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).
15 Severity : enhancement
17 Description: Re-disable certain /proc logging
18 Details : Enable and disable client's offset_stats, extents_stats and
19 extents_stats_per_process stats logging on the fly.
22 Frequency : Only on FC kernels 2.6.22+
24 Description: oops in statahead
25 Details : Do not drop reference count for the dentry from VFS when lookup,
26 VFS will do that by itself.
28 Severity : enhancement
30 Description: Generic /proc file permissions
31 Details : Set /Proc file permissions in a more generic way to enable non-
32 root users operate on some /proc files.
36 Description: Hitting mdc_commit_close() ASSERTION
37 Details : Properly handle request reference release in
38 ll_release_openhandle().
42 Frequency : only patchless client
43 Description: add workaround for race between add/remove dentry from hash
45 Severity : enhancement
47 Description: Allow OST glimpses to return PW locks
51 Description: LBUG when llog conf file is full
52 Details : When llog bitmap is full, ENOSPC should be returned for plain
57 Description: Prevent import from entering FULL state when server in recovery
61 Description: service mount cannot take device name with ":"
62 Details : Only when device name contains ":/" will mount treat it as
68 Description: replace ptlrpcd with the statahead thread to interpret the async
69 statahead RPC callback
73 Frequency : on recovery
74 Description: I/O failures after umount during fail back
75 Details : if client reconnected to restarted server we need join to recovery
76 instead of find server handler is changed and process self eviction
77 with cancel all locks.
79 Severity : enhancement
81 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
83 Severity : enhancement
85 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
87 Severity : enhancement
89 Description: Upcall on Lustre log has been dumped
90 Details : Allow for a user mode script to be called once a Lustre log has
91 been dumped. It passes the filename of the dumped log to the
92 script, the location of the script can be specified via
93 /proc/sys/lnet/debug_log_upcall.
98 Description: avoid idr_remove called for id which is not allocated.
99 Details : Move assigment s_dev for clustered nfs to end of initialization,
100 to avoid problem with error handling.
105 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
106 Details : When connection is reused this not moved from CONN_UNUSED_HASH
107 into CONN_USED_HASH and this prodice warning when put connection
108 again in unused hash.
110 Severity : enhancement
112 Description: File striping can now be set to use an arbitrary pool of OSTs.
114 Severity : enhancement
116 Description: Export bytes_read/bytes_write count on OSC/OST.
120 Description: Early reply size mismatch, MGC loses connection
121 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
122 the connect flags are properly negotiated.
126 Description: Properly propagate oinfo flags from lov to osc for statfs
127 Details : restore missing copy oi_flags to lov requests.
129 Severity : enhancement
131 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
132 : l_getgroups(8), lst(8), routerstat(8)
133 Details : included man pages for llobdstat(8), llstat(8),
134 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
136 Severity : enhancement
138 Description: configure's --enable-quota should check the
139 : kernel .config for CONFIG_QUOTA
140 Details : configure is terminated if --enable-quota is passed but
141 : no quota support is in kernel
145 Description: Remove portals compatibility
146 Details : Remove portals compatibility, not interoperable with releases
151 Description: Resolve device initialization race
152 Details : Prevent proc handler from accessing devices added to the
153 obd_devs array but yet be intialized.
155 Severity : enhancement
157 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
159 Severity : enhancement
161 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
166 Description: Using wrong pointer in osc_brw_prep_request
167 Details : Access to array[-1] can produce panic if kernel compiled with
168 CONFIG_PAGE_ALLOC enabled
172 Description: Client runs out of low memory
173 Details : Consider only lowmem when counting initial number of llap pages
177 Description: *optional* service tags registration
178 Details : if the "service tags" package is installed on a Lustre node
179 When the filesystem is mounted, a local-node service tag will
180 be created. See http://inventory.sun.com/ for more information
181 about the Service Tags asset management system.
185 Description: Kernel BUG tries to release flock
186 Details : Lustre does not destroy flock lock before last reference goes
187 away. So always drop flock locks when client is evicted and
188 perform unlock regardless of successfulness of speaking to MDS.
192 Description: add recount protection for osc callbacks, so avoid panic on shutdown
196 Description: sanity test 65a fails if stripecount of -1 is set
197 Details : handle -1 striping on filesystem in ll_dirstripe_verify
202 Description: ASSERTION(CheckWriteback(page,cmd)) failed
203 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
206 Severity : enhancement
208 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
212 Description: do not process already freed flock
213 Details : flock can possibly be freed by another thread before it reaches
214 to ldlm_flock_completion_ast.
218 Description: LBUG during stress test
219 Details : Need properly lock accesses the flock deadlock detection list.
223 Description: oops in page fault handler
224 Details : kernel page fault handler can return two special 'pages' in error case, don't
225 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
229 Description: timeout with invalidate import.
230 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
231 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
232 ptlrpc_invalidate_import and as result LASSERT.
234 Severity : enhancement
236 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
240 Description: enable MGS and MDT services start separately
241 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
242 a co-located MGS without starting the MGS, which is a complement
243 to 'nosvc' mount option.
247 Frequency : after recovery
248 Description: precreate to many object's after del orphan.
249 Details : del orphan st in oscc last_id == next_id and this triger growing
250 count of precreated objects. Set flag LOW to skip increase count
251 of precreated objects.
255 Frequency : rare, on clear nid stats
256 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
257 Details : when clean nid stats sometimes try destroy live entry,
258 and this produce panic in free.
262 Description: Stack overflow during MDS log replay
263 ease stack pressure by using a thread dealing llog_process.
267 Description: wait until IO finished before start new when do lock cancel.
268 Details : VM protocol want old IO finished before start new, in this case
269 need wait until PG_writeback is cleared until check dirty flag and
270 call writepages in lock cancel callback.
272 Severity : enhancement
274 Description: using special macro for print time and cleanup in includes.
278 Description: mds_mfd_close() ASSERTION(rc == 0)
279 Details : In mds_mfd_close(), we need protect inode's writecount change
280 within its orphan write semaphore to prevent possible races.
284 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
288 Frequency : rare, on shutdown ost
289 Description: don't hit live lock with umount ost.
290 Details : shrink_dcache_parent can be in long loop with destroy dentries,
291 use shrink_dcache_sb instead.
295 Description: don't panic with use echo client
296 Details : echo client pass NULL as client nid pointer and this produce null
301 Description: fix build on ppc32
302 Details : compile code with -m64 flag produce wrong object file for ppc32.
306 Description: add message levels for liblustreapi
310 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
311 Details : Move the 'good_osts' check before the 'total_bavail' check. This
312 will result in an -EAGAIN and in the exit call path we call
313 alloc_rr() which will with increasing aggressiveness attempt to
314 aquire precreated objects on the minimum number of required OSCs.
318 Description: Use old size assignment to avoid deadlock
319 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
320 the scheduling while holding a spinlock. We do not need locking
321 for size in ll_update_inode() because size is only updated from
322 the MDS for directories or files without objects, so there is no
323 other place to do the update, and concurrent access to such inodes
324 are protected by the inode lock.
328 Description: resolve "_IOWR redefined" build error on SLES10
332 Description: dump the memory debugging after all modules are unloaded to
333 suppress false negative in conf_sanity test 39
335 Severity : enhancement
337 Description: build kernel-ib packages for OFED 1.3 in our release cycle
342 Description: fix SLES kernel versioning
343 Details : the kernel version for our SLES 10 kernel did not include a "-"
344 before the "smp" at the end. while this was not a problem in
345 general, it did mean that software trying to use the kernel
346 version to try to detect a vendor specific kernel would fail.
347 this was most evident by the OFED build scripts.
351 Description: Don't update lov_desc members until making sure they are valid
352 Details : When updating lov_desc members via proc fs, need fix their
353 validities before doing the real update.
357 Description: don't put request into delay list while invalidate in flight.
358 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
359 import in flight. this produce timeout for invalidate and sometimes
360 can cause stale data.
364 Frequency : on ppc only
365 Description: not convert ost objects for directory because it's not exist.
366 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
370 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
372 Severity : enhancement
374 Description: Update to support 2.6.22.14 vanilla kernel.
378 Frequency : rare, on recovery
379 Description: read procfs can produce deadlock in some situation
380 Details : Holding lprocfs lock which send rpc can produce block for destroy
381 obd objects and this also block reconnect with -EALREADY. This isn't
382 fix all lprocfs bugs - but make it rare.
384 Severity : enhancement
386 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
389 Frequency : frequent on X2 node
391 Description: mdc_set_open_replay_data LBUG
392 Details : Set replay data for requests that are eligible for replay.
396 Description: lustre_mgs: operation 101 on unconnected MGS
397 Details : When MGC is disconnected from MGS long enough, MGS will evict the
398 MGC, and late on MGC cannot successfully connect to MGS and a lot
399 of the error messages complaining that MGS is not connected.
403 Frequency : on network error
404 Description: panic with double free request if network error
405 Details : mdc_finish_enqueue is finish request if any network error ocuring,
406 but it's true only for synchronus enqueue, for async enqueue
407 (via ptlrpcd) this incorrect and ptlrpcd want finish request
410 Severity : enhancement
412 Description: client-side metadata stat-ahead during readdir(directory readahead)
413 Details : perform client-side metadata stat-ahead when the client detects
414 readdir and sequential stat of dir entries therein
417 Frequency : on start mds
419 Description: Implement get_info(last_id) in obdfilter.
422 Frequency : occasional
424 Description: Correctly check stale fid, not start epoch if ost not support SOM
425 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
426 overwrited with new generated fid, but mds can anwer with one of these
427 two fids and both is not stale. setattr incorectly start epoch and
428 assume will be called done_writeting, but without SOM done_writing
432 Frequency : rare, depends on device drivers and load
434 Description: MDS or OSS nodes crash due to stack overflow
435 Details : Code changes in 1.8.0 increased the stack usage of some functions.
436 In some cases, in conjunction with device drivers that use a lot
437 of stack the MDS (or possibly OSS) service threads could overflow
438 the stack. One change which was identified to consume additional
439 stack has been reworked to avoid the extra stack usage.
442 Frequency : occasional
444 Description: Do not fail import if osc_interpret_create gets -EAGAIN
445 Details : If osc_interpret_create got -EAGAIN it immediately exits and
446 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
447 oscc_has_objects and see OSC has no objests and call
448 oscc_internal_create to resend create request.
450 Severity : enhancement
452 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
454 Severity : enhancement
456 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
459 Frequency : very rare
461 Description: Fix replay if there is an un-replied request and open
462 Details : In some cases, older replay request will revert the
463 mcd->mcd_last_xid on MDS which is used to record the client's
466 Severity : enhancement
468 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
470 Severity : enhancement
472 Description: Add rhel5 support to HEAD.
474 Serverity : enhancement
476 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
481 Description: Don't allow skipping OSTs if index has been specified.
482 Details : Don't allow skipping OSTs if index has been specified, make locking
483 in internal create lots better.
487 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
488 Details : access to bitfield in structure is always rounded to long
489 and this produce problem with not atomic change any bit.
493 Description: Lustre make rpms failed.
494 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
495 builds Lustre rpms from tarball.
498 Frequency : rare on shutdown ost
500 Description: If llog cancel was not send before clean_exports phase, this can
501 produce deadlock in llog code.
502 Details : If llog thread has last reference to obd and call class_import_put
503 this produce deadlock because llog_cleanup_commit_master wait when
504 last llog_commit_thread exited, but this never success because was
505 called from llog_commit_thread.
509 Description: allow userland application know is lost one of stripes.
510 Details : fill lvb_blocks with error code on ost and return it to
511 application if error flag found.
515 Description: NULL lov_tgts causing MDS oops
516 Details : more safe checks for NULL lov_tgts for avoid oops.
518 Severity : enhancement
520 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
524 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
525 Details : If an OST is down the MDS will hang indefinitely in
526 obd_statfs_rqset() waiting for the statfs data. While for
527 MDS QOS usage of statfs, it should not stuck in waiting.
529 Severity : enhancement
531 Description: remote_acl support
532 Details : Support ACL-based permission check for remote user.
533 Support setfacl/getfacl for remote user with the utils
534 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
535 the system "{s,g}etfacl" utils.
537 Severity : enhancement
539 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
541 Severity : enhancement
543 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
547 Description: make mgs_setparam() handle fsname containing dash
548 Details : fsname containing a dash does not work with lctl conf_param
550 Severity : enhancement
552 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
554 Severity : enhancement
556 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
560 Frequency : rare, at shutdown
561 Description: access already free / zero obd_namespace.
562 Details : if client_disconnect_export was called without force flag set,
563 and exist connect request in flight, this can produce access to
564 NULL pointer (or already free pointer) when connect_interpret
565 store ocd flags in obd_namespace.
569 Frequency : only at startup
570 Description: not alloc memory with spinlock held.
571 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
572 if any spinlock held.
574 Severity : enhancement
576 Description: make lustre randomly fail allocating memory
577 Details : Make lustre randomly failed allocating memory for testing purpose.
579 Severity : enhancement
581 Description: lost problems with lov objid file
582 Details : Fixes some scability and access to not inited memory problems
583 in work with lov objdid file.
588 Description: lfs find does not continue on file error
589 Details : Continue other files processing when a file/dir is absent.
593 Description: Inconsistent usage of lustre_pack_reply()
594 Details : Standardize the usage of lustre_pack_reply() such that it
595 always generate a CERROR on failure.
598 Frequency : occasional
600 Description: MDS hang or stay in waiting lock
601 Details : If client receive lock with CBPENDING flag ldlm need send lock
602 cancel as separate rpc, to avoid situation when cancel request
603 can't processed due all i/o threads stay in wait lock.
607 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
611 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
615 Description: Update to SLES9 kernel-2.6.5-7.287.3.
619 Description: Update to RHEL4 latest kernel.
621 Severity : enhancement
623 Description: Build SLES10 patchless client fails
624 Details : The configure was broken by run ./configure with
625 --with-linux-obj=.... argument for patchless client. When the
626 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
627 can't search header adequately. Use absolute path such as
628 -I($LINUX)/include instead.
632 Description: interrupt oig_wait produce painc on resend.
633 Details : brw_redo_request can be used for resend requests from ptlrpcd and
634 private set, and this produce situation when rq_ptlrpcd_data not
635 copyed to new allocated request and triggered LBUG on assert
636 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
637 ptlrpcd set if request is changed and can be safety changed to use
642 Description: LASSERT_{REQ,REP}SWAB macros are buggy
643 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
644 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
645 these macros and replace them with their difinitions instead.
649 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
650 Details : Update which_patch & target file for SLES10 latest kernel.
653 Description: Remove client patches from SLES 10 kernel.
654 Details : This causes SLES 10 clients to behave as patchless clients
655 even on a Lustre-patched (server) kernel.
657 Severity : enhancement
659 Description: self-adjustable client's lru lists
660 Details : use adaptive algorithm for managing client cached locks lru
661 lists according to current server load, other client's work
662 pattern, memory activities, etc. Both, server and client
663 side namespaces provide number of proc tunables for controlling
666 Severity : enhancement
668 Description: light-weight GSS support
669 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
670 and reduce performance overhead.
672 Severity : enhancement
674 Description: Linux keyring support
675 Details : Support using service of Linux keyring for Lustre GSS internal
676 context refresh/cache mechanism.
680 Description: Fix errors in lfs documentation
681 Details : Fixes man pages
685 Description: don't use GFP_* in generic Lustre code.
686 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
690 Description: obdclass is limited by single OBD_ALLOC(idarray)
691 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
695 Description: warnings with build patchless client with vanila 2.6.19 and up
696 Details : change old ctl_table style and replace ctl_table/ctl_table_header
697 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
701 Frequency : Only for SLES
702 Description: sanity_quota fail test_1
703 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
704 file, which makes TSTID[2] inunique.
708 Frequency : Only for SLES9
709 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
710 Details : Update target/ChangeLog/which_patch .
714 Frequency : Only for test_10 in sanity-quota.sh
715 Description: a bug in quota test code
716 Details : checking a test flag in a wrong place
720 Frequency : Only for test_13 in sanity-quota.sh
721 Description: a bug in quota test script
722 Details : shouldn't check overfull allocated quota
726 Description: jbd statistics
727 Details : Port older jbd statistics patch for sles10
729 Severity : enhancement
731 Description: Add printing inode info into message about error in writepage.
736 Description: reply_lock_interpret crash due to race with it and lock cancel.
737 Details : Do not replay locks that are being cancelled. Do not reference
738 locks by their address during replay, just by their handle.
742 Frequency : When flocks are used.
743 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
744 Details : Flock locks might destroy just granted lock if it could be merged
745 with another existing flock, this is done in completion handler,
746 so teach ldlm_cli_enquque_fini that this is a valid case for
752 Description: Oops in read and write path when failing to allocate lock.
753 Details : Check if lock allocation failed and return error back.
757 Description: lstripe command fails for valid OST index
758 Details : The stripe offset is compared to lov->desc.ld_tgt_count
759 instead of lov->desc.ld_active_tgt_count.
763 Description: sanity.sh failed test 103
764 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
769 Description: df doesn't work properly if diskfs blocksize != 4K
770 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
774 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
775 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
777 Severity : enhancement
779 Description: Adaptive timeouts
780 Details : RPC timeouts adapt to changing server load and network
781 conditions to reduce resend attempts and improve recovery time.
785 Description: llapi_file_create() does not allow some changes
786 Details : add llapi_file_open() that allows specifying the mode and
787 open flags, and also returns an open file handle.
791 Description: merge and cleanup kernel patches.
792 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
796 Description: Add journal checksum support.(Kernel part)
797 Details : The journal checksum feature adds two new flags i.e
798 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
799 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
800 indicates that the commit block contains the checksum for
801 the blocks described by the descriptor blocks. Now commit
802 record can be sent to disk without waiting for descriptor
803 blocks to be written to disk. This behavior is controlled
804 using JBD2_FEATURE_ASYNC_COMMIT flag.
808 Description: OSS needs mutliple precreate threads
809 Details : Add ability to start more than one create thread per OSS.
813 Description: Sanity.sh test_65k failed: setstripe should have succeeded
814 Details : execute lfs setstripe on client
818 Description: mds_obd_create error creating tmp object
819 Details : When the user sets quota on root, llog will be affected and can't
820 create files and write files.
823 Frequency : Always on ia64 patchless client, and possibly others.
825 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
826 Details : This allows the patchless client to be loaded on architectures
831 Description: RedHat Update kernel for RHEL5
832 Details : Add the kernel config file for RHEL5.
836 Description: RedHat Update kernel for RHEL5
837 Details : Modify the kernel config file more closer RHEL5.
841 Description: Build failure against Centos5 (RHEL5)
842 Details : Define PAGE_SIZE when it isn't present.
846 Description: test_7 on sanity-quota.sh can't be used on separate machines
847 Details : add facet to handle it
851 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
852 Details : with NFS, the anon dentry's parent was set to itself in
853 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
854 corresponding dentry other than use rec->ur_name.
856 Severity : enhancement
858 Description: lfs setstripe enhancement
859 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
863 Description: enable data checksumming by default
864 Details : enable checksum by default, allow --disable-checksum
865 configure option and "-o nochecksum" mount option. Checksums
866 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
870 Description: lustre support for RHEL5
871 Details : Add support for RHEL5.
875 Description: Client eviction due to failover config
876 Details : after a connection loss, the lustre client should attempt to
877 reconnect to the last active server first before trying the
878 other potential connections.
882 Description: when mds and osts use different quota unit(32bit and 64bit),
883 quota will be released repeatly.
884 Details : void sending multiple quota reqs to mds, which will keep the
885 status between the reqs.
889 Description: osts not allocated evenly to files
890 Details : change the condition to increase offset_idx
893 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
895 Description: Data corruption for OSTs that are formatted larger than 2TB
897 Details : When generating the bio request for lustre file writes the
898 sector number would overflow a temporary variable before being
899 used for the IO. The data reads correctly from Lustre (which
900 will overflow in a similar manner) but other file data or
901 filesystem metadata may be corrupted in some cases.
905 Description: Tune the kernel for good SCSI performance.
906 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
907 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
912 Description: rewrite ext2-derived code in obdclass/uuid.c
913 Details : rewrite inherited code (uuid parsing code from ext2 utils)
914 from scratch preserving functionality.
918 Description: "lfs find -obd UUID" prints directories
919 Details : "lfs find -obd UUID" will return all directory names instead
920 of just file names. It is incorrect because the directories
921 do not reside on the OSTs.
925 Description: use i_size_read and i_size_write in 2.6 port
926 Details : replace inode->i_size access with i_size_read/write()
929 Frequency : when removing large files
931 Description: scheduling issue during removal of large Lustre files
932 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
933 It causes scheduling issues when removing large files (17TB in the
937 Frequency : only with liblustre clients on XT3
939 Description: evictions taking too long
940 Details : allow llrd to evict clients directly on OSTs
943 Frequency : only on ppc
945 Description: /proc/fs/lustre/devices broken on ppc
946 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
947 We should make sure the type of 'index' is loff_t and then cast
948 explicitly as needed below. Do not assign an explicitly cast
952 Frequency : only for rhel5
954 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
955 Details : Modify the target file & which_kernel.
957 Severity : enhancement
959 Description: omit set fsid for export NFS
960 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
961 as FS_REQUIRES_DEV for avoid problems with generate fsid.
965 Frequency : Always, for kernels after 2.6.16
966 Description: Fix warning idr_remove called for id=.. which is not allocated.
967 Details : Last kernels save old s_dev before kill super and not allow
968 to restore from callback - restore it before call kill_anon_super.
972 Description: buffer overruns could theoretically occur
973 Details : llapi_semantic_traverse() modifies the "path" argument by
974 appending values to the end of the origin string, and a
975 overrun may occur. Adding buffer overrun check in liblustreapi.
979 Description: Fix error on 'ls .' at the top of the Lustre mount.
980 Details : Don't revalidate dentry if it is a root dentry.
984 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
985 Details : Modify vm-tunables-rhel4.patch.
989 Description: Kernel config for 2.6.18-vanilla.
990 Details : Modify targets/2.6-vanilla.target.in.
991 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
992 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
993 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
994 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
998 Description: improve handling recoverable errors
999 Details : if request processig with error which can be recoverable on server
1000 request should be resend, otherwise page released from cache and
1005 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1006 Details : Modify target file & which_patch.
1007 A flaw was found in the IA32 system call emulation provided
1008 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1009 value could be stored in the %RAX register, which could trigger an
1010 out-of-bounds system call table access. An untrusted local user
1011 could exploit this flaw to run code in the kernel
1012 (ie a root privilege escalation). (CVE-2007-4573).
1016 Description: O_DIRECT bypasses client statistics.
1017 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1018 read_ahead_stats were updated. copied the stats section from
1019 osc_send_oap_rpc() into async_internal().
1023 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1027 Description: change order of libsysio includes
1028 Details : '#include sysio.h' should always come before '#include xtio.h'
1032 Description: handle "serious error: objid * already exists" more gracefully
1033 Details : If LAST_ID value on disk is smaller than the objects existing in
1034 the O/0/d* directories, it indicates disk corruption and causes an
1035 LBUG(). If the object is 0-length, then we should use the existing
1036 object. This will help to avoid a full fsck in most cases.
1038 Severity : enhancement
1040 Description: adapt the lustre_config script to support the upgrade case
1041 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1042 server targets to 1.6 in parallel.
1046 Description: To avoid grant space > avaible space when the disk is almost
1047 full. Without this patch you might see the error "grant XXXX >
1048 available" or some LBUG about grant, when the disk is almost
1050 Details : In filter_check_grant, for non_grant cache write, we should
1051 check the left space by if (*left > ungranted + bytes), instead
1052 of (*left > ungranted), because only we are sure the left space
1053 is enough for another "bytes", then the ungrant space should be
1054 increase. In client, we should update cl_avail_grant only there
1055 is OBD_MD_FLGRANT in the reply.
1059 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1060 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1061 described in CVE-2007-4573. This problem could allow untrusted
1062 local users to gain root access.
1065 Frequency : when using O_DIRECT and quotas
1067 Description: Incorrect file ownership on O_DIRECT output files
1068 Details : block usage reported by 'lfs quota' does not take into account
1069 files that have been written with O_DIRECT.
1074 Description: touch file failed when fs is not full
1075 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1076 otherwise we can get ENOSP while fs is not full.
1080 Description: parallel lock callbacks
1081 Details : Instead of sending blocking and completion callbacks as separated
1082 requests, adding them to a set and sending in parallel.
1085 Frequency : only for Cray XT3
1086 Bugzilla : 12829/13455
1087 Description: Changing primary group doesn't change the group lustre assigns to
1089 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1090 overridden with the primary group provided by the group upcall,
1091 whereas the supplied fsgid can be trusted if it is in the list of
1092 supplementary groups returned by the group upcall.
1094 Severity : enhancement
1096 Description: Allow masking D_WARNING, D_ERROR messages from console
1097 Details : Console messages can now be disabled via lnet.printk.
1101 Description: User code with malformed file open parameter crashes client node
1102 Details : Before packing join_file req, all the related reference should be
1103 checked carefully in case some malformed flags cause fake
1104 join_file req on client.
1108 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1109 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1110 mode is equal to the granted mode. However, it should also check
1111 that LDLM_FL_CP_REQD is not set.
1115 Description: Heavy nfs access might result in deadlocks
1116 Details : After ELC code landed, it is now improper to enqueue any mds
1117 locks under och_sem, because enqueue might want to decide to
1118 cancel open locks for same inode we are holding och_sem for.
1122 Description: Client eviction while running blogbench
1123 Details : A lot of unlink operations with concurrent I/O can lead to a
1124 deadlock causing evictions. To address the problem, the number of
1125 oustanding OST_DESTROY requests is now throttled to
1126 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1127 callbacks are processed in priority.
1131 Description: enable ACLs on MDS by default
1132 Details : ACLs must be enabled on MDS by default.
1135 Frequency : PPC/PPC64 only
1137 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1138 Details : fix duplicated definitions between asm-ppc64/types.h and
1139 lustre_types.h on PPC.
1142 Frequency : PPC/PPC64 only
1144 Description: asm-ppc/segment.h does not exist
1145 Details : fix compile issue on PPC.
1149 Description: better handle error messages in extents code
1152 Frequency : RHEL4 only
1154 Description: mkfs is very slow on IA64/RHEL4
1155 Details : A performance regression has been discovered in the MPT Fusion
1156 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1157 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1158 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1160 Severity : enhancement
1162 Description: SNMP support enhancement
1163 Details : Adding total number of sampled request for an MDS node in snmp
1166 Severity : enhancement
1168 Description: Optimize ldlm waiting list processing for PR extent locks
1169 Details : When processing waiting list for read extent lock and meeting read
1170 lock that is same or wider to it that is not contended, skip
1171 processing rest of the list and immediatelly return current
1172 status of conflictness, since we are guaranteed there are no
1173 conflicting locks in the rest of the list.
1177 Description: Time out and refuse to reconnect
1178 Details : When the failover node is the primary node, it is possible
1179 to have two identical connections in imp_conn_list. We must
1180 compare not conn's pointers but NIDs, otherwise we can defeat
1181 connection throttling.
1185 Description: port llog fixes from b1_6 into HEAD
1186 Details : Port llog reference couting and some llog cleanups from b1_6
1187 (bug 10800) into HEAD, for protect from panic and access to already
1188 free llog structures.
1192 Description: Detect stride IO mode in read-ahead
1193 Details : When a client does stride read, read-ahead should detect that and
1194 read-ahead pages according to the detected stride pattern.
1198 Description: data checksumming impacts single node performance
1199 Details : add support for several checksum algorithm. Currently, only CRC32
1200 and Adler-32 are supported. The checksum type can be changed on
1201 the fly via /proc/fs/lustre/osc/*/checksum_type.
1205 Description: use adler32 for page checksums
1206 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1211 Description: build for x2 fails
1212 Details : fix compile issue on Cray systems.
1216 Description: Properly match for duplicate locks
1217 Details : Due to different lock order from skiplists code, we need to
1218 traverse entire list for now
1221 Frequency : only on PPC/SLES10
1223 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1224 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1225 be defined. Add a hack in mkfs_lustre.c to work around this compile
1230 Description: LASSERT on MDS when client holding flock lock dies
1231 Details : ldlm pool logic depends on number of granted locks equal to
1232 number of released locks which is not true for flock locks, so
1233 just exclude such locks from consideration.
1237 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1238 Details : Use fsfilt_sync() for both the whole filesystem sync and
1239 individual file sync to eliminate dangerous inode locking
1240 with I_LOCK that can lead to a deadlock.
1244 Description: performance in 1.6.3
1245 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1246 sure that 1MB requests are not fragmented by the block layer.
1248 Severity : enhancement
1250 Description: organize the server-side client stats on per-nid basis
1251 Details : Change the structure of stats under obdfilter and mds to
1259 The "uuid"s file would list the uuids of _active_ exports.
1260 And the clear entry is to clear all stats and stale nids.
1262 Severity : enhancement
1264 Description: eliminate client locks in face of contention
1265 Details : file contention detection and lockless i/o implementation
1266 for contended files.
1270 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1273 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1275 Description: MDS may lose file striping (and hence file data) in some cases
1276 Details : If there are additional extended attributes stored on the MDS,
1277 in particular ACLs, SELinux, or user attributes (if user_xattr
1278 is specified for the client mount options) then there is a risk
1279 of attribute loss. Additionally, the Lustre file striping
1280 needs to be larger than default (e.g. striped over all OSTs),
1281 and an additional attribute must be stored initially in the
1282 inode and then increase in size enough to be moved to the
1283 external attribute block (e.g. ACL growing in size) for file
1288 Description: skiplist implementation simplification
1289 Details : skiplists are used to group compatible locks on granted list
1290 that was implemented as tracking first and last lock of each
1291 lock group the patch changes that to using doubly linked lists
1295 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1296 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1297 because LASSERT on both the data supplied by a client, and the
1298 data on disk is dangerous and incorrect.
1300 Severity : enhancement
1302 Description: Slow truncate/writes to huge files at high offsets.
1303 Details : Directly associate cached pages to lock that protect those pages,
1304 this allows us to quickly find what pages to write and remove
1305 once lock callback is received.
1309 Description: more ldlm soft lockups
1310 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1311 starve other threads from the resource lock for a long time in
1312 case of long waiting queue, so change the debug level from
1313 D_OTHER to the less frequently used D_INFO.
1315 Severity : enhancement
1317 Description: add -gid, -group, -uid, -user options to lfs find
1321 Description: Hung threads in invalidate_inode_pages2_range
1322 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1323 one is completed. As a result, some RPCs are stuck in the queue
1328 Description: filter threads hungs on waiting journal commit
1329 Details : Cleanup filter group llog code, then only filter group llog will
1330 be only created in the MDS/OST syncing process.
1334 Description: Procfs and llog threads access destoryed import sometimes.
1335 Details : Sync the import destoryed process with procfs and llog threads by
1336 the import refcount and semaphore.
1338 Severity : enhancement
1340 Description: openlock cache of b1_6 port to HEAD
1345 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1346 Details : If insertion of an extent fails, then discard the inode
1347 preallocation and free data blocks else it can lead to duplicate
1352 Description: don't always update ctime in ext3_xattr_set_handle()
1353 Details : Current xattr code updates the inode ctime in ext3_xattr_set_handle.
1354 In some cases the ctime should not be updated, for example for
1355 2.0->1.8 compatibility it is necessary to delete an xattr and it
1356 should not update the ctime.
1360 Bugzilla : 15713/16362
1361 Description: Assertion in iopen_connect_dentry in 1.6.3
1362 Details : looking up an inode via iopen with the wrong generation number can
1363 populate the dcache with a disconneced dentry while the inode
1364 number is in the process of being reallocated. This causes an
1365 assertion failure in iopen since the inode's dentry list contains
1366 both a connected and disconnected dentry.
1370 Description: assertion failure in ldlm_handle2lock()
1371 Details : fix a race between class_handle_unhash() and class_handle2object()
1372 introduced in lustre 1.6.5 by bug 13622.
1377 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1378 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1383 Description: oops in sd_iostats_seq_show()
1384 Details : unloading/reloading the scsi low level driver triggers a kernel
1385 bug when trying to access the sd iostat file.
1390 Description: Kernel panics during QLogic driver reload
1391 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1392 patch, causing memory corruption.
1397 Description: journal_dev option does not work in b1_6
1398 Details : pass mount option during pre-mount.
1400 Severity : enhancement
1402 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1403 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1404 extent information of a file. It can be used to map logical blocks
1405 in a file to physical blocks in the block device.
1409 Description: LDLM soft lockups - improvement
1410 Details : It is be possible to send the lock handle along with each read
1411 or write request because the client is already doing a lock match
1412 itself so there isn't any reason the OST should have to re-do that
1416 Frequency : only on Cray X2
1418 Description: X2 build failures
1419 Details : fix build failures on Cray X2.
1423 Description: xid & resent requests
1424 Details : Initialize RPC XID from clock at startup (randomly if clock is
1427 Severity : enhancement
1429 Description: Add lustre_start utility to start or stop multiple Lustre servers
1434 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1435 Details : In case of memory pressure, list_del() can be called twice on
1436 req->rq_history_list, causing a kernel oops.
1438 --------------------------------------------------------------------------------
1440 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1442 * Support for kernels:
1443 2.6.5-7.283 (SLES 9),
1444 2.6.9-55.EL (RHEL 4),
1445 2.6.16.46-0.14 (SLES 10),
1446 2.6.18.8 vanilla (kernel.org)
1447 * Client support for unpatched kernels:
1448 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1449 2.6.16 - 2.6.22 vanilla (kernel.org)
1450 * Due to recently discovered recovery problems, we do not recommend
1451 using patchless RHEL 4 clients with this or any earlier release.
1452 * Recommended e2fsprogs version: 1.39.cfs8
1453 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1454 * Starting with this release, the ldiskfs backing filesystem required
1455 by Lustre is now in its own package, lustre-ldiskfs. This package
1456 should be installed. It is versioned separately from Lustre and
1457 may be released separately in future.
1461 Description: block reactivating mgc import until all deactivates complete
1462 Details : Fix race when failing back MDT/MGS to itself (testing)
1465 Frequency : at statup only
1467 Description: mds_lov_synchronize race leads to various problems
1468 Details : simultaneous MDT->OST connections at startup can cause the
1469 sync to abort, leaving the OSC in a bad state.
1471 Severity : enhancement
1473 Description: add optional extra BUILD_VERSION info
1474 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1475 to override the lustre version.
1478 Frequency : 2.6.18 servers only
1480 Description: ll_kern_mount() doesn't release the module reference
1481 Details : The ldiskfs module reference count never drops down to 0
1482 because ll_kern_mount() doesn't release the module reference.
1487 Description: server LBUG when using old ost_num_threads parameter
1488 Details : Accept the old ost_num_threads parameter but warn that it
1489 is deprecated, and fix an off-by-one error that caused an LBUG.
1494 Description: Transient SCSI error results in persistent IO issue
1495 Details : iobuf->dr_error is not reinitialized to 0 between two
1499 Frequency : sometimes when underlying device returns I/O errors
1501 Description: OSTs not going read-only during write failures
1502 Details : OSTs are not remounted read-only when the journal commit threads
1503 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1504 instead of the ext3 wrappers.
1508 Description: poor connect scaling with increasing client count
1509 Details : Don't run filter_grant_sanity_check for more than 100 exports
1510 to improve scaling for large numbers of clients.
1513 Frequency : SLES10 only
1515 Description: sanity-quota.sh quotacheck failed: rc = -22
1516 Details : Quotas cannot be enabled on SLES10.
1519 Frequency : liblustre clients only
1521 Description: getdirentries does not give error when run on compute nodes
1522 Details : getdirentries does not fail when the size specified as an argument
1523 is too small to contain at least one entry
1525 Severity : enhancement
1527 Description: Add LNET router traceability for debug purposes
1528 Details : If a checksum failure occurs with a router as part of the
1529 IO path, the NID of the last router that forwarded the bulk data
1530 is printed so it can be identified.
1535 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1536 Details : Due to a race condition, liblustre clients were occasionally
1537 evicted incorrectly.
1539 Severity : enhancement
1541 Description: lfs setstripe use optional parameters instead of postional
1544 Severity : enhancement
1546 Description: Nanosecond timestamp support for ldiskfs
1547 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1548 resolution timestamps. There is not yet support for this at
1549 the Lustre filesystem level.
1552 Frequency : during server recovery
1554 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1555 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1559 Frequency : nfs export on patchless client
1561 Description: connectathon hang when test nfs export over patchless client
1562 Details : Disconnected dentry cannot be found with lookup, so we do not need
1563 to unhash it or make it invalid
1566 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1567 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1568 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1573 Description: open req refcounting wrong on reconnect
1574 Details : If reconnect happened between getting open reply from server and
1575 call to mdc_set_replay_data in ll_file_open, we will schedule
1576 replay for unreferenced request that we are about to free.
1577 Subsequent close will crash in variety of ways.
1578 Check that request is still eligible for replay in
1579 mdc_set_replay_data().
1584 Description: disable writes to filesystem when reading health_check file
1585 Details : the default for reading the health_check proc file has changed
1586 to NOT do a journal transaction and write to disk, because this
1587 can cause reads of the /proc file to hang and block HA state
1588 checking on a healthy but otherwise heavily loaded system. It
1589 is possible to return to the previous behaviour during configure
1590 with --enable-health-write.
1592 Severity : enhancement
1594 Description: 64-bit inode version
1595 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1596 to the inode. This will be required for version-based recovery.
1601 Description: MDS fails to start if a duplicate client export is detected
1602 Details : in some rare cases it was possible for a client to connect to
1603 an MDS multiple times. Upon recovery the MDS would detect this
1604 and fail during startup. Handle this more gracefully.
1606 Severity : enhancement
1608 Description: Add -o localflock option to simulate old noflock
1610 Details : This will achieve local-only flock/fcntl locks
1616 Description: log_commit_thread vs filter_destroy race leads to crash
1617 Details : Take import reference before releasing llog record semaphore
1622 Description: Wrong request locking in request set processing
1623 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1624 imp_delayed_list, in this place should be used imp_lock.
1627 Frequency : when reconnection
1629 Description: Grant Leak when osc reconnect to OST
1630 Details : When osc reconnect ost, OST(filter) should check whether it
1631 should grant more space to client by comparing fed_grant and
1632 cl_avail_grant, and return the granted space to client instead
1633 of "new granted" space, because client will call osc_init_grant
1634 to update the client grant space info.
1637 Frequency : when client reconnect to OST
1639 Description: Grant Leak when osc do resend and replay bulk write
1640 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1641 bulk write request, because the grant info while be sync between
1642 OSC and OST when reconnect, and we should ignore the grant info
1643 these of resend/replay write req.
1648 Description: Grant space more than avaiable left space sometimes.
1649 Details : When then OST is about to be full, if two bulk writing from
1650 different clients came to OST. Accord the avaliable space of the
1651 OST, the first req should be permitted, and the second one
1652 should be denied by ENOSPC. But if the seconde arrived before
1653 the first one is commited. The OST might wrongly permit second
1654 writing, which will cause grant space > avaiable space.
1657 Frequency : when client is evicted
1659 Description: Grant might be wrongly erased when osc is evicted by OST
1660 Details : when the import is evicted by server, it will fork another
1661 thread ptlrpc_invalidate_import_thread to invalidate the
1662 import, where the grant will be set to 0. While the original
1663 thread will update the grant it got when connecting. So if
1664 the former happened latter, the grant will be wrongly errased
1665 because of this race.
1670 Description: Checking Stale with correct fid
1671 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1672 to check whether it is stale, because sometimes, we want the
1673 enqueue happened anyway, and op_data.fid2 will not be initialized.
1675 Severity : enhancement
1677 Description: update patchless client
1678 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1681 Frequency : only with 2.4 kernel
1683 Description: random memory corruption
1684 Details : size of struct ll_inode_info is to big for union inode.u and this
1685 can be cause of random memory corruption.
1690 Description: Memory leak in recovery
1691 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1692 It should also check obd_fail before fsfilt_start, otherwise if
1693 fsfilt_start return -EROFS,(failover mds during mds recovery).
1694 then the req will return with repmsg->transno = 0 and rc = EROFS.
1695 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1696 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1697 be freed no matter whether fsfilt_commit success or not.
1700 Frequency : only with huge count clients
1702 Description: Prevents from taking the superblock lock in llap_from_page for
1704 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1705 for llap_from_page call in ll_removepage prevents from taking the
1706 superblock lock for a soon died page.
1711 Description: Not check open intent error before release open handle
1712 Details : in some rare cases, the open intent error is not checked before
1713 release open handle, which may cause
1714 ASSERTION(open_req->rq_transno != 0), because it tries to release
1715 the failed open handle.
1720 Description: Set cat log bitmap only after create log success.
1721 Details : in some rare cases, the cat log bitmap is set too early. and it
1722 should be set only after create log success.
1726 Description: Accessing a block bevice can re-enable I/O when Lustre is
1727 tearing down a device.
1728 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1733 Description: service threads may hog cpus when there are a lot of requests
1734 Details : Insert cond_resched to give other threads a chance to use some CPU
1739 Description: the cat log was not initialized in recovery
1740 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1741 unlink log on mds will not be initialized until mds post
1742 recovery. And also in mds post recovery, the unlink log will
1743 initialization will be done asynchronausly, so there will be race
1744 between add unlink log and unlink log initialization.
1748 Description: brw_stats were being printed incorrectly
1749 Details : brw_stats were being printed as log2 but all of them were not
1750 recorded as log2. Also remove some code duplication arising from
1751 filter_tally_{read,write}.
1755 Frequency : rare, only in recovery.
1756 Description: ASSERTION(req->rq_type != LI_POISON) failed
1757 Details : imp_lock should be held while iterating over imp_sending_list for
1758 prevent destroy request after get timeout in ptlrpc_queue_wait.
1762 Description: replay-single.sh test 52 fails
1763 Details : A lock's skiplist need to be cleanup when it being unlinked
1764 from its resource list.
1768 Description: Short directio read returns full requested size rather than
1770 Details : Direct I/O operations should return actual amount of bytes
1771 transferred rather than requested size.
1773 Severity : enhancement
1775 Description: metadata RPC reduction (e.g. for rm performance)
1776 Details : decrease the amount of synchronous RPC between clients and servers
1777 by canceling conflicing lock before the operation on the client
1778 and packing thier handles into the main operation RPC to server.
1780 Severity : enhancement
1782 Description: Async OSC create to avoid the blocking unnecessarily.
1783 Details : If a OST has no remain object, system will block on the creating
1784 when need to create a new object on this OST. Now, ways use
1785 pre-created objects when available, instead of blocking on an
1786 empty osc while others are not empty. If we must block, we block
1787 for the shortest possible period of time.
1791 Description: improve handling recoverable errors
1792 Details : if request processig with error which can be recoverable on server
1793 request should be resend, otherwise page released from cache and
1796 Severity : enhancement
1798 Description: refine locking for avoid write wrong info into lov_objid file
1799 Details : fix possible races with add new target and write/update data in
1802 --------------------------------------------------------------------------------
1804 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1809 Frequency : on some architectures
1811 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1812 Details : Uninitialized flags sometimes cause configuration commands to
1816 Frequency : patchless clients only
1818 Description: missing __iget() symbol export
1819 Details : The __iget() symbol export is missing. To avoid the need for
1820 this on patchless clients the deathrow inode reaper is turned
1821 off, and we depend on the VM to clean up old inodes. This
1822 dependency was during via the fix for bug 12181.
1824 --------------------------------------------------------------------------------
1826 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1828 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1829 INTEROPERATE with older versions automatically. In many cases a
1830 special upgrade step is needed. Please read the
1831 user documentation before upgrading any part of a live system.
1832 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1833 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1834 * WARNING: Lustre configuration and startup changes are required with
1835 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1839 Severity : enhancement
1841 Description: MountConf
1842 Details : Lustre configuration is now managed via mkfs and mount
1843 commands instead of lmc and lconf. New obd types (MGS, MGC)
1844 are added for dynamic configuration management. See
1845 https://mail.clusterfs.com/wikis/lustre/MountConf for
1848 Severity : enhancement
1850 Description: dynamic OST addition
1851 Details : OSTs can now be added to a live filesystem
1853 Severity : enhancement
1855 Description: startup order invariance
1856 Details : MDTs and OSTs can be started in any order. Clients only
1857 require the MDT to complete startup.
1859 Severity : enhancement
1861 Description: parallel, asynchronous orphan cleanup
1862 Details : orphan cleanup is now performed in separate threads for each
1863 OST, allowing parallel non-blocking operation.
1865 Severity : enhancement
1867 Description: optimized stripe assignment
1868 Details : stripe assignments are now made based on ost space available,
1869 ost previous usage, and OSS previous usage, in order to try
1870 to optimize storage space and networking resources.
1872 Severity : enhancement
1874 Description: Permanently set tunables
1875 Details : All writable /proc/fs/lustre tunables can now be permanently
1876 set on a per-server basis, at mkfs time or on a live system.
1878 Severity : enhancement
1880 Description: Lustre message v2
1881 Details : Add lustre message format v2.
1883 Severity : enhancement
1885 Description: client OST exclusion list
1886 Details : Clients can be started with a list of OSTs that should be
1887 declared "inactive" for known non-responsive OSTs.
1891 Description: SPEC SFS validation failure on NFS v2 over lustre.
1892 Details : Changes the blocksize for regular files to be 2x RPC size,
1893 and not depend on stripe size.
1895 Severity : enhancement
1897 Description: Multiple MD RPCs in flight.
1898 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1899 To avoid overly-overloading MDS, introduce a limit on number of
1900 MDS RPCs in flight for a single client and add /proc controls
1901 to adjust this limit.
1903 Severity : enhancement
1905 Description: client read/write statistics
1906 Details : Add client read/write call usage stats for performance
1907 analysis of user processes.
1908 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1909 file access. extents_stats shows chunk size distribution.
1910 extents_stats_per_process show chunk size distribution per
1913 Severity : enhancement
1915 Description: mds statistics
1916 Details : Add detailed mds operations statistics in
1917 /proc/fs/lustre/mds/*/stats.
1921 Description: Failure of copying files with lustre special EAs.
1922 Details : Client side always return success for setxattr call for lustre
1923 special xattr (currently only "trusted.lov").
1928 Description: Refcount LNET uuids
1929 Details : The global LNET uuid list grew linearly with every startup;
1930 refcount repeated list entries instead of always adding to
1933 Severity : enhancement
1935 Description: Dynamic service threads
1936 Details : Within a small range, start extra service threads
1937 automatically when the request queue builds up.
1940 Frequency : mixed-endian client/server environments
1942 Description: mixed-endian crashes
1943 Details : The new msg_v2 system had some failures in mixed-endian
1946 Severity : enhancement
1948 Description: Easy OST removal
1949 Details : OSTs can be permanently deactivated with e.g. 'lctl
1950 conf_param lustre-OST0001.osc.active=0'
1952 Severity : enhancement
1954 Description: MGS proc entries
1955 Details : Added basic proc entries for the MGS showing what filesystems
1958 Severity : enhancement
1960 Description: provide MGS failover
1961 Details : Added config lock reacquisition after MGS server failover.
1963 Severity : enhancement
1965 Description: add Linux 2.4 support
1966 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1970 Description: a large application tries to do I/O to the same resource and dies
1971 in the middle of it.
1972 Details : Check the req->rq_arrival time after the call to
1973 ost_brw_lock_get(), but before we do anything about
1974 processing it & sending the BULK transfer request. This
1975 should help move old stale pending locks off the queue as
1976 quickly as obd_timeout.
1979 Frequency : when an incorrect nid is specified during startup
1981 Description: ptlrpc connect to non-existant node causes kernel crash
1982 Details : LNET can't be re-entered from an event callback, which
1983 happened when we expire a message after the export has been
1984 cleaned up. Instead, hand the zombie cleanup off to another
1987 Severity : enhancement
1989 Description: plain/inodebits lock performance improvement
1990 Details : Grouping plain/inodebits in granted list by their request modes
1991 and bits policy, thus improving the performance of search through
1995 Frequency : only if OST filesystem is corrupted
1997 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1998 Details : for a short time RPCs with bulk IO are in the replay list,
1999 but replay of bulk IOs is unimplemented. If the OST filesystem
2000 is corrupted due to disk cache incoherency and then replay is
2001 started it is possible to trip an assertion. Avoid putting
2002 committed RPCs into the replay list at all to avoid this issue.
2005 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2007 Description: System hang on startup
2008 Details : This bug allowed the liblustre (e.g. catamount) client to
2009 return to the app before handling all startup RPCs. This
2010 could leave the node unresponsive to lustre network traffic
2011 and manifested as a server ptllnd timeout.
2013 Severity : enhancement
2015 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2016 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2017 boolean to control whether to print peer debug info when a
2018 client's RPC times out.
2021 Frequency : only for kernels with patches from Lustre below 1.4.3
2023 Description: Remove old rdonly API
2024 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2027 Frequency : only for devices with external journals
2029 Description: Set external device read-only also
2030 Details : During a commanded failover stop, we set the disk device
2031 read-only while the server shuts down. We now also set any
2032 external journal device read-only at the same time.
2035 Frequency : when upgrading from 1.4 while trying to change parameters
2037 Description: The wrong (new) MDC name was used when setting parameters for
2038 upgraded MDT's. Also allows changing of OSC (and MDC)
2039 parameters if --writeconf is specified at tunefs upgrade time.
2042 Frequency : when setting specific ost indicies
2044 Description: QOS code breaks on skipped indicies
2045 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2046 created with --index need not be sequential.
2048 Severity : enhancement
2050 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2051 Details : The uninit_groups feature works in conjunction with the kernel
2052 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2053 up the pass1 processing of e2fsck. This is a read-only feature
2054 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2055 mount filesystems that have had this feature enabled.
2057 Severity : enhancement
2059 Description: Improve multi-block allocation algorithm to avoid fragmentation
2060 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2061 allocation locality and avoid filesystem fragmentation.
2063 ------------------------------------------------------------------------------
2065 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2067 * Support for kernels:
2068 2.4.21-47.0.1.EL (RHEL 3)
2069 2.6.5-7.283 (SLES 9)
2070 2.6.9-42.0.10.EL (RHEL 4)
2071 2.6.12.6 vanilla (kernel.org)
2072 2.6.16.27-0.9 (SLES 10)
2073 * Recommended e2fsprogs version: 1.39.cfs5
2075 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2079 Frequency : occasional, depends on client load and configuration
2080 Bugzilla : 12181, 12203
2081 Description: data loss for recently-modified files
2083 Details : In some cases it is possible that recently written or created
2084 files may not be written to disk in a timely manner (this should
2085 normally be within 30s unless client IO load is very high).
2086 The problem appears as zero-length files or files that are a
2087 multiple of 1MB in size after a client crash or client eviction
2088 that are missing data at the end of the file.
2090 This problem is more likely to be hit on clients where files are
2091 repeatedly created and unlinked in the same directory, clients
2092 have a large amount of RAM, have many CPUs, the filesystem has
2093 many OSTs, the clients are rebooted frequently, and/or the files
2094 are not accessed by other nodes after being written.
2096 The presence of the problem can be detected by looking at
2097 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2098 smaller than the second (nr_unused) then dirty files will not
2099 be flushed automatically to disk. "sync; sleep 10" should be
2100 run several times on the node before unmounting it to update
2101 Lustre (this is also safe to run on nodes without this problem).
2103 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2104 that can cause this same problem, so customers using that kernel
2105 also need to update the kernel in addition to Lustre. In order
2106 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2108 It is normal that files written just before a client crash (less
2109 than 30s) may not yet have been flushed to disk, even for local
2113 Frequency : frequent on thin XT3 nodes
2115 Description: UUID collision on thin XT3 Linux nodes
2116 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2117 randomly, since we relied on an insufficiently-seeded PRNG.
2122 Description: OSS hangs after "All ost request buffers busy"
2123 Details : A deadlock between quota and journal operations caused OSS
2124 hangs after printing "All ost request buffers busy."
2127 Frequency : always on liblustre builds
2129 Description: Cleanup compiler warnings on liblustre
2132 Frequency : always on liblustre builds on XT3
2134 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2138 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2139 > 32000 subdirectories
2140 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2141 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2142 subdirectory count crosses 32000. This will aid e2fsck to
2143 correctly handle more than 32000 subdirectories.
2146 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2148 Description: System hang on startup
2149 Details : This bug allowed the liblustre (e.g. catamount) client to
2150 return to the app before handling all startup RPCs. This
2151 could leave the node unresponsive to lustre network traffic
2152 and manifested as a server ptllnd timeout.
2154 Severity : enhancement
2156 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2157 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2158 boolean to control whether to print peer debug info when a
2159 client's RPC times out.
2164 Description: make O_SYNC working on 2.6 kernels
2165 Details : 2.6 kernels use different method for mark pages for write,
2166 so need add a code to lustre for O_SYNC work.
2171 Description: Failure to close file and release space on NFS
2172 Details : Put inode details into lock acquired in ll_intent_file_open.
2173 Use mdc_intent_lock in ll_intent_open to properly
2174 detect all kind of errors unhandled by mdc_enqueue.
2179 Description: proc file read during shutdown sometimes raced obd removal,
2181 Details : Add lock to prevent obd access after proc file removal.
2184 Frequency : Only for files larger than 4GB on 32-bit clients.
2186 Description: improperly doing page alignment of locks
2187 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2188 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2193 Description: RPCs being resent when they shouldn't be
2194 Details : Some RPCs that should not be resent are being resent. This
2195 can cause inconsistencies in the RPC state machine. Do not
2196 resend such requests.
2199 Frequency : rare, only with NFS export
2201 Description: Crash on NFS re-export node
2202 Details : under very unusual load conditions an assertion is hit in
2203 ll_intent_file_open()
2206 Frequency : only if OST filesystem is corrupted
2208 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2209 Details : for a short time RPCs with bulk IO are in the replay list,
2210 but replay of bulk IOs is unimplemented. If the OST filesystem
2211 is corrupted due to disk cache incoherency and then replay is
2212 started it is possible to trip an assertion. Avoid putting
2213 committed RPCs into the replay list at all to avoid this issue.
2218 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2219 Details : Large single O_DIRECT read and write calls can fail to allocate
2220 a sufficiently large buffer to process the request. In case of
2221 allocation failure the allocation is retried with a smaller
2222 buffer and broken into smaller requests.
2224 Severity : enhancement
2226 Description: Add -o localflock option to simulate old noflock behaviour.
2227 Details : This will achieve local-only flock/fcntl locks coherentness.
2232 Description: versioning check is incomplete
2233 Details : Checking the version difference of client vs. server, report
2234 error if the gap is too big.
2239 Description: add support PG_writeback bit
2240 Details : add support for PG_writeback bit for Lustre, for more carefull
2241 work with page cache in 2.6 kernel. This also fix some deadlocks
2242 and remove hack for work O_SYNC with 2.6 kernel.
2244 Severity : enhancement
2246 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2247 Details : The uninit_groups feature works in conjunction with the kernel
2248 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2249 up the pass1 processing of e2fsck. This is a read-only feature
2250 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2251 mount filesystems that have had this feature enabled.
2253 Severity : enhancement
2255 Description: Improve multi-block allocation algorithm to avoid fragmentation
2256 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2257 allocation locality and avoid filesystem fragmentation.
2259 ------------------------------------------------------------------------------
2261 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2263 * Support for kernels:
2264 2.6.9-42.0.3.EL (RHEL 4)
2265 2.6.5-7.276 (SLES 9)
2266 2.4.21-47.0.1.EL (RHEL 3)
2267 2.6.12.6 vanilla (kernel.org)
2268 2.6.16.21-0.8 (SLES10)
2269 * Recommended e2fsprogs version: 1.39.cfs2-0
2271 * The backwards-compatible /proc/sys/portals symlink has been removed
2272 in this release. Before upgrading, please ensure that you change
2273 any configuration scripts or /etc/sysctl.conf files that access
2274 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2275 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2276 in advance of the upgrade on any system running Lustre 1.4.6 or
2277 newer, since /proc/sys/lnet was added in that version.
2278 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2282 Frequency : only when quota is used
2284 Description: avoid scanning export list for quota master
2285 Details : Change the algorithms to avoid scanning export list in order
2286 to improve the efficiency.
2289 Frequency : MDS failover only, very rarely
2291 Description: "went back in time" messages on mds failover
2292 Details : The greatest transno may be lost when the current operation
2293 finishes with an error (transno==0) and the client's last_rcvd
2294 record is over-written. Save the greatest transno in the
2295 mds_last_transno for this case.
2298 Frequency : always for specific kernels and striping counts
2300 Description: client may get "Matching packet too big" without ACL support
2301 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2302 when trying to access files in certain configurations. The
2303 clients should in fact be denied when mounting because they do
2304 not understand ACLs.
2307 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2309 Description: many clients connecting with IO in progress causes connect timeouts
2310 Details : Avoid synchronous journal commits to avoid delays caused by many
2311 clients connecting/disconnecting when bulk IO is in progress.
2312 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2313 OST_IO_PORTAL to avoid delays behind potentially many pending
2317 Frequency : occasionally with multiple writers to a single file
2319 Description: shared writes to file may result in wrong size reported by stat()
2320 Details : Allow growing of kms when extent lock is cancelled
2323 Frequency : always with random mmap IO to multi-striped file
2325 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2326 Details : Only if the hole is at the end of OST object so that kms is too
2327 small. Fix is to increase kms accordingly in ll_nopage.
2330 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2332 Description: writes to a missing object would leak memory on the OST
2333 Details : If there is an inconsistency between the MDS and OST filesystems,
2334 such that the MDS references an object that doesn't exist, writes
2335 to that object will leak memory due to incorrect cleanup in the
2336 error handling path, eventually running out of memory on the OST.
2341 Description: Creating too long symlink causes lustre errors
2342 Details : Check symlink and name lengths before sending requests to MDS.
2345 Frequency : only if flock is enabled (not on by default)
2347 Description: posix locks not released on fd closure on 2.6.9+
2348 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2349 kernels, this caused such locks not to be released on fd close and
2350 then assertions on fs unmount about still used locks.
2353 Frequency : MDS failover only, very rarely
2355 Description: clients may get ASSERTION(granted_lock != NULL)
2356 Details : When request was taking a long time, and a client was resending
2357 a getattr by name lock request. The were multiple lock requests
2358 with the same client lock handle and
2359 mds_getattr_name->fixup_handle_for_resent_request found one of the
2360 lock handles but later failed with ASSERTION(granted_lock != NULL).
2365 Description: handle->h_buffer_credits > 0, assertion failure
2366 Details : h_buffer_credits is zero after truncate, causing assertion
2367 failure. This patch extends the transaction or creates a new
2371 Frequency : NFS re-export or patchless client
2372 Bugzilla : 11179, 10796
2373 Description: Crash on NFS re-export node (__d_move)
2374 Details : We do not want to hash the dentry if we don't have a lock.
2375 But if this dentry is later used in d_move, we'd hit uninitialised
2376 list head d_hash, so we just do this to init d_hash field but
2377 leave dentry unhashed.
2380 Frequency : NFS re-export or patchless client
2382 Description: NFS exports has problem with symbolic link
2383 Details : lustre client didn't properly install dentry when re-exported
2384 to NFS or running patchless client.
2387 Frequency : NFS re-export or patchless client
2389 Description: Various nfs/patchless fixes.
2390 Details : fixes reuse disconected alias for lookup process - this fixes
2391 warning "find_exported_dentry: npd != pd",
2392 fix permission error with open files at nfs.
2393 fix apply umask when do revalidate.
2396 Frequency : occasional
2398 Description: Crash on NFS re-export node
2399 Details : calling clear_page() on the wrong pointer triggered oops in
2400 generic_mapping_read().
2403 Frequency : rarely, using O_DIRECT IO
2405 Description: unaligned directio crashes client with LASSERT
2406 Details : check for unaligned buffers before trying any requests.
2409 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2411 Description: stale data returned from RAID cache
2412 Details : If only a small amount of IO is done to the RAID device before
2413 reading it again it is possible to get stale data from the RAID
2414 cache instead of reading it from disk.
2417 Frequency : always for sles10 kernel
2419 Description: sles10 support
2420 Details : ll_follow_link: compile fixes and using of nd_set_link
2421 under newer kernels.
2424 Frequency : depends on arch, kernel and compiler version, always on sles10
2427 Description: recursive or deep enough symlinks cause stack overflow
2428 Details : getting rid of large stack-allocated variable in
2432 Frequency : depends on hardware
2434 Description: lustre write performance loss in the SLES10 kernel
2435 Details : the performance loss is caused by using of write barriers in the
2436 ext3 code. The SLES10 kernel turns barrier support on by
2437 default. The fix is to undo that change for ldiskfs.
2439 ------------------------------------------------------------------------------
2441 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2443 * Support for kernels:
2444 2.6.9-42.0.3EL (RHEL 4)
2445 2.6.5-7.276 (SLES 9)
2446 2.4.21-47.0.1.EL (RHEL 3)
2447 2.6.12.6 vanilla (kernel.org)
2451 Frequency : quota enabled and large files being deleted
2453 Description: releasing more than 4GB of quota at once hangs OST
2454 Details : If a user deletes more than 4GB of files on a single OST it
2455 will cause the OST to spin in an infinite loop. Release
2456 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2461 Description: statfs data retrieved from /proc may be stale or zero
2462 Details : When reading per-device statfs data from /proc, in the
2463 {kbytes,files}_{total,free,avail} files, it may appear
2464 as zero or be out of date.
2467 Frequency : systems with MD RAID1 external journal devices
2469 Description: lconf's call to blkid is confused by RAID1 journal devices
2470 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2471 of returning all block devices that match the journal UUID.
2474 Frequency : always, for aggregate stripe size over 4GB
2476 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2477 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2478 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2481 Frequency : always if "lfs find" used on a local file/directory
2483 Description: "lfs find" segfaults if used on a local file/directory
2484 Details : The case where a directory component was not specified wasn't
2485 handled correctly. Handle this properly.
2488 Frequency : always on ppc64
2490 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2491 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2492 perform endianness conversion.
2495 Frequency : rarely (truncate to non-zero file size after write under load)
2496 Bugzilla : 10730, 10687
2497 Description: Files padded with zeros to next 4K multiple
2498 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2499 it is possible that files that are truncated to a non-zero size
2500 immediately after being written are filled with zero bytes beyond
2501 the truncated size. No file data is lost.
2503 Severity : enhancement
2505 Description: Allow recovery/failover for liblustre clients.
2506 Details : liblustre clients were unaware of failover configurations until
2509 Severity : enhancement
2511 Description: user file locks should fail when not mounting with flock option
2512 Details : Set up an error-returning stub in ll_file_operations.lock field
2513 to prevent incorrect behaviour when client is mounted without
2514 flock option. Also, set up properly f_op->flock field for
2518 Frequency : always on ia64
2520 Description: "lfs df" loops on printing out MDS statfs information
2521 Details : The obd_ioctl_data was not initialized and in some systems
2522 this caused a failure during the ioctl that did not return
2523 an error. Initialize the struct and return an error on failure.
2526 Frequency : SLES 9 only
2528 Description: Error of copying files with lustre special EAs as root
2529 Details : Client side always return success for setxattr call for lustre
2530 special xattr (currently only "trusted.lov").
2533 Frequency : rarely on clusters with both ia64+i386 clients
2535 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2536 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2537 clients are doing concurrent writes to the same file it is
2538 possible that the ia64 clients may LASSERT because the OST
2539 extent locks are not PAGE_SIZE aligned. Ensure that grown
2540 locks are always aligned on the request boundary.
2543 Frequency : specific use, occasional
2545 Description: Overwriting in use executable truncates on-disk binary image
2546 Details : If one node attempts to overwrite an executable in use by
2547 another node, we now correctly return ETXTBSY instead of
2548 truncating the file.
2550 Severity : enhancement
2552 Description: Async OSC create to avoid the blocking unnecessarily.
2553 Details : If a OST has no remain object, system will block on the creating
2554 when need to create a new object on this OST. Now, ways use
2555 pre-created objects when available, instead of blocking on an
2556 empty osc while others are not empty. If we must block, we block
2557 for the shortest possible period of time.
2562 Description: chmod on Lustre root is propagated to other clients
2563 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2564 invalid by the follow_mount time.
2567 Frequency : liblustre clients only
2569 Description: Race in 'instant cancel' lock handling could lead to such locks
2570 never to be granted in case of SMP MDS
2571 Details : Do not destroy not yet granted but cbpending locks in
2575 Frequency : replay/resend of open
2577 Description: non null lock assetion failure in mds_intent_policy
2578 Details : Trying to replay/resend lockless open requests resulted in
2579 mds_open() returning 0 with no lock. Now it sets a flag if
2580 a lock is going to be returned.
2582 Severity : enhancement
2584 Description: Checksum enhancements
2585 Details : New checksum enhancements allow for resending RPCs that failed
2588 Severity : enhancement
2590 Description: Tunables on number of dirty pages in cacche
2591 Details : Allow to set limit on number of dirty pages cached.
2596 Description: client crash on unmount - lock still has references
2597 Details : In some error handling cases it was possible to leak a lock
2598 reference on a client while accessing a file. This was not
2599 harmful to the client during operation, but would cause the
2600 client to crash when the filesystem is unmounted.
2603 Frequency : specific case, rare
2605 Description: ETXTBSY on mds though file not in use by client
2606 Details : ETXTBSY is no longer incorrectly returned when attempting to
2607 chmod or chown a directory that the user previously tried to
2608 execute or a currently-executing binary.
2611 Frequency : extremely rare except on liblustre-based clients
2613 Description: Lustre space not freed when files are deleted
2614 Details : Clean up open-unlinked files after client eviction. Previously
2615 the unlink was skipped and the files remained as orphans.
2620 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2621 Details : In some cases it was possible to send a blocking callback to
2622 a client doing a glimpse, even though that client didn't get
2623 a lock granted. When the glimpse lock is cancelled on the OST
2624 the freed lock is left on the waiting list and corrupted the list.
2627 Frequency : all core dumps
2629 Description: Broke core dumps to lustre
2630 Details : Negative dentry may be unhashed if parent does not have UPDATE
2631 lock, but some callers, e.g. do_coredump, expect dentry to be
2632 hashed after successful create, hash it in ll_create_it.
2634 ------------------------------------------------------------------------------
2636 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2638 * Support for kernels:
2639 2.6.9-42.0.2.EL (RHEL 4)
2640 2.6.5-7.276 (SLES 9)
2641 2.4.21-40.EL (RHEL 3)
2642 2.6.12.6 vanilla (kernel.org)
2646 Frequency : always on RHEL 3
2648 Description: Number of open files grows over time
2649 Details : The number of open files grows over time, whether or not
2650 Lustre is started. This was due to a filp leak introduced
2651 by one of our kernel patches.
2653 ------------------------------------------------------------------------------
2655 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2657 * Support for kernels:
2658 2.6.9-42.EL (RHEL 4)
2659 2.6.5-7.276 (SLES 9)
2660 2.4.21-40.EL (RHEL 3)
2661 2.6.12.6 vanilla (kernel.org)
2666 Bugzilla : 5719, 9635, 9792, 9684
2667 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2668 Details : If a server is under heavy load and cannot reply to new
2669 connection requests before the client resends the (re)connect,
2670 the connection handling code can behave badly if two service
2671 threads are concurrently handing separate (re)connections from
2672 the same client. Add better locking to the connection handling
2673 code, and ensure that only a single connection will be processed
2674 for a given client UUID, even if the lock is dropped.
2676 Severity : enhancement
2678 Description: add TCP zero-copy support to kernel
2679 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2680 sends if the hardware supports scatter-gather and checksumming.
2681 This allows socklnd to do client-write and server-read more
2682 efficiently and reduce CPU utilization from skbuf copying.
2685 Frequency : only if NFS exporting from client
2687 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2688 Details : If setting attributes on a file created under NFS that had
2689 never been opened it would be possible to oops the client
2690 if the file had no objects.
2693 Frequency : always for liblustre
2695 Description: liblustre client does MDS+OSTs setattr RPC for each write
2696 Details : When doing a write from a liblustre client, the client
2697 incorrectly issued an RPC to the MDS and each OST the file was
2698 striped over in order to update the timestamps. When writing
2699 with small chunks and many clients this could overwhelm the MDS
2700 with RPCs. In all cases it would slow down the write because
2701 these RPCs are unnecessary.
2703 Severity : enhancement
2705 Description: allow number of MDS service threads to be changed at module load
2706 Details : It is now possible to change the number of MDS service threads
2707 running. Adding "options mds mds_num_threads={N}" to the MDS's
2708 /etc/modprobe.conf will set the number of threads for the next
2709 time Lustre is restarted (assuming the "mds" module is also
2710 reloaded at that time). The default number of threads will
2711 stay the same, 32 for most systems.
2716 Description: OST crash if filesystem is unformatted or corrupt
2717 Details : If an OST is started on a device that has never been formatted
2718 or if the filesystem is corrupt and cannot even mount then the
2719 error handling cleanup routines would dereference a NULL pointer.
2724 Description: NULL pointer deref in llap_from_page.
2725 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2726 incorrect) mapping if the page was truncated/reclaimed while it was
2727 searched for. Check for this condition and skip such pages when
2728 doing readahead. Introduce extra check to llap_from_page() to
2729 verify page->mapping->host is non-NULL (so page is not anonymous).
2732 Frequency : Sometimes when using sys_sendfile
2734 Description: "page not covered by a lock" warnings from ll_readpage
2735 Details : sendfile called ll_readpage without right page locks present.
2736 Now we introduced ll_file_sendfile that does necessary locking
2737 around call to generic_file_sendfile() much like we do in
2741 Frequency : with certain MDS communication failures at client mount time
2743 Description: NULL pointer deref after failed client mount
2744 Details : a client connection request may delayed by the network layer
2745 and not be sent until after the PTLRPC layer has timed out the
2746 request. If the client fails the mount immediately it will try
2747 to clean up before the network times out the request. Add a
2748 reference from the request import to the obd device and delay
2749 the cleanup until the network drops the request.
2752 Frequency : occasionally during client (re)connect
2754 Description: assertion failure during client (re)connect
2755 Details : processing a client connection request may be delayed by the
2756 client or server longer than the client connect timeout. This
2757 causes the client to resend the connection request. If the
2758 original connection request is replied in this interval, the
2759 client may trip an assertion failure in ptlrpc_connect_interpret()
2760 which thought it would be the only running connect process.
2763 Frequency : only with obd_echo servers and clients that are rebooted
2765 Description: kernel BUG accessing uninitialized data structure
2766 Details : When running an obd_echo server it did not start the ping_evictor
2767 thread, and when a client was evicted an uninitialized data
2768 structure was accessed. Start the ping_evictor in the RPC
2769 service startup instead of the OBD startup.
2771 Severity : enhancement
2772 Bugzilla : 10193 (patchless)
2773 Description: Remove dependency on various unexported kernel interfaces.
2774 Details : No longer need reparent_to_init, exit_mm, exit_files,
2775 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2778 Frequency : rare (only users of deprecated and unsupported LDAP config)
2780 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2781 Details : LDAP apparently contains 'lustreName' attributes instead of
2782 'name'. A simple remapping of the name is sufficient.
2785 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2787 Description: waiting_locks_callback trips kernel BUG if client is evicted
2788 Details : Running with the dump_on_timeout debug flag turned on makes
2789 it possible that the waiting_locks_callback() can try to dump
2790 the Lustre kernel debug logs from an interrupt handler. Defer
2791 this log dumping to the expired_lock_main() thread.
2793 Severity : enhancement
2795 Description: Support NFS exporting on 2.6 kernels.
2796 Details : Implement non-rawops metadata methods for NFS server to use without
2797 changing NFS server code.
2800 Frequency : very rare (synthetic metadata workload only)
2802 Description: two racing renames might cause an MDS thread to deadlock
2803 Details : Running the "racer" program may cause one MDS thread to rename
2804 a file from being the source of a rename to being the target of
2805 a rename at exactly the same time that another thread is doing
2806 so, and the second thread has already enqueued these locks after
2807 doing a lookup of the target and is trying to relock them in
2808 order. Ensure that we don't try to re-lock the same resource.
2811 Frequency : only very large systems with liblustre clients
2813 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2814 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2815 all OSTs in parallel. This allows the eviction of stale liblustre
2816 clients to proceed much faster than if they were done in series,
2817 and also offers similar improvements for other set_info RPCs.
2822 Description: excessive CPU usage during initial read phase on client
2823 Details : During the initial read phase on a client, it would agressively
2824 retry readahead on the file, consuming too much CPU and impacting
2825 performance (since 1.4.5.8). Improve the readahead algorithm
2826 to avoid this, and also improve some other common cases (read
2827 of small files in particular, where "small" is files smaller than
2828 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2833 Description: MDS crash when receiving packet with unknown intent.
2834 Details : Do not LBUG in unknown intent case, just return -EFAULT
2836 Severity : enhancement
2837 Bugzilla : 9293, 9385
2838 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2839 Details : Do not serialize getattr (non-intent version) and statfs.
2842 Frequency : occasional, when OST network is overloaded/intermittent
2844 Description: client evicted by OST after bulk IO timeout
2845 Details : If a client sends a bulk IO request (read or write) the OST
2846 may evict the client if it is unresposive to its data GET/PUT
2847 request. This is incorrect if the network is overloaded (takes
2848 too long to transfer the RPC data) or dropped the OST GET/PUT
2849 request. There is no need to evict the client at all, since
2850 the pinger and/or lock callbacks will handle this, and the
2851 client can restart the bulk request.
2854 Frequency : Always when mmapping file with no objects
2856 Description: client crashes when mmapping file with no objects
2857 Details : Check that we actually have objects in a file before doing any
2858 operations on objects in ll_vm_open, ll_vm_close and
2864 Description: Request leak when working with deleted CWD
2865 Details : Introduce advanced request refcount tracking for requests
2866 referenced from lustre intent.
2868 Severity : Enhancement
2870 Description: Cache open file handles on client.
2871 Details : MDS now will return special lock along with openhandle, if
2872 requested and client is allowed to hold openhandle, even if unused,
2873 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2874 closing files for every read/write openration.
2876 Severity : Enhancement
2878 Description: Cache open negative dentries on client when possible.
2879 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2880 negative dentries on lock revocation.
2885 Description: Remounting a client read-only wasn't possible with a zconf mount
2886 Details : It wasn't possible to remount a client read-only with llmount.
2888 Severity : enhancement
2889 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2890 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2891 MPICH is included to add Lustre-specific ADIO interfaces.
2892 This is based closely on the UFS ADIO layer and only differs
2893 in file creation, in order to allow the OST striping to be set.
2894 This is user-contributed code and not supported by CFS.
2899 Description: extended inode attributes (immutable, append-only) work improperly
2900 when 2.4 and 2.6 kernels are used on client/server or vice versa
2901 Details : Introduce kernel-independent values for these flags.
2903 Severity : enhancement
2906 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2907 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2908 granularity. Now, Fractional MB granularity is be supported,
2909 this is very useful for low memory system.
2911 Severity : enhancement
2913 Description: Getattr by fid
2914 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2915 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2916 have locks and avoids getattr rpc requests for it on every stat.
2919 Frequency : Always, for filesystems larger than 2TB
2921 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2922 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2923 filesystems larger than 2TB to be mounted with mballoc enabled.
2926 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2928 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2929 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2930 filesystems larger than 2TB is mounted then the kernel will
2931 silently corrupt the start of the filesystem. CONFIG_LBD is
2932 enabled for all CFS-supported kernels, but the possibility of
2933 this happening with a modified kernel config exists.
2935 Severity : enhancement
2937 Description: add client O_DIRECT support for 2.6 kernels
2938 Details : It is now possible to do O_DIRECT reads and writes to files
2939 in the Lustre client mountpoint on 2.6 kernel clients.
2941 Severity : enhancement
2943 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2944 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2945 parallel, not waiting for response from every OST before sending
2946 a rpc to the next OST.
2951 Description: setattr vs write race when updating file timestamps
2952 Details : Client processes that update a file timestamp into the past
2953 right after writing to the file (e.g. tar) it is possible that
2954 the updated file modification time can be reset to the current
2955 time due to a race between processing the setattr and write RPC.
2957 Severity : enhancement
2959 Description: Bring 'lfs find' closer in line with regular Linux find.
2960 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2961 -print0 options and obtains all the needed info through the lustre
2964 Severity : enhancement
2966 Description: support up to 1024 configured devices on one node
2967 Details : change obd_dev array from statically allocated to dynamically
2968 allocated structs as they are first used to reduce memory usage
2973 Description: Flush dirty partially truncated pages during truncate
2974 Details : Immediatelly flush partially truncated pages in filter_setattr,
2975 this way we completely avoid having any pages in page cache on OST
2976 and can retire ugly workarounds during writes to flush such pages.
2981 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2982 Details : protect inode from truncation within vfs_unlink() context
2983 just take a reference before calling vfs_unlink() and release it
2984 when parent's i_sem is free.
2989 Description: last_id value checked outside lock on OST caused LASSERT failure
2990 Details : If there were multiple MDS->OST object precreate requests in
2991 flight, it was possible that the OST's last object id was checked
2992 outside a lock and incorrectly tripped an assertion. Move checks
2993 inside locks, and discard old precreate requests.
2996 Frequency : always, if extents are used on OSTs
2998 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2999 Details : OSTs using the extents format would not zero the high 16 bits of
3000 the index physical block number. This is not a problem for any
3001 OST filesystems smaller than 16TB, and no kernels support ext3
3002 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3003 new/modified files) and can be fixed for existing filesystems
3004 with e2fsprogs-1.39-cfs1.
3009 Description: import connection selection may be incorrect if timer wraps
3010 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3011 import connections to be ignored if the 32-bit jiffies counter
3012 wraps. Use a 64-bit jiffies counter.
3015 Frequency : very large clusters immediately after boot
3017 Description: LNET request buffers exhausted under heavy short-term load
3018 Details : If a large number of client requests are generated on a service
3019 that has previously never seen so many requests it is possible
3020 that the request buffer growth cannot keep up with the spike in
3021 demand. Instead of dropping incoming requests, they are held in
3022 the LND until the RPC service can accept more requests.
3025 Frequency : Sometimes during replay
3027 Description: Assertion failure in ll_local_open after replay.
3028 Details : If replay happened on an open request reply before we were able
3029 to set replay handler, reply will become not swabbed tripping the
3030 assertion in ll_local_open. Now we set the handler right after
3031 recognising of open request
3034 Frequency : very rare
3036 Description: kernel reports "badness in vsnprintf"
3037 Details : Reading from the "recovery_status" /proc file in small chunks
3038 may cause a negative length in lprocfs_obd_rd_recovery_status()
3039 call to vsnprintf() (which is otherwise harmless). Exit early
3040 if there is no more space in the output buffer.
3042 Severity : enhancement
3044 Description: clear OBD RPC statistics by writing to them
3045 Details : It is now possible to clear the OBD RPC statistics by writing
3046 to the "stats" file.
3051 Description: Client mtime is not the same on different clients after utimes
3052 Details : In some cases, the client was using the utimes() syscall on
3053 a file cached on another node. The clients now validate the
3054 ctime from the MDS + OSTs to determine which one is right.
3059 Description: Inability to activate failout mode
3060 Details : lconf script incorrectly assumed that in pythong string's numeric
3061 value is used in comparisons.
3064 Frequency : always with multiple stripes per file
3066 Description: Inefficient object allocation for mutli-stripe files
3067 Details : When selecting which OSTs to stripe files over, for files with
3068 a stripe count that divides evenly into the number of OSTs,
3069 the MDS is always picking the same starting OST for each file.
3070 Return the OST selection heuristic to the original design.
3075 Description: mount failures may take full timeout to return an error
3076 Details : Under some heavy load conditions it is possible that a
3077 failed mount can wait for the full obd_timeout interval,
3078 possibly several minutes, before reporting an error.
3079 Instead return an error as soon as the status is known.
3081 Frequency : quota enabled and large files being deleted
3083 Description: releasing more than 4GB of quota at once hangs OST
3084 Details : If a user deletes more than 4GB of files on a single OST it
3085 will cause the OST to spin in an infinite loop. Release
3086 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3091 Description: statfs data retrieved from /proc may be stale or zero
3092 Details : When reading per-device statfs data from /proc, in the
3093 {kbytes,files}_{total,free,avail} files, it may appear
3094 as zero or be out of date.
3097 Frequency : systems with MD RAID1 external journal devices
3099 Description: lconf's call to blkid is confused by RAID1 journal devices
3100 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3101 of returning all block devices that match the journal UUID.
3104 Frequency : always, for aggregate stripe size over 4GB
3106 Description: assertion fails when trying to use 4GB stripe size
3107 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3108 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3111 Frequency : always on ppc64
3113 Description: the first write on an ext3 filesystem with mballoc got stuck
3114 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3115 perform endianness conversion.
3117 ------------------------------------------------------------------------------
3119 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3121 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3122 INTEROPERATE with older versions automatically. Please read the
3123 user documentation before upgrading any part of a live system.
3124 * WARNING: Lustre networking configuration changes are required with
3125 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3128 * Support for kernels:
3129 2.6.9-22.0.2.EL (RHEL 4)
3130 2.6.5-7.244 (SLES 9)
3131 2.6.12.6 vanilla (kernel.org)
3134 Severity : enhancement
3135 Bugzilla : 7981/8208
3136 Description: Introduced Lustre Networking (LNET)
3137 Details : LNET is new networking infrastructure for Lustre, it includes
3138 a reorganized network configuration mode (see the user
3139 documentation for full details) as well as support for routing
3140 between different network fabrics. Lustre Networking Devices
3141 (LNDs) for the supported network fabrics have also been
3142 created for this new infrastructure.
3144 Severity : enhancement
3145 Description: Introduced Access control lists
3146 Details : clients can set ACLs on files and directories in order to have
3147 more fine-grained permissions than the standard Unix UGO+RWX.
3148 The MDS must be started with the "-o acl" mount option.
3150 Severity : enhancement
3151 Description: Introduced filesystem quotas
3152 Details : Administrators may now establish per-user quotas on the
3155 Severity : enhancement
3157 Description: Configuration change for the XT3
3158 The PTLLND is now used to run Lustre over Portals on the XT3
3159 The configure option(s) --with-cray-portals are no longer used.
3160 Rather --with-portals=<path-to-portals-includes> is used to
3161 enable building on the XT3. In addition to enable XT3 specific
3162 features the option --enable-cray-xt3 must be used.
3167 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3168 Details : A race between allocating a new llcd and re-getting the llcd_lock
3169 allowed another thread to grab newly-allocated llcd.
3171 Severity : enhancement
3173 Description: 2.6 OST async journal commit and locking fix to improve performance
3174 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3175 2.6 kernels are now async as they already were in 2.4 kernels so
3176 that they can commit concurrently with the network bulk transfer.
3177 For block-allocated files the filter allocation semaphore is held
3178 to avoid filesystem fragmentation during allocation. BKL lock
3179 removed for 2.6 xattr operations where it is no longer needed.
3184 Description: lconf incorrectly determined whether two IP networks could talk
3185 Details : In some more complicated routing and multiple-network
3186 configurations, lconf will avoid trying to make a network
3187 connection to a disjoint part of the IP space. It was doing the
3188 math incorrectly for one set of cases.
3193 Description: Fix for potential infinite loop processing records in an llog.
3194 Details : If an llog record is corrupted/zeroed, it is possible to loop
3195 forever in llog_process(). Validate the llog record length
3196 and skip the remainder of the block on error.
3199 Frequency : occasional (liblustre only)
3201 Description: liblustre could not open files whose last component is a symlink
3202 Details : sysio_path_walk() would incorrectly pass the open intent to
3203 intermediate path components.
3206 Frequency : rare (liblustre only with non-standard tuning)
3207 Bugzilla : 7201 (7350)
3208 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3209 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3210 after it has been released, only do this for referenced locks
3212 Severity : enhancement
3214 Description: specifying an (invalid) directory default stripe_size of -1
3215 would reset the directory default striping
3216 Details : stripe_size -1 was used internally to signal directory stripe
3217 removal, now use "all default" to signal dir stripe removal
3218 as a directory striping of "all default" is not useful
3221 Frequency : common for large clusters running liblustre clients
3223 Description: doing an ls when liblustre clients are running is slow
3224 Details : sending a glimpse AST to a liblustre client waits for every AST
3225 to time out, as liblustre clients will not respond. Since they
3226 cannot cache data we refresh the OST lock LVB from disk instead.
3228 Severity : enhancement
3230 Description: doing an ls at the same time as file IO can be slow
3231 Details : enqueue and other "small" requests can be blocked behind many
3232 large IO requests. Create a new OST IO portal for non-IO
3233 requests so they can be processed faster.
3236 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3238 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3239 created file with mode 000
3240 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3241 allow an ftruncate() as a non-root user to a file with mode 000.
3242 Lustre now respects this flag to disable mode checks when
3243 truncating a file owned by the user
3246 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3249 Description: Revoking locks from clients that went dead or catatonic might take
3251 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3252 cancellation of such locks instant on servers without waiting for
3253 any reply from clients. Clients drops these locks when cancel
3254 notification from server is received without replying.
3257 Frequency : liblustre-only, when liblustre client dies or becomes busy
3259 Description: Doing ls on Linux clients can take a long time with active
3261 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3262 granting such locks to it in the first place if possible. Locks
3263 are taken by proxy on the OST during the read or write and
3264 dropped immediately afterward. Add connect flags handling, do
3265 not grant locks to liblustre clients for glimpse ASTs.
3267 Severity : enhancement
3269 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3270 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3271 sequential reads and behaves badly with applications doing random
3272 reads. Improve it by reading ahead at least read region, and
3273 avoiding excessive large RPC for small reads.
3275 Severity : enhancement
3277 Description: Creating more than 1000 files for a single job may cause a load
3278 imbalance on the OSTs if there are also a large number of OSTs.
3279 Details : qos_prep_create() uses an OST index reseed value that is an
3280 even multiple of the number of available OSTs so that if the
3281 reseed happens in the middle of the object allocation it will
3282 still utilize the OSTs as uniformly as possible.
3287 Description: OST or MDS may oops in ping_evictor_main()
3288 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3289 but doesn't restart at beginning of obd_exports_timed list
3292 Severity : enhancement
3294 Description: improve by-nid export eviction on the MDS and OST
3295 Details : allow multiple exports with the same NID to be evicted at one
3296 time without re-searching the exports list.
3299 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3301 Description: MDS may oops in groups_free()
3302 Details : in rare race conditions a newly allocated group_info struct is
3303 freed again, and this can be NULL. The 2.4 compatibility code
3304 for groups_free() checked for a NULL pointer, but 2.6 did not.
3307 Frequency : common for liblustre clients doing little filesystem IO
3308 Bugzilla : 9352, 7313
3309 Description: server may evict liblustre clients accessing contended locks
3310 Details : if a client is granted a lock or receives a completion AST
3311 with a blocking AST already set it would not reply to the AST
3312 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3315 Frequency : lfs setstripe, only systems with more than 160 OSTs
3317 Description: unable to set striping with a starting offset beyond OST 160
3318 Details : llapi_create_file() incorrectly limited the starting stripe
3319 index to the maximum single-file stripe count.
3322 Frequency : LDAP users only
3324 Description: lconf did not handle in-kernel recovery with LDAP properly
3325 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3327 Severity : enhancement
3329 Description: bind OST threads to NUMA nodes to improve performance
3330 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3331 node and do their allocations there to localize memory access
3333 Severity : enhancement
3335 Description: llmount can determine client NID directly from Myrinet (GM)
3336 Details : the client NID code from gmnalnid was moved directly into
3337 llmount, removing the need to use this or specifying the
3338 client NID explicitly when mounting GM clients with zeroconf
3341 Frequency : if client is started with down MDS
3343 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3344 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3345 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3346 ptlrpc_import_delayed_req() because we didn't check if the
3347 request was interrupted, and we also didn't break out of the
3348 event loop if there was no timeout
3353 Description: data loss during non-page-aligned writes to a single file from
3354 both multiple nodes and multiple threads on one node at same time
3355 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3356 inconsistency led to false short-reads, that were cached and later
3357 used by ->prepare_write() to fill in partially written page,
3358 leading to data loss.
3361 Frequency : always, if lconf --abort_recovery used
3363 Description: lconf --abort_recovery fails with 'Operation not supported'
3364 Details : lconf was attempting to abort recovery on the MDT device and not
3367 Severity : enhancement
3369 Description: remove cleanup logs
3370 Details : replace lconf-generated cleanup logs with lustre internal
3371 cleanup routines. Eliminates the need for client-cleanup and
3374 Severity : enhancement
3376 Description: add support for EAs (user and system) on lustre filesystems
3377 Details : it is now possible to store extended attributes in the Lustre
3378 client filesystem, and with the user_xattr mount option it
3379 is possible to allow users to store EAs on their files also
3381 Severity : enhancement
3383 Description: Add possibility (config option) to show minimal available OST free
3385 Details : When compiled with --enable-mindf configure option, statfs(2)
3386 (and so, df) will return least minimal free space available from
3387 all OSTs as amount of free space on FS, instead of summary of
3388 free spaces of all OSTs.
3390 Severity : enhancement
3392 Description: do not expand extent locks acquired on OST-side
3393 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3394 by server: they are not cached anyway.
3397 Frequency : when mmap is used/binaries executed from Lustre
3399 Description: Unmmap pages before throwing them away from read cache.
3400 Details : llap_shrink cache now attempts to unmap pages before discarding
3401 them (if unmapping failed - do not discard). SLES9 kernel has
3402 extra checks that trigger if this unmapping is not done first.
3407 Description: lconf didn't resolve symlinks before checking to see whether a
3408 given mountpoint was already in use
3411 Frequency : when migrating failover services
3412 Bugzilla : 6395, 9514
3413 Description: When migrating a subset of services from a node (e.g. failback
3414 from a failover service node) the remaining services would
3415 time out and evict clients.
3416 Details : lconf --force (implied by --failover) sets the global obd_timeout
3417 to 5 seconds in order to quickly disconnect, but this caused
3418 other RPCs to time out too quickly. Do not change the global
3419 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3421 Severity : enhancement
3422 Frequency : if MDS is started with down OST
3423 Bugzilla : 9439,5706
3424 Description: Allow startup/shutdown of an MDS without depending on the
3425 availability of the OSTs.
3426 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3427 Add appropriate locking and lov-osc refcounts for safe
3428 cleaning. Add osc abort_inflight calls in case the
3429 synchronize never started.
3432 Frequency : occasional (Cray XT3 only)
3434 Description: root not authorized to access files in CRAY_PORTALS environment
3435 Details : The client process capabilities were not honoured on the MDS in
3436 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3437 been accessed by an authorized user then root was able to access
3438 the file on the local client also. The root user capabilities
3439 are now allowed on the MDS, as this environment has secure UID.
3442 Frequency : occasional
3444 Description: ldiskfs "too long searching" message happens too often
3445 Details : A debugging message (otherwise harmless) prints too often on
3446 the OST console. This has been reduced to only happen when
3447 there are fragmentation problems on the filesystem.
3452 Description: Division by zero in statfs when all OSCs are inactive
3453 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3454 lov_statfs divides by value returned by lov_get_stripecnt().
3458 Bugzilla : 9489, 3273
3459 Description: First write from each client to each OST was only 4kB in size,
3460 to initialize client writeback cache, which caused sub-optimal
3461 RPCs and poor layout on disk for the first writen file.
3462 Details : Clients now request an initial cache grant at (re)connect time
3463 and so that they can start streaming writes to the cache right
3464 away and always do full-sized RPCs if there is enough data.
3465 If the OST is rebooted the client also re-establishes its grant
3466 so that client cached writes will be honoured under the grant.
3471 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3472 Details : Now I/O RPCs go to different portal number and (presumably) fast
3473 lock requests (and glimses) and other RPCs get their own service
3474 threads pool that should be able to service those RPCs
3477 Severity : enhancement
3479 Description: Ability to exchange lustre version between client and servers and
3480 issue warnings at client side if client is too old. Also for
3481 liblustre clients there is ability to refuse connection of too old
3483 Details : New 'version' field is added to connect data structure that is
3484 filled with version info. That info is later checked by server and
3488 Frequency : rare, liblustre only.
3489 Bugzilla : 9296, 9581
3490 Description: Two simultaneous writes from liblustre at offset within same page
3491 might proceed at the same time overwriting eachother with stale
3493 Details : I/O lock withing llu_file_prwv was released too early, before data
3494 actually was hitting the wire. Extended lock-holding time until
3495 server acknowledges receiving data.
3498 Frequency : extremely rare. Never observed in practice.
3500 Description: avoid generating lustre_handle cookie of 0.
3501 Details : class_handle_hash() generates handle cookies by incrementing
3502 global counter, and can hit 0 occasionaly (this is unlikely, but
3503 not impossible, because initial value of cookie counter is
3504 selected randonly). Value of 0 is used as a sentinel meaning
3505 "unassigned handle" --- avoid it. Also coalesce two critical
3506 sections in this function into one.
3508 Severity : enhancement
3510 Description: allow liblustre clients to delegate truncate locking to OST
3511 Details : To avoid overhead of locking, liblustre client instructs OST to
3512 take extent lock in ost_punch() on client's behalf. New connection
3513 flag is added to handle backward compatibility.
3515 Severity : enhancement
3516 Bugzilla : 4928, 7341, 9758
3517 Description: allow number of OST service threads to be specified
3518 Details : a module parameter allows the number of OST service threads
3519 to be specified via "options ost ost_num_threads={N}" in the
3520 OSS's /etc/modules.conf or /etc/modprobe.conf.
3524 Bugzilla : 6146, 9635, 9895
3525 Description: servers crash with bad pointer in target_handle_connect()
3526 Details : In rare cases when a client is reconnecting it was possible that
3527 the connection request was the last reference for that export.
3528 We would temporarily drop the export reference and get a new
3529 one, but this may have been the last reference and the export
3530 was just destroyed. Get new reference before dropping old one.
3532 Severity : enhancement
3533 Frequency : if client is started with failover MDS
3535 Description: Allow multiple MDS hostnames in the mount command
3536 Details : Try to read the configuration from all specified MDS
3537 hostnames during a client mount in case the "primary"
3540 Severity : enhancement
3542 Description: Stop sending data to evicted clients as soon as possible.
3543 Details : Check if the client we are about to send or are sending data to
3544 was evicted already. (Check is done every second of waiting,
3545 for which l_wait_event interface was extended to allow checking
3546 of exit condition at specified intervals).
3549 Frequency : rare, normally only when NFS exporting is done from client
3551 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3553 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3556 Frequency : always, if srand() is called before liblustre initialization
3558 Description: Liblustre uses system PRNG disturbing its usage by user application
3559 Details : Introduce internal to lustre fast and high-quality PRNG for
3560 lustre usage and make liblustre and some other places in generic
3561 lustre code to use it.
3563 Severity : enhancement
3564 Bugzilla : 9477, 9557, 9870
3565 Description: Verify that the MDS configuration logs are updated when xml is
3566 Details : Check if the .xml configuration logs are newer than the config
3567 logs stored on the MDS and report an error if this is the case.
3568 Request --write-conf, or allow starting with --old_conf.
3570 Severity : enhancement
3572 Description: Handle symlinks in the path when checking if Lustre is mounted.
3573 Details : Resolve intermediate symlinks when checking if a client has
3574 mounted a filesystem to avoid duplicate client mounts.
3579 Description: lconf can hit an error exception but still return success.
3580 Details : The lconf command catches the Command error exception at the top
3581 level script context and will exit with the associated exit
3582 status, but doesn't ensure that this exit status is non-zero.
3587 Description: failure of ptlrpc thread startup can cause oops
3588 Details : Starting a ptlrpc service thread can fail if there are a large
3589 number of threads or the server memory is very fragmented.
3590 Handle this without oopsing.
3593 Frequency : always, only if liblustre and non-default acceptor port was used
3595 Description: liblustre cannot connect to servers with non-default acceptor port
3596 Details : tcpnal_set_default_params() was not called and was therefore
3597 ignoring the environment varaible TCPNAL_PORT, as well as other
3598 TCPNAL_ environment variables
3603 Description: two objects could be created on the same OST for a single file
3604 Details : If an OST is down, in some cases it was possible to create two
3605 objects on a single OST for a single file. No problems other
3606 than potential performance impact and spurious error messages.
3610 Bugzilla : 5681, 9562
3611 Description: Client may oops in ll_unhash_aliases
3612 Details : Client dcache may become inconsistent in race condition.
3613 In some cases "getcwd" can fail if the current directory is
3619 Description: Inode refcounting problems in NFS export code
3620 Details : link_raw functions used to call d_instantiate without obtaining
3621 extra inode reference first.
3625 Bugzilla : 9942, 9903
3626 Description: Referencing freed requests leading to crash, memleaks with NFS.
3627 Details : We used to require that call to ll_revalidate_it was always
3628 followed by ll_lookup_it. Also with revalidate_special() it is
3629 possible to call ll_revalidate_it() twice for the same dentry
3630 even if first occurence returned success. This fix changes semantic
3631 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3632 reference on a request referred from the intent.
3633 ll_intent_release() then releases such a request.
3636 Frequency : rare, normally benchmark loads only
3638 Description: unlinked inodes were kept in memory on the client
3639 Details : If a client is repeatedly creating and unlinking files it
3640 can accumulate a lot of stale inodes in the inode slab cache.
3641 If there is no other client load running this can cause the
3642 client node to run out of memory. Instead flush old inodes
3643 from client cache that have the same inode number as a new inode.
3646 Frequency : SLES9 2.6.5 kernel and long filenames only
3647 Bugzilla : 9969, 10379
3648 Description: utime reports stale NFS file handle
3649 Details : SLES9 uses out-of-dentry names in some cases, which confused
3650 the lustre dentry revalidation. Change it to always use the
3654 Frequency : rare, unless heavy write-truncate concurrency is continuous
3655 Bugzilla : 4180, 6984, 7171, 9963, 9331
3656 Description: OST becomes very slow and/or deadlocked during object unlink
3657 Details : filter_destroy() was holding onto the parent directory lock
3658 while truncating+unlinking objects. For very large objects this
3659 may block other threads for a long time and slow overall OST
3660 responsiveness. It may also be possible to get a lock ordering
3661 deadlock in this case, or run out of journal credits because of
3662 the combined truncate+unlink. Solution is to do object truncate
3663 first in one transaction without parent lock, and then do the
3664 final unlink in a new transaction with the parent lock. This
3665 reduces the lock hold time dramatically.
3668 Frequency : rare, 2.4 kernels only
3670 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3671 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3672 before unmounting their filesystem in order to avoid deadlock.
3673 The kernel_locked() function in 2.4 kernels only checks whether
3674 the kernel lock is held, not whether it is this process that is
3675 holding it as 2.6 kernels do.
3680 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3681 Details : The client ptlrpc code may be trying to reconnect to a down
3682 server before a previous connection attempt has timed out.
3683 Increase the reconnect interval to be longer than the connection
3684 timeout interval to avoid sending duplicate connections to
3688 Frequency : echo_client brw_test command
3690 Description: fix echo_client to work with OST preallocated code
3691 Details : OST preallocation code (5137) didn't take echo_client IO path
3692 into account: echo_client calls filter methods outside of any
3693 OST thread and, hence, there is no per-thread preallocated
3694 pages and buffers to use. Solution: hijack pga pages for IO. As
3695 a byproduct, this avoids unnecessary data copying.
3699 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3700 Description: Client can oops in mdc_commit_close() after open replay
3701 Details : It was possible for the MDS to return an open request with no
3702 transaction number in mds_finish_transno() if the client was
3703 evicted, but without actually returning an error. Clients
3704 would later try to replay that open and may trip an assertion
3705 Simplify the client close codepath, and always return an error
3706 from the MDS in case the open is not successful.
3709 Frequency : rare, 2.6 OSTs only
3711 Description: OST may deadlock under high load on fragmented files
3712 Details : If there was a heavy load and highly-fragmented OST filesystems
3713 it was possible to have all the OST threads deadlock waiting on
3714 allocation of biovecs, because the biovecs were not released
3715 until the entire RPC IO was completed. Instead, release biovecs
3716 as soon as they are complete to ensure forward IO progress.
3718 Severity : enhancement
3720 Description: Support for specifying external journal device at mount
3721 Details : If an OST or MDS device is formatted with an external journal
3722 device, this device major/minor is stored in the ext3 superblock
3723 and may not be valid for failover. Allow detecting and
3724 specifying the external journal at mount time.
3729 Description: Mounting an MDS with pending unlinked files may cause oops
3730 Details : target_finish_recovery() calls mds_postrecov() which returned
3731 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3732 considers this an error and immediately begins cleaning up the
3733 lov, just after starting the mds_lov process
3735 Severity : enhancement
3737 Description: Implement 'lfs df' to report actual free space on per-OST basis
3738 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3739 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3740 usage of objects; '-h' to report in human readable format.
3742 ------------------------------------------------------------------------------
3744 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3751 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3752 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3753 at startup during orphan unlinking. Instead, skip buddy update
3754 before setup, as it will be regenerated after recovery is complete.
3759 Description: If an OST is inactive, its locks might reference stale inodes.
3760 Details : lov_change_cbdata() must iterate over all namespaces, even if
3761 they are inactive to clear inode references from the lock.
3763 Severity : enhancement
3764 Frequency : occasional, if non-standard max_dirty_mb used
3766 Description: Client will block write RPCs if not enough grant
3767 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3768 then the client will block writes while waiting for another RPC
3769 to complete instead of consuming its dirty limit. With change
3770 we get improved performance when max_dirty_mb is small.
3772 Severity : enhancement
3773 Bugzilla : 3389, 6253
3774 Description: Add support for supplementary groups on the MDS.
3775 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3776 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3777 lookups for user supplementary groups into a cache.
3781 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3782 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3783 for liblustre O_EXCL is enforced by MDS.
3787 Description: readdir never returns NULL in liblustre.
3788 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3789 dentry in struct dirent.
3793 Frequency : liblustre only, depends on application IO pattern
3794 Description: liblustre clients evicted if not contacting servers
3795 Details : Don't put liblustre clients into the ping_evictor list, so
3796 they will not be evicted by the pinger ever.
3798 Severity : enhancement
3800 Description: Add ability to evict clients by NID from MDS.
3801 Details : By echoing "nid:$NID" string into
3802 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3803 $NID would be instantly evicted from this MDS and from all active
3804 OSTs connected to it.
3808 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3809 Details : lookup_it_finish() used to query file size from OSTs that was not
3814 Description: service threads change working directory to that of init
3815 Details : Starting lustre service threads may pin the working directory
3816 of the parent thread, making that filesystem busy. Threads
3817 now change to the working directory of init to avoid this.
3821 Frequency : during shutdown only
3822 Description: shutdown with a failed MDS or OST can cause unmount to hang
3823 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3828 Frequency : 2.6 only
3829 Description: chmod/chown may include an extra supplementary group
3830 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3831 supplementary group and if none is specified this is used.
3834 Bugzilla : 5479 (6816)
3836 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3837 Details : If lookup is in progress on a file that is unlinked we might try
3838 to revalidate the inode and fail in revalidate after lookup is
3839 complete and ll_file_open() enqueues the open again but
3840 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3843 Frequency : always, if lconf --abort_recovery used
3845 Description: lconf --abort_recovery fails with 'Operation not supported'
3846 Details : lconf was attempting to abort recovery on the MDT device and not
3849 ------------------------------------------------------------------------------
3851 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3856 Frequency : rare (only unsupported configurations with a node running as an
3858 Bugzilla : 6514, 5137
3859 Description: Mounting a Lustre file system on a node running as an OST could
3861 Details : OSTs now preallocates memory needed to write out data at
3862 startup, instead of when needed, to avoid having to
3863 allocate memory in possibly low memory situations.
3864 Specifically, if the file system is mounted on on OST,
3865 memory pressure could force it to try to write out data,
3866 which it needed to allocate memory to do. Due to the low
3867 memory, it would be unable to do so and the node would
3868 become unresponsive.
3870 Severity : enhancement
3872 Description: Addition of lconf --service command line option
3873 Details : lconf now accepts a '--service <arg>' option, which is
3874 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3876 Severity : enhancement
3878 Description: Failover mode is now the default for OSTs.
3879 Details : By default, OSTs will now run in failover mode. To return to
3880 the old behaviour, add '--failout' to the lmc line for OSTs.
3882 Severity : enhancement
3884 Description: Health checks are now provided for MDS and OSTs
3885 Details : Additional detailed health check information on MSD and OSTs
3886 is now provided through the procfs health_check value.
3889 Frequency : occasional, depends on IO load
3891 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3892 after numerous create/delete cycles
3893 Details : The ext3 inode allocation policy would not allocate new inodes
3894 very well on the OSTs because there are no new directories
3895 being created. Instead we look for groups with free space if
3896 the parent directories are nearly full.
3901 Description: Network or server problems during mount may cause partially
3902 mounted clients instead of returning an error.
3903 Details : The config llog parsing code may overwrite the error return
3904 code during mount error handling, returning success instead
3910 Description: MDS can fail to allocate large reply buffers
3911 Details : After long uptimes the MDS can fail to allocate large reply
3912 buffers (e.g. zconf client mount config records) due to memory
3913 fragmentation or consumption by the buffer cache. Preallocate
3914 some large reply buffers so that these replies can be sent even
3915 under memory pressure.
3919 Frequency : rare (liblustre)
3920 Description: fsx running with liblustre complained that using truncate() to
3921 extend the file doesn't work. This patch corrects that issue.
3922 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3923 ATTR_SIZE and lsm use in llu_setattr_raw.
3927 Frequency : rare, only 2.6 kernels
3928 Description: Unusual file access patterns on the MDS may result in inode
3929 data being lost in very rare circumstances.
3930 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3931 code in the 2.6 kernel caused the inode and/or EA data not to
3932 be read from disk, causing single-file corruption.
3936 Frequency : rare, only 2.6 filesystems using extents
3937 Description: Heavy concurrent write and delete load may cause data corruption.
3938 Details : It was possible under high-load situations to have an extent
3939 metadata block in the block device cache from a just-unlinked
3940 file overwrite a newly-allocated data block. We now unmap any
3941 metadata buffers that alias just-allocated data blocks.
3945 Frequency : filesystems with default stripe_count larger than 77
3946 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3947 Details : lconf specifies an inode size of 4096 bytes when the default
3948 stripe_count is larger than 77. This conflicts with the default
3949 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3950 this case to avoid pinning too much memory for large EAs.
3952 ------------------------------------------------------------------------------
3954 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3959 Frequency : rare (extremely heavy IO load with hundreds of clients)
3961 Description: Client is evicted, gets IO error writing to file
3962 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3963 caused clients to be evicted for AST timeouts. The fixes in
3964 bug 5192 mean we no longer need to have such short AST timeouts
3965 so ldlm_timeout has been increased.
3968 Frequency : occasional during --force or --failover shutdown under load
3969 Bugzilla : 5949, 4834
3970 Description: Server oops/LBUG if stopped with --force or --failover under load
3971 Details : a collection of import/export refcount and cleanup ordering
3972 issues fixed for safer force cleanup
3975 Frequency : only filesystems larger than 120 OSTs
3976 Bugzilla : 5990, 6223
3977 Description: lfs getstripe would oops on a very large filesystem
3978 Details : lov_getconfig used kfree on vmalloc'd memory
3981 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3982 Bugzilla : 6242, 6243
3983 Description: reading from files that had been truncated to a non-zero size
3984 but never opened returned no data
3985 Details : ll_file_read() reads zeros from no-object files to EOF
3990 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3991 Details : There's little harm in aborting MDS/OSS recovery and letting it
3992 try again, so I removed the LASSERT and return an error instead.
3994 Severity : enhancement
3996 Description: New debugging infrastructure for tracking down data corruption
3997 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3998 (b) cover more of the client-side code path, and (c) try to narrow
3999 down where problems occurred
4003 Bugzilla : 3819, 4364, 4397, 6313
4004 Description: Racing close and eviction MDS could cause assertion in mds_close
4005 Details : It was possible to get multiple mfd references during close and
4006 client eviction, leading to one thread referencing a freed mfd.
4008 Severity: : enhancement
4009 Bugzilla : 3262, 6359
4010 Description: Attempts to reconnect to servers are now more aggressive.
4011 Details : This builds on the enhanced upcall-less recovery that was added
4012 in 1.4.2. When trying to reconnect to servers, clients will
4013 now try each server in the failover group every 10 seconds. By
4014 default, clients would previously try one server every 25 seconds.
4019 Description: After recovery, certain operations trigger a failed
4020 assertion on a client.
4021 Details : Failing over an mds, using lconf -d --failover, while a
4022 client was doing a readdir() call would cause the client to
4023 LBUG after recovery completed and the readdir() was resent.
4025 Severity : enhancement
4027 Description: Default groups are now added by lconf
4028 Details : You can now run lconf --group <servicename> without having to
4029 manually add groups with lmc.
4032 Frequency : occasional
4034 Description: Nodes with an elan id of 0 trigger a failed assertion
4037 Frequency : always when accessing e.g. tty/console device nodes
4039 Description: tty and some other devices nodes cannot be used on lustre
4040 Details : file's private_data field is used by device data and lustre
4041 values in there got lost. New field was added to struct file to
4042 store fs-specific private data.
4045 Frequency : when exporting Lustre via NFS
4047 Description: NFSD failed occasionally when looking up a path component
4048 Details : NFSD is looking up ".." which was broken in ext3 directories
4049 that had grown large enough to become hashed.
4052 Frequency : Clusters with multiple interfaces not on the same subnet
4054 Description: Nodes will repeatedly try to reconnect to an interface which it
4055 cannot reach and report an error to the log.
4056 Details : Extra peer list entries will be created by lconf with some peers
4057 unreachable. lconf now validates the peer before adding it.
4060 Frequency : Only if a default stripe is set on the filesystem root.
4062 Description: Setting a default stripe on the filesystem root prevented the
4063 filesystem from being remounted.
4064 Details : The client was sending extra request flags in the root getattr
4065 request and did not allocate a reply buffer for the dir EA.
4068 Frequency : occasional, higher if lots of files are accessed by one client
4069 Bugzilla : 6159, 6097
4070 Description: Client trips assertion regarding lsm mismatch/magic
4071 Details : While revalidating inodes the VFS looks up inodes with ifind()
4072 and in rare cases can find an inode that is being freed.
4073 The ll_test_inode() code will free the lsm during ifind()
4074 when it finds an existing inode and then the VFS later attaches
4075 this free lsm to a new inode.
4079 Bugzilla : 6422, 7030
4080 Description: MDS deadlock between mkdir and client eviction
4081 Details : Creating a new file via mkdir or mknod (starting a transaction
4082 and getting the ns lock) can deadlock with client eviction
4083 (gets ns lock and trying to finish a synchronous transaction).
4086 Frequency : occasional
4087 Description: While starting a server, the fsfilt_ext3 module could not be
4089 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4090 kernels. Previously, lconf would still use the ext3 name
4091 when trying to load modules. Now, it will correctly use
4092 ext3 on 2.4 and ldiskfs on 2.6.
4094 Severity : enhancement
4095 Description: The default stripe count has been changed to 1
4096 Details : The interpretation of the default stripe count (0, to lfs
4097 or lmc) has been changed to mean striping across a single
4098 OST, rather than all available. For general usage we have
4099 found a stripe count of 1 or 2 works best.
4101 Severity : enhancement
4102 Description: Add support for compiling against Cray portals.
4103 Details : Conditional compiling for some areas that are different
4107 Frequency : occasional
4108 Bugzilla : 6409, 6834
4109 Description: Creating files with an explicit stripe count may lead to
4110 a failed assertion on the MDS
4111 Details : If some OSTs are full or unavailable, creating files may
4112 trigger a failed assertion on the MDS. Now, Lustre will
4113 try to use other servers or return an error to the
4117 Frequency : occasional
4119 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4120 caused lock timeout during readahead (since 1.4.2).
4121 Details : Processes doing readahead might match a lock that hasn't been
4122 granted yet if there are overlapping and conflicting lock
4123 requests. The readahead process waits on ungranted lock
4124 (original lock is CBPENDING), while OST waits for that process
4125 to cancel CBPENDING read lock and eventually evicts client.
4127 Severity : enhancement
4129 Description: Initial enabling of flock support for clients
4130 Details : Implements fcntl advisory locking and file status functions.
4131 This feature is provided as an optional mount flag (default
4132 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4133 locking are implemented yet, and those that are are not guaranteed
4134 to be completely correct in production environments.
4135 mount -t lustre -o [flock|noflock] ...
4138 Frequency : occasional
4140 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4141 assertion in the ext3 JBD (journaling) layer.
4142 Details : The b_committed_data struct is protected by the big kernel lock
4143 in 2.4 kernels, serializing journal_commit_transaction() and
4144 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4145 access to this struct. In 2.6 kernels there is finer grained
4146 locking to improve SMP performance of the JBD layer.
4150 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4155 Description: merge and cleanup kernel patches.
4158 -----------------------------------------------------------------------------
4160 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4162 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4163 versions of Lustre. Please update all servers and clients to
4164 version 1.4.2 or later at the same time. You must also run
4165 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4166 to update the configuration logs.
4168 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4169 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4170 - swab llog records of type '0' so we get proper header size/idx (5861)
4171 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4172 - fix rename of one directory over another leaking an inode (5953)
4173 - avoid SetPageDirty on 2.6 (5981)
4174 - don't re-add just-being-destroyed locks to the waiting list (5653)
4175 - when creating new directories, inherit the parent's custom
4176 striping settings if present parent (3048)
4177 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4178 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4179 - don't hold client locks on temporary worklist from l_lru (5666)
4180 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4181 - automatically evict dead clients (5921)
4182 - Update file size properly in create+truncate+fstat case (6196)
4183 - Do not unhash mountpoint dentries, do not allow removal of
4185 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4186 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4187 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4188 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4189 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4190 - Correctly handle reads of files with no objects (6243)
4191 - lctl recover will also mark a device active if deactivate used (5933)
4193 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4194 - allow --write-conf on an MDS with different nettype than client (5619)
4195 - don't write config llogs to MDS for mounts not from that MDS (5617)
4196 - lconf should create multiple TCP connections from a client (5201)
4197 - init scripts are now turned off by default; run chkconfig --on
4198 lustre and chkconfig --on lustrefs to use them
4199 - upcalls are no longer needed for clients to recover to failover
4201 - add --abort-recovery option to lconf to abort recovery on device
4203 - add support for an arbitrary number of OSTs (3026)
4204 - Quota support protocol changes.
4205 - forward compatibility changes to wire structs (6007)
4206 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4207 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4208 - improved "lustre status" script
4209 - initialize blocksize for non-regular files (6062)
4210 - added --disable-server and --disable-client configure options (5782)
4211 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4212 - Vanilla 2.4.29 support
4213 - increase maximum number of obd devices to 520 (6242)
4214 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4215 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4216 SLES 9 2.6.5 kernels
4217 - limit stripes per file to 160 (the maximum EA size) (6093)
4219 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4222 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4223 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4224 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4225 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4226 - small ext3 extents cleanups and fixes (5733)
4227 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4228 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4229 - avoid needless client->OST connect, fix handle mismatch (5317)
4230 - fix DLM error path that led to out-of-sync client, long delays (5779)
4231 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4232 - fix several locking issues related to i_size (5492,5624,5654,5672)
4233 - don't move pending lock onto export if it is already evicted (5683)
4234 - fix kernel oops when creating .foo in unlinked directory (5548)
4235 - fix deadlock in obdfilter statistics vs. object create (5811)
4236 - use time_{before,after} to avoid timer jiffies wrap (5882)
4237 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4238 - Do not leak request if server was not able to process it (5154)
4239 - If mds_open unable to find parent dir, make that negative lookup(5154)
4240 - don't create new directories with extent-mapping (5909, 5936)
4242 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4243 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4244 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4245 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4246 - support for RHEL4 kernel on i686 (5773)
4247 - provide error messages when incompatible logs are encountered (5898)
4249 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4250 * version 1.4.0.10 (1.4.1 release candidate 1)
4252 - don't keep a lock reference when lock is not granted (4238)
4253 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4254 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4255 - drop import inflight refcount on signal_completed_replay error (5255)
4256 - unlock page after async write error during send (3677)
4257 - handle missing objects in filter_preprw_read properly (5265)
4258 - no transno return for symlink open, don't save no-trasno open (3440)
4259 - don't try to complete elan receive that already failed (4012)
4260 - free RPC server reply state on error (5406)
4261 - clean up thread from ptlrpc_start_thread() on error (5160)
4262 - readahead could read extra page into cache that wasn't ejected (5388)
4263 - prevent races in class_attach/setup/cleanup/detach (5260)
4264 - don't dereference de->d_inode after l_dput of de (5458)
4265 - use "int" for stripe value returned from lock_to_stripe (5544)
4266 - mballoc allocation and error-checking fixes in 2.6 (5504)
4267 - block device patches to fix I/O request sizes in 2.6 (5482)
4268 - look up hostnames for IB nals (5602)
4269 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4270 - don't start multiple acceptors for the same port (5277)
4271 - fix incorrect LASSERT in mds_getattr_name (5635)
4272 - export a proc file for general "ping" checking (5628)
4273 - fix "lfs check" to not block when the MDS is down (5628)
4275 - service request history (4965)
4276 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4277 - create an "evict_client" /proc entry on OSTs, like the MDS has
4278 - fix mount usage message, return errors per mount(8) (5168)
4279 - change grep [] to grep "[]" in tests so they work in more UMLs
4280 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4281 - remove ancient LOV_MAGIC_V0 EA support (5047)
4282 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4283 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4284 - allow readahead allocations to fail when low on memory (5383)
4285 - mmap locking landed again, after considerable improvement (2828)
4286 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4288 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4291 - send OST transaction number in read/write reply to free req (4966)
4292 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4293 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4294 - the watchdog thread now runs as interruptible (5246)
4295 - flock/lockf fixes (but it's still disabled, pending 5135)
4296 - don't use EXT3 constants in llite code (5094)
4297 - memory shortage at startup could cause assertion (5176)
4299 - reorganization of lov code
4300 - single portals codebase
4302 - add extents/mballoc support (5025)
4303 - direct I/O reads in the obdfilter (4048)
4304 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4306 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4309 - send OST transaction number in read/write reply to free req (4966)
4310 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4311 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4312 - the watchdog thread now runs as interruptible (5246)
4313 - handle missing objects in filter_preprw_read properly (5265)
4314 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4315 - drop import inflight refcount on signal_completed_replay error (5255)
4316 - unlock page after async write error during send (3677)
4317 - return original error code on reconstructed replies (3761)
4318 - no transno return for symlink open, don't save no-trasno open (3440)
4320 - add pid to ldlm debugging output (4922)
4321 - bump the watchdog timeouts -- we can't handle 30sec yet
4322 - extra debugging for orphan dentry/inode bug (5259)
4324 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4327 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4328 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4329 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4330 - fix NULL dereference in /proc/sys/portals/routes (4827)
4331 - allow failed mdc_close() operations to be interrupted (4561)
4332 - stop precreate on OST before MDS would time out on it (4778)
4333 - don't send partial-page writes before EOF from client (4410)
4334 - discard client grant for sub-page writes on large-page clients (4520)
4335 - don't free dentries not owned by NFS code, check generation (4806)
4336 - fix lsm leak if mds_create_objects() fails (4801)
4337 - limit debug_daemon file size, always print CERROR messages (4789)
4338 - use transno after validating reply (3892)
4339 - process timed out requests if import state changes (3754)
4340 - update mtime on OST during writes, return in glimpse (4829)
4341 - add mkfsoptions to LDAP (4679)
4342 - use ->max_readahead method instead of zapping global ra (5039)
4343 - don't interrupt __l_wait_event() during strace
4345 - add software watchdogs to catch hung threads quickly (4941)
4346 - make lustrefs init script start after nfs is mounted
4347 - fix CWARN/ERROR duplication (4930)
4348 - return async write errors to application if possible (2248)
4349 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4350 - print NAL number in %x format (4645)
4351 - update barely-supported suse-2.4.21-171 series (4842)
4352 - support for sles 9 %post scripts
4353 - support for building 2.6 kernel-source packages
4354 - support for sles km_* packages
4356 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4359 - ignore -ENOENT errors in osc_destroy (3639)
4360 - notify osc create thread that OSC is being cleaned up (4600)
4361 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4362 - reconstruct ost_handle() like mds_handle() (4657)
4363 - create a new thread to do import eviction to avoid deadlock (3969)
4364 - let lconf resolve symlinked-to devices (4629)
4365 - don't unlink "objects" from directory with default EA (4554)
4366 - hold socknal file ref over connect in case target is down (4394)
4367 - allow more than 32000 subdirectories in a single directory (3244)
4368 - fix blocks count for O_DIRECT writes (3751)
4369 - OST returns ENOSPC from object create when no space left (4539)
4370 - don't send truncate RPC if file size isn't changing (4410)
4371 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4372 - bind to privileged port in socknal and tcpnal (3689)
4374 - rate limit CERROR/CWARN console message to avoid overload (4519)
4375 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4376 - basic mmap support (3918)
4377 - kernel patch series update from b1_4 (4711)
4379 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4382 - avoid crash during MDS cleanup with OST shut down (2775)
4383 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4384 - don't use bad inodes on the MDS (3744)
4385 - dynamic object preallocation to improve recovery speed (4236)
4386 - don't hold spinlock over lock dumping or change debug flags (4401)
4387 - don't zero obd_dev when it is force cleaned (3651)
4388 - print grants to console if they go negative (4431)
4389 - "lctl deactivate" will stop automatic recovery attempts (3406)
4390 - look for existing locks in ldlm_handle_enqueue() (3764)
4391 - don't resolve lock handle twice in recovery avoiding race (4401)
4392 - revalidate should check working dir is a directory (4134)
4394 - don't always mark "slow" obdfilter messages as errors (4418)
4396 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4399 - don't close LustreDB during write_conf until it is done (3860)
4400 - fix typo in lconf for_each_profile (3821)
4401 - allow dumping logs from multiple threads at one time (3820)
4402 - don't allow multiple threads in OSC recovery (3812)
4403 - fix debug_size parameters (3864)
4404 - fix mds_postrecov to initialize import for llog ctxt (3121)
4405 - replace config semaphore with spinlock (3306)
4406 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4407 - don't allow enqueue to complete on a destroyed export (3822)
4408 - down write_lock before checking llog header bitmap (3825)
4409 - recover from lock replay timeout (3764)
4410 - up llog sem before sending rpc (3652)
4411 - reduce ns lock hold times when setting kms (3267)
4412 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4413 - fix NULL deref and obd_dev leak on setup error (3312)
4414 - replace some LBUG about llog ops with error handling (3841)
4415 - don't match INVALID dentries from d_lookup and spin (3784)
4416 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4417 - fix invalid assertion in ptlrpc_set_wait (3880)
4419 - add libwrap support for the TCP acceptor (3996)
4420 - add /proc/sys/portals/routes for non-root route listing (3994)
4421 - allow setting MDS UUID in .xml (2580)
4422 - print the stack of a process that LBUGs (4228)
4424 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4427 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4428 - make sure to unset replay flag from failed open requests (3440)
4429 - if default stripe count is 0, use OST count for inode size (3636)
4430 - update parent mtime/ctime on client for create/unlink (2611)
4431 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4432 - free recovery state on server during a forced cleanup (3571)
4433 - unregister_reply for resent reqs (3063)
4434 - loop back devices mounting and status check on 2.6 (3563)
4435 - fix resource-creation race that can provoke i_size == 0 (3513)
4436 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4437 - more debugging for page-accounting assertion (3746)
4438 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4439 - avoid deadlock after precreation failure (3758)
4440 - fix race and lock order deadlock in orphan handling (3450, 3750)
4441 - add validity checks when grabbing inodes from l_ast_data (3599)
4443 - add /proc/.../recovery_status to obdfilter (3428)
4444 - lightweight CDEBUG infrastructure, debug daemon (3668)
4445 - change default OSC RPC parameters to be better on small clusters
4446 - turn off OST read cache for files smaller than 32MB
4447 - install man pages and include them in rpms (3100)
4448 - add new init script for (un)mounting lustre filesystems (2593)
4449 - run chkconfig in %post for init scripts (3701)
4450 - drop scimac NAL (unmaintained)
4452 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4455 - clean kiobufs before and after use (3485)
4456 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4457 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4458 - change init script to fail more clearly if not run as root (1528)
4459 - allow clients to reconnect during replay (1742)
4460 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4461 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4462 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4463 - fix cond_resched (3554)
4464 - only evict unfinished clients after recovery (3515)
4465 - allow bulk resend, prevent data loss (3570)
4466 - dynamic ptlrpc request buffer allocation (2102)
4467 - don't allow unlinking open directory if it isn't empty (2904)
4468 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4469 - remove extraneous obd dereference causing LASSERT failure (3334)
4470 - don't use get_cycles() when creating temp. files on the mds (3156)
4471 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4472 - handle EEXIST for set-stripe, set proper directory name (3336)
4474 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4475 - fix ksocknal_fmb_callback() error messages (2918)
4477 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4480 - don't copy lvb into (possibly NULL) reply on error (2983)
4481 - don't deref dentry after dput, don't free lvb on error (2922)
4482 - use the kms to determine writeback rpc length (2947)
4483 - increment oti_logcookies when osc is inactive (2948)
4484 - update client's i_blocks count via lvb messages (2543)
4485 - handle intent open/close of special files properly (1557)
4486 - mount MDS with errors=remount-ro, like obdfilter (2009)
4487 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4488 - don't use cancelling-locks' kms values (2947)
4489 - use highest lock extent for kms, not last one (2925)
4490 - don't dereference ERR_PTR() dentry in error handling path (3107)
4491 - fix thread race in portals_debug_dumplog() (3122)
4492 - create lprocfs device entries at setup instead of at attach (1519)
4493 - common AST error handler, don't evict client on completion race (3145)
4494 - zero nameidata in detach_mnt in 2.6 (3118)
4495 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4496 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4497 - initialize RPC timeout timer earlier for 2.6 (3219)
4498 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4499 - print nal/nid for unknown nid (3258)
4500 - additional checks for oscc recovery before doing precreate (3284)
4501 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4502 - don't crash in mdc_close for bad permissions on open (3285)
4503 - zero i_rdev for non-device files (3147)
4504 - clear page->private before handing to FS, better assertion (3119)
4505 - tune the read pipeline (3236)
4506 - fix incorrect decref of invalidated dentry (2350)
4507 - provide read-ahead stats and refine rpc in flight stats (3328)
4508 - don't hold journal transaction open across create RPC (3313)
4509 - update atime on MDS at close time (3265)
4510 - close LDAP connection when recovering to avoid server load (3315)
4511 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4512 - don't leak open file on MDS after open resend (3325)
4513 - serialize filter_precreate and filter_destroy_precreated (3329)
4514 - loop device shouldn't call sync_dev() for nul device (3092)
4515 - clear page cache after eviction (2766)
4516 - resynchronize MDS->OST in background (2824)
4517 - refuse to mount the same filesystem twice on same mountpoint (3394)
4518 - allow llmount to create routes for mounting behind routers (3320)
4519 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4520 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4521 - fix rare problem with rename on htree directories (3417)
4523 - allow default OST striping configuration per directory (1414)
4524 - fix compilation for qswnal for 2.6 kernels (3125)
4525 - increase maximum number of MDS request buffers for large systems
4526 - change liblustreapi to be useful for external progs like lfsck (3098)
4527 - increase local configuration timeout for slow disks (3353)
4528 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4530 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4533 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4534 - don't overwrite extent policy data in reply if lock was blocked (2901)
4535 - drop filter export grants atomically with removal from device (2663)
4536 - del obd_self_export from work_list in class_disconnect_exports (2908)
4537 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4538 - swab reply message in mdc_close, other PPC fixes (2464)
4539 - fix destroying of named logs (2325)
4540 - overwrite old logs when running lconf --write_conf (2264)
4541 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4542 - fix race in target_handle_connect (2898)
4543 - mds_reint_create() should take same inode create lock (2926)
4544 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4545 - don't close files for self_export to avoid uninitialized obd (2936)
4546 - allow MDS with the same name as client node (2939)
4547 - hold dentry reference for closed log files for unlink (2325)
4548 - reserve space for all logs during transactions (2059)
4549 - don't evict page beyond end of stripe extent (2925)
4550 - don't oops on a deleted current working directory (2399)
4551 - handle hard links to targets without a parent properly (2517)
4552 - don't dereference NULL lock when racing during eviction (2867)
4553 - don't grow lock extents when lots of conflicting locks (2919)
4555 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4558 - account for cache space usage on clients to avoid data loss (974)
4559 - lfsck support in lustre kernel code (2349)
4560 - reduce journal credits needed for BRW writes (2370)
4561 - orphan handling to avoid losing space on client/server crashes
4562 - ptlrpcd can be blocked, stopping ALL progress (2477)
4563 - use lock value blocks to assist in proper KMS, faster stat (1021)
4564 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4565 - recovery for initial connections (2355)
4566 - fixes for mds_cleanup_orphans (1934)
4567 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4568 - block all file creations until orphan recovery completes (1901)
4569 - client remove rq_connection from request struct (2423)
4570 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4571 - recovery timer race (2670)
4572 - mdc_close recovey bug (2532)
4573 - ptlrpc cleanup bug (2710)
4574 - mds timeout on local locks (2588)
4575 - namespace lock held during RPCs (2431)
4576 - handle interrupted sync write properly (2503)
4577 - don't try to handle a message that hasn't been replied to (2699)
4578 - client assert failure during cleanup after abort recovery (2701)
4579 - leak mdc device after failed mount (2712)
4580 - ptlrpc_check_set allows timedout requests to complete (2714)
4581 - wait for inflight reqs when ptlrpcd finishes (2710)
4582 - make sure unregistered services are removed from the srv_list
4583 - reset bulk XID's when resending them (caught by 1138 test)
4584 - unregister_bulk after timeout
4585 - fix lconf error (2694)
4586 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4587 - readahead locks pages, leaves pending causing memory pressure (2673)
4588 - increase OST request buffers to 4096 on large machines (2729)
4589 - fix up permission of existing directories in simple_mkdir (2661)
4590 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4591 - don't assign transno to errored transactions (2742)
4592 - don't delete objects on OST if given a bogus objid from MDS (2751)
4593 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4594 - if rq_no_resend, then timeout request after recovery (2432)
4595 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4596 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4597 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4598 - MDS recovery shouldn't delete objects that it has given out (2730)
4599 - if enqueue arrives after completion, don't clobber LVB (2819)
4600 - don't unlock pages twice when trigger_group_io returns error (2814)
4601 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4602 - don't write pages to disk if there was an error (1450)
4603 - don't ping imports that have recovery disabled (2676)
4604 - take buffered bytes into account when balancing socknal conn (2817)
4605 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4606 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4607 - remove little-endian swabbing of llog records (1987)
4608 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4609 - retry reposting request buffers if they fail (1191)
4610 - grow extent at grant time to avoid granting a revoked lock (2809)
4611 - lock revoke doesn't evict page if covered by a second lock (2765)
4612 - disable VM readahead to avoid reading outside lock extents (2805)
4614 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4615 - updated kernel patches for hp-2.4.20 kernel (2681)
4617 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4620 - fix truncated write corruption (2366)
4621 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4623 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4624 - ensure len(uuid) < 37 in lmc (1171)
4625 - fix ia64 OOPS in llog_test (2255)
4626 - zero end of page at obdfilter for partial page writes (2648)
4627 - don't leave stale dentries around after renames (bug 2428)
4628 - fix timeouts when evicting a client with a single lock held (2642)
4629 - set deadline for the initial HELLO message to drain (2634)
4630 - print out dotted-quad IP addresses in the socknal (2302)
4632 - additional debugging for MDS client eviction problem (2443)
4633 - fix mkfsoptions support for osts (2603, 2604)
4635 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4638 - add series for the vanilla 2.6.0 kernel
4639 - add series for the vanilla 2.4.24 kernel
4640 - add series for a cray x86/64 UL kernel drop
4641 - fix xattr patches for the vanilla 2.4.19 series
4643 - generate true UUIDs in lmc (1171)
4644 - have portals stack dumping break in UML (2466)
4645 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4646 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4647 - wake callback waiting threads on client eviction (2460)
4648 - Add --ptldebug and --subsystem to lmc (1719)
4649 - update assertion to allow safe interrupt allocation
4650 - set rq_no_resend for cancel requests (2432)
4651 - recalculate ptlrpcd timeout after resend (2494)
4652 - call vfs_rmdir when removing pending directories (2368)
4653 - fix renaming a file to itself (2429)
4654 - lmc creates a default one-stripe lov (2454)
4655 - expand procfs space to handle large clusters (2326)
4656 - increase UML stack to avoid overflow
4657 - update lconf's list of debug and subsystem masks
4658 - fix lfs find --obd (2510)
4659 - /proc tunable for disabling filter read caching (2591)
4660 - stop rpm packages from altering slapd.conf (2301)
4661 - disable nagle in the socknal under 0conf (2578)
4662 - choose mds inode size based on stripe count (2572)
4663 - fix kernel-source rpm problems (2516)
4665 - add --disable-doc to avoid pdf generation (2421)
4666 - update documentation, tests, type-os, comments
4667 - avoid format warnings on ia64
4668 - remove the TOE NAL
4669 - tiny code cleanups by removing unused fields
4671 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4674 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4675 - remove the most blatant lies from BUILDING file (2371)
4676 - change default debug level to reasonable production setting
4677 - reduce client side cache size to reduce cache flush time
4678 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4679 - make TCP zerocopy and pinger support enabled by default (2476)
4680 - sync writes completed after process exits caused crashes (2319)
4681 - maintain correct mount count on the MDS (2356)
4682 - backout 1557, because 2316 wasn't really fixed
4683 - better file I/O statistics gathering in /proc
4684 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4685 - another kernel patch to fix zero-copy TCP function export
4686 - don't take duplicate lock when processing re-sent getattr (2420)
4687 - lctl uses obd_self_export instead of creating new conn (2353)
4688 - MDS/OST recovery case which requires object creation asserted (2425)
4689 - move lfs from /usr/sbin to /usr/bin in packages
4690 - fix race between mds_client_add and mds_client_free (2417)
4691 - use kmalloc instead of slabs in portals (2430)
4692 - don't create duplicate records when a failover MDS is present (2442)
4693 - remove unnecessary mount age check (2332)
4694 - don't remove directory inodes from locks prematurely (2451)
4695 - don't break if MDS service name is the same as hostname (2103)
4696 - fix races in client write RPC generation when cache full (2482)
4698 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4701 - remove now-unused request->rq_obd (278)
4702 - if an allocation fails, print out how much memory we've used (1933)
4703 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4704 - add the "configurable stack size" patch to most series files (1256)
4705 - ability to write large log records, for 100+ OST configs (2306)
4706 - fix NULL deref when filter_prep fails (2314)
4707 - fix operator precedence error in filter_sync
4708 - dynamic allocation of socknal TX descriptors (2315)
4709 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4710 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4711 - ensure that CWARN messages in Portals always get to the syslog
4712 - __init/__exit are not for prototype decls (ldlm_init/exit)
4713 - x86-64 compile warning fixes
4714 - fix gateway LMC keyword conflict (2318)
4715 - fix MDS lock inversions in getattr/reint paths (1844)
4716 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4717 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4718 - DLM race condition prevented some lock evictions (2328)
4719 - ENOMEM detection and retry on socknal sends (2230)
4720 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4721 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4722 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4723 - better error messages when a client is rejected during recovery (1505)
4724 - avoid cancelling locks which were never granted, after failure (2330)
4725 - fix i_sem/journal inversion in mds_client_add (2333)
4726 - fix truncate/getattr lock cycle deadlock (2334)
4727 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4728 - fix two rare exit paths which could leak an l_lock() ref (2321)
4729 - fencepost error in MDS/OST orphan recovery (2226)
4730 - make log record alignment 8 bytes (1988)
4731 - lstripe now fails when requested offset > ost_count (2237)
4732 - ensure that all kernel series have a complete list.h (1607)
4733 - fix crashes in special-file operations (2316)
4734 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4736 - allow configurable automake binary, for testing new versions
4737 - small update to the lfs documentation
4739 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4741 * fix negative export reference count in fsfilt_sync (2312)
4743 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4744 * release candidate 0.9.1
4746 - orphans are moved into the PENDING directory for possible recovery
4747 - replayed opens now open by fid for orphan/rename safety (1042)
4748 - last close of an orphan inode generates a transno (683)
4749 - chdir() and mount() now pin the directory entry (1020)
4750 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4751 - discard very old requests without processing them (1502)
4752 - remove some common, well-understood CERRORs (1505)
4753 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4754 - clear "grant" flags in OST replies until OST grant code lands (1644)
4755 - fix read performance by not clobbering i_blksize on client (1598)
4756 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4757 - make LRU size a /proc tunable, clears locks when reduced (707)
4758 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4759 - server locks take references on exports now (1558)
4760 - build fixes for 2.4.20-rh trees (1663)
4761 - return an error from lov_create if all OSCs are inactive (1751)
4762 - fix import levels when a reconnect happens without a timeout (1597)
4763 - exit early from mds_open if we get a lookup error (1749)
4764 - partial page read at EOF wouldn't wait for disk before sending (1642)
4765 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4766 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4767 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4768 - write/truncate lock inversion (1639)
4769 - Don't auto-load obdclass, portals modules during cleanup (1495)
4770 - fix timestamps from jumping to "now" (1763)
4771 - extra journal assertions (1648)
4772 - add an extra multiunlink test (1771)
4773 - fix read_record/write_record API (1776)
4774 - fix leak of offset_extent, possible incorrect i_size later (1772)
4775 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4776 - Debugging for the kqswnal_get_idle_tx problems (1820)
4777 - Allow recovery to be attempted multiple times (1536)
4778 - Write out MDS last_rcvd file after it is first created (1600)
4779 - Fix tx_descriptor leak in failed transmit situations (1827)
4780 - ext3 journaling fixes for assertion failure after IO error (1871)
4781 - class_export_put() on freed export after completion AST error (1896)
4782 - Fix revalidate looping in VFS (1322)
4783 - Don't access a freed export during MDS_REINT timeout (1521)
4784 - Add open-unlink recovery support on the MDS (1673,1764)
4785 - Return an error if no MDS data was read from last_rcvd (1946)
4786 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4787 - Don't setup a disk device that doesn't match exported UUID (317)
4788 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4789 - avoid committing NULL handle in force close
4790 - local.sh is now a one-stripe LOV configuration
4791 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4792 - fix ext3 htree duplicate directory entry corruption (1516)
4793 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4794 - update to new LOV EA format (2097)
4795 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4797 2003-06-15 Phil Schwan <phil@clusterfs.com>
4800 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4801 - per-import/export recovery handling (958, 931, 959)
4802 - multiple last-rcvd slots, for serving multiple FSes (949)
4803 - connections are again shared between multiple imp/exports (963, 964)
4804 - "umount -f" would hang if any requests needed to be sent (393, 978)
4805 - avoid pinning large req buffer by copying for queued messages (989)
4806 - add "uuid" to "lctl device" command to help upcalls (991)
4807 - "open" RPCs with transnos would confuse recovery counters (1037)
4808 - do proper endian conversion of all wire messages (288, 340, 891)
4809 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4810 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4811 - fix ptlrpc_connection leak in target_handle_connect (1174)
4812 - fix import refcounting bug in OST and MDS cleanup (1134)
4813 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4814 - fix very unlikely obd_types race condition (501)
4815 - remove osc_open hack for echo_client (1187)
4816 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4817 - a failure in read_inode2 leads to deadlock (1139)
4818 - cancel ack-locks as soon as transaction is committed (1072)
4819 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4820 - make sure to commitrw after any preprw to avoid deadlock (1162)
4821 - failing to execute a file in a lustre FS would lock inode (1203)
4822 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4823 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4824 - obd_uuid2tgt would walk off the end of the list (1255)
4825 - on IA64 the osc would give portals incorrect bulk size (1258)
4826 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4827 - fix lock inversion caused by new llite matching code (1282)
4828 - limit the number of dirty pages on a client to 10MB (1286)
4829 - timed out locks were not being corrected cancelled (1289)
4830 - fix O_DIRECT above 4GB on IA-32 (1292)
4831 * major user-visible changes
4832 - fail out/fail over policy now controlled by the upcall (993)
4834 - add OBD_PING to check server availability and failure (954)
4835 - lustre messages are now sent in sending host order (288, 340, 891)
4836 - add eadatalen to MDS getattr reply (340)
4837 - OST read replies may contain second buffer, with per-page status (593)
4839 2003-03-11 Phil Schwan <phil@clusterfs.com>
4842 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4843 - failed open()s could cause deadlock; fixed (867, 869)
4844 - stop cancelling OST locks when files are closed (481)
4845 - overlapping XID spaces caused network corruption (851, 853)
4846 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4847 - setattr_raw added, to do single-RPC, server-side setattrs
4848 - lmc/lconf syntax change for OST UUIDs
4849 - fix crashy race condition between ptlrpc_free_req and osc_close
4850 - don't use request in mdc_enqueue if we hit a timeout (889)
4851 - don't set the inode i_size for regular files from the MDS (896)
4852 - handle out of order completion AST (842)
4853 - don't LBUG if a lock request times out after receiving AST (913)
4854 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4855 - if a bad lock AST arrives, send an error instead of dropping entirely
4856 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4857 - fix leak in bulk IO when only partially completed (899, 900, 926)
4858 - fix O_DIRECT for ia64 (55)
4859 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4860 - C-z after timeout could hang a process forever; fixed (977)
4862 - client-side I/O cache (678, 924, 929, 941, 970)
4864 - READPAGE and SETATTRs which don't take server-side locks get
4867 2003-02-11 Phil Schwan <phil@clusterfs.com>
4870 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4871 - Fix fsfilt_extN_readpage() to read a full page of directory
4872 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4873 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4874 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4875 - Most intent operations were removed, in favour of a new RPC mode
4876 that does a single RPC to the server and bypasses most of the VFS
4877 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4878 - Aggressively cancel local locks on DLM servers
4879 - mds_reint_unlink sends EA to the client if it's the last nlink.
4880 client uses that EA to unlink OST objects.
4881 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4882 - recursive symlinks were fixed (439)
4883 - fixed NULL deref in DEBUG_REQ
4884 - filter_update_lastobjid no longer calls sync, which annoyed extN
4885 - fixed multi-client small-writes to a single file problem (445)
4886 - fixed mtime updates during file writes (607)
4887 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4888 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4889 - fixed Linux OST size reporting problem (444, 656)
4890 - OST now updates object mtime with writes or setattr (607, 619)
4891 - client verifies file size before zeroing page past EOF (445)
4892 - OST now writes last allocated objid to disk with allocation (108)
4893 - LOV on echo now works (409)
4895 - mds_reint_unlink sends a new buffer, with the EA included. this
4896 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4897 set if a regular file was being unlinked, and it was the last link
4898 - use PtlGet from the target for bulk writes (315)
4899 - OST now updates object mtime with writes or setattr (607, 619)
4900 - LDLM now has a grant-time callback to revalidate locked items, if
4902 - Many MDS operations were reorganized to combat race conditions
4904 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4905 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4907 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4910 - Fully reactivate OST imports after reconnection (512, others)
4911 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4912 - More graceful error handling for truncating on dead OST (515)
4913 - Don't error out unless we're actually accessing dead stripes (474)
4914 - Fix garbage sizes when stripes are missing (410)
4915 - LRU counters were broken, causing constant lock purge (433, 432)
4916 - garbage on read from stripes with failed OSTs (441)
4917 - mark OSCs as active before reconnecting during recovery (438)
4918 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4919 - lfind did not preserve OST order in output (443)
4920 - symlinks cause hung clients, incorrect data (439)
4921 - stop dereferencing request after dropping refcount (457)
4922 - don't LASSERT(spin_is_locked) on non-SMP (455)
4923 - fixes for many rename() bugs
4924 - fstat didn't correctly synchronize attributes (399)
4925 - server must handle lock cancellation during blocking AST prep (487)
4926 - bulk descriptors were free()d too soon (511)
4927 - fix paths in lconf, which would load incorrect modules (451, 507)
4928 - fix confusing lconf 'host not found' error message (386)
4929 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4930 - fix race condition in mdc_blocking_ast() for inode access (526)
4931 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4932 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4933 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4934 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4935 - fix lstripe in several ways (kernel side) (527)
4936 - fix request leak in ldlm_cli_enqueue (262)
4937 - incorrect OSC was marked inactive after OST failure
4938 - call mds_fs_cleanup before unmounting filesystem (524)
4939 - fix races between taking ns_lock and ldlm_lock_change_resource
4940 - fix races updating LOV export open file list
4941 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4942 - fix recovery NULL deref in ldlm_cli_cancel_unused
4943 - fix some DLM races by using new hash table for lock handles (419)
4944 - permit the client to specify desired inodes, at replay
4945 - duplicate requests when we queue them for replay reintegration
4946 - fix last_rcvd offset calculation
4947 - sync after each recovered transaction, so we always make progress
4948 - never, not always, ERESTART requests without transnos
4949 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4951 - skip replay if the MDS says that the client is already connected
4952 - don't check for a recovery-enabled export to match lctl's UUID
4953 - don't INC_USE_COUNT for phantom exports
4954 - don't crash when cleaning up phantom exports (567)
4955 - don't double-finish or set replay data for errored mdc_open requests
4956 - abort requests when they time out, so we don't get old replies
4957 - send/receive replies for AST messages again
4958 - if the client says that it doesn't have the lock, cancel it on the
4960 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4961 mark it as destroyed, it will all work out when decref is called
4962 - fix module use counts (22, 581)
4964 - ASTs now expect a reply (server cancels lock on error reply)
4966 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4969 - fix many simultaneous client startup (392)
4970 - fix dentry->d_it clobbering
4971 - credentials weren't being shipped for readdir/getattr operations
4972 - remove invalid assertions triggered during some concurrent MD
4974 - proper Lustre versions added (336, 389)
4975 - fix memory leak for create error case (398)
4976 - fix LOV locking bug that would get cli/srv out of sync
4977 - fix echo client over LOV (409)
4978 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4979 - fix double-O_EXCL intent crash (424)
4980 - avoid sending multiple lock CANCELs (352)
4982 - MDS can do multi-client recovery (modulo bugs in new code)
4984 - many updates, edits, cleanups
4986 2002-11-18 Phil Schwan <phil@clusterfs.com>
4989 - fix null d_it dereference (346)
4990 - fix full OST/dbench hang (333)
4991 - fix permission problem with file removal (286)
4992 - fix removal of OSCs from LOV when they fail
4993 - fix NULL deref during bulk timeout (214)
4994 - fix problems related to multiple filesystems on one MDS (241)
4995 - fixed serious subtle metadata locking bugs
4996 - free locks on clients when inodes are removed due to memory
4998 - fix inode pointer in lock data (285)
4999 - partial support for multiple MDS on a single host (241)
5000 - data locks weren't cancelled at clear_inode time (290, 311)
5001 - intent locks could lead to unbounded lock growth (205)
5002 - added a maximum lock count, an LRU list, and a flusher
5003 - fix multiple rename (365)
5004 - properly abstracted the echo client
5005 - OSC locked 1 byte too many; fixed
5006 - rewrote brw callback code:
5007 - fixed recovery bugs related to LOVs (306)
5008 - fixed too-many-pages-in-one-write crash (191)
5009 - fixed (again) crash in sync_io_timeout (214)
5010 - probably fixed callback-related race (385)
5012 - Add capability to MDS protocol
5013 - LDLM cancellations and callbacks on different portals
5015 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5018 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5019 - defer open object destruction to close time (601981, 138)
5020 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5021 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5022 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5023 - don't call obd_setattr in ll_file_release for destroyed objects
5024 * protocol change to lustre_msg: move |version| and add |flags|
5025 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5026 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5027 * added replay of create, unlink, link and rename operations during
5028 MDS failover; recovery should be much more robust now
5029 * remove failed OSCs from LOVs (only lov_create uses this so far)
5030 * the lustre-HOWTO was brought (more) up to date (582544)
5032 2002-10-23 Phil Schwan <phil@clusterfs.com>
5035 - in-use dentries weren't being reused properly (617851)
5036 - prevent multiple LDLM setup (599178)
5037 - fix LOV size calculations for truncate (617853)
5038 - fix client handling of MDS intent errors (POSIX)
5039 - fix permission bug in lovstripe.c test (624321)
5040 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5041 - truncate past end of file could corrupt data
5042 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5043 - a race in recovery could return ETIMEDOUT to apps (623947)
5044 - building outside the source directory was fixed
5045 * the lustre-HOWTO was brought (more) up to date (582544)
5046 * major progress was made on recovery functionality
5048 2002-10-10 Phil Schwan <phil@clusterfs.com>
5051 - recovery deadlock fix
5052 - rm -rf causes LBUG fix (617817)
5053 - file open by multiple tasks fix (618962)
5054 - directory permissions bugs (602707 and 620007)
5055 - journal_stop fixed with locking (611313)
5056 - O_APPEND failures resolved (618273, perhaps 614459)
5057 - lconf PATH fix (619770)
5058 - IA64 build fix (621450)
5059 - RPC buffer sizes scale with amount of memory
5061 2002-10-01 Phil Schwan <phil@clusterfs.com>
5064 - locks would be cancelled without throwing away data pages,
5065 resulting in inconsistent data (605627)
5066 - inode attributes were not always being refreshed (605627, 612449)
5067 - lconf now continues to cleanup after lctl reports an error
5068 - MDS now enforces user permissions (602707)
5069 - lprocfs cleanup fixed, but not yet enabled (614157)
5070 - fixed infinite server hang, should a client not respond to an AST
5071 - avoid going into recovery if user calls readlink() with a buffer
5072 that's too small (613941)
5073 - AST RPCs no longer require replies (614867) -- this may be changed
5074 - don't crash server if client sends an IOV that's too big (611336)
5075 - fixed lock conversion deadlock (611892)
5076 - fixed the following of symlinks (614622)
5077 * recovery: the server can remove locks from a client that dies, other
5078 clients can make progress
5079 * more extN patch fixes
5080 * compile-time configurable ptlrpc buffer allocations
5082 - collaborative read cache document
5083 - Lustre Lite Performance CDR document-in-progress
5085 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5088 - fix typo in patch-2.4.18
5090 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5093 - clear ptlrpc request each time in handle_incoming_request()
5094 - unlink of files now destroys the object on the OST
5096 2002-09-19 Peter Braam <braam@clusterfs.com>
5098 * add hard link support
5099 * change obdfile creation method
5100 * kernel patch changed
5102 2002-09-19 Peter Braam <braam@clusterfs.com>
5105 - stack overflow bug in extN fixed
5107 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5109 * documentation updates
5110 - add man pages for config tools
5111 - update tests/README to describe testing with new config tools
5112 - finish metadata API descriptions
5113 * bug fixes and cleanups
5114 - statfs workaround for 16TB limit
5115 - LOV stripe allocation improved, can stripe on subset of OSTs
5116 - LOV file size/IO offset was wrong for files > 4GB in size
5117 - object EA data was being dropped, caused files to be unreadable
5118 - memory overflow with non-LOV OST caused memory corruption
5119 - fixed regression tests to work with new config tools, obdfilter
5120 - fixed bug when directory size became larger than 1 block
5121 - fixed bug (for single client case) when PWD was deleted
5122 - invalidate local directory pages when doing intent-based ops
5123 - avoid LDLM oops when lock callback contained bad data
5125 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5127 * documentation updates
5128 * bug fixes and cleanups
5129 - configuration tools
5132 - 64-bit compile warnings
5133 - 64-bit internal statfs data
5135 * test_brw on persistent OST devices
5137 * lprocfs (disabled)
5139 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5141 * documentation updates
5142 * bug fixes and cleanups
5143 * configuration tools
5145 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5148 * many small fixes to 0_5_4
5149 * io/network handling
5150 * thinkos in MDS operations
5152 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5155 * crucial basic fixes to 0.5.3
5156 * IOR, Iozone work over Elan
5159 2002-08-07 Phil Schwan <phil@clusterfs.com>
5160 * version 0_5_3, our first alpha
5161 * we use the new Portals iovs
5162 * documentation updates
5163 * bug fixes and cleanups
5164 * small changes in the DLM wire protocol
5166 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5167 * version 0_5_1 with some initial stability,
5168 * locking on MD and file I/O.
5169 * documentation updates
5170 * several bug fixes since 0.5.0
5171 * small changes in wire protocol
5173 2002-07-18 Phil Schwan <phil@clusterfs.com>
5175 * delivered as Lustre Light Alpha
5176 * fixed a crash after handling invalid MDS requests
5177 * fixed directory pages for architectures with non-4k pages sizes
5179 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5180 * release version v0_4_4
5181 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5183 * Fixes a number of interruption problems with OST operations.
5184 * Update documentation for portals header changes
5185 * Move all wire protocol structs/defines to lustre_idl.h
5186 * Fixes symlink length bug.
5187 * Add tcpdump to repository.
5189 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5190 * release version v0_4_3
5191 * Fixes statfs for inodes on extN.
5192 * Fixes bug in runtests which would delete /etc/hosts.
5193 * Use 64-bit object IDs wherever possible (not into VFS though)
5194 Remove ost_get_info, which is unused by lustre, and out of date.
5196 2002-07-03 Peter Braam <braam@clusterfs.com>
5197 * release version v0_4_2 Fixes a lookup error (type not passed)
5198 * move forward to head of Portals
5199 * move forward to latest Lustre kernel
5201 2002-06-25 Peter Braam <braam@clusterfs.com>
5202 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob