but not yet enabled (614157)
5401 - fixed infinite server hang, should a client not respond to an AST
5402 - avoid going into recovery if user calls readlink() with a buffer
5403 that's too small (613941)
5404 - AST RPCs no longer require replies (614867) -- this may be changed
5405 - don't crash server if client sends an IOV that's too big (611336)
5406 - fixed lock conversion deadlock (611892)
5407 - fixed the following of symlinks (614622)
5408 * recovery: the server can remove locks from a client that dies, other
5409 clients can make progress
5410 * more extN patch fixes
5411 * compile-time configurable ptlrpc buffer allocations
5413 - collaborative read cache document
5414 - Lustre Lite Performance CDR document-in-progress
5416 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5419 - fix typo in patch-2.4.18
5421 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5424 - clear ptlrpc request each time in handle_incoming_request()
5425 - unlink of files now destroys the object on the OST
5427 2002-09-19 Peter Braam <braam@clusterfs.com>
5429 * add hard link support
5430 * change obdfile creation method
5431 * kernel patch changed
5433 2002-09-19 Peter Braam <braam@clusterfs.com>
5436 - stack overflow bug in extN fixed
5438 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5440 * documentation updates
5441 - add man pages for config tools
5442 - update tests/README to describe testing with new config tools
5443 - finish metadata API descriptions
5444 * bug fixes and cleanups
5445 - statfs workaround for 16TB limit
5446 - LOV stripe allocation improved, can stripe on subset of OSTs
5447 - LOV file size/IO offset was wrong for files > 4GB in size
5448 - object EA data was being dropped, caused files to be unreadable
5449 - memory overflow with non-LOV OST caused memory corruption
5450 - fixed regression tests to work with new config tools, obdfilter
5451 - fixed bug when directory size became larger than 1 block
5452 - fixed bug (for single client case) when PWD was deleted
5453 - invalidate local directory pages when doing intent-based ops
5454 - avoid LDLM oops when lock callback contained bad data
5456 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5458 * documentation updates
5459 * bug fixes and cleanups
5460 - configuration tools
5463 - 64-bit compile warnings
5464 - 64-bit internal statfs data
5466 * test_brw on persistent OST devices
5468 * lprocfs (disabled)
5470 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5472 * documentation updates
5473 * bug fixes and cleanups
5474 * configuration tools
5476 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5479 * many small fixes to 0_5_4
5480 * io/network handling
5481 * thinkos in MDS operations
5483 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5486 * crucial basic fixes to 0.5.3
5487 * IOR, Iozone work over Elan
5490 2002-08-07 Phil Schwan <phil@clusterfs.com>
5491 * version 0_5_3, our first alpha
5492 * we use the new Portals iovs
5493 * documentation updates
5494 * bug fixes and cleanups
5495 * small changes in the DLM wire protocol
5497 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5498 * version 0_5_1 with some initial stability,
5499 * locking on MD and file I/O.
5500 * documentation updates
5501 * several bug fixes since 0.5.0
5502 * small changes in wire protocol
5504 2002-07-18 Phil Schwan <phil@clusterfs.com>
5506 * delivered as Lustre Light Alpha
5507 * fixed a crash after handling invalid MDS requests
5508 * fixed directory pages for architectures with non-4k pages sizes
5510 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5511 * release version v0_4_4
5512 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5514 * Fixes a number of interruption problems with OST operations.
5515 * Update documentation for portals header changes
5516 * Move all wire protocol structs/defines to lustre_idl.h
5517 * Fixes symlink length bug.
5518 * Add tcpdump to repository.
5520 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5521 * release version v0_4_3
5522 * Fixes statfs for inodes on extN.
5523 * Fixes bug in runtests which would delete /etc/hosts.
5524 * Use 64-bit object IDs wherever possible (not into VFS though)
5525 Remove ost_get_info, which is unused by lustre, and out of date.
5527 2002-07-03 Peter Braam <braam@clusterfs.com>
5528 * release version v0_4_2 Fixes a lookup error (type not passed)
5529 * move forward to head of Portals
5530 * move forward to latest Lustre kernel
5532 2002-06-25 Peter Braam <braam@clusterfs.com>
5533 * release version v0_4_1. Hopefully stable on single node use.