1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.31 (SLES 10),
5 2.6.18-92.1.17.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: compact fld format with extents
19 Details : Store range of seq rather than every seq in FLD. Seq
20 controller update FLD rather than clients. In Case of CMD, mdt0
21 has FLD, all other metadata server act as non persistent proxy
22 for FLD queries and cache fld entries in fld cache.
27 Description: don't skip ost target if they assigned to file
28 Details : Drop slow OSCs if we can, but not for requested start idx.
29 This means "if OSC is slow and it is not the requested
30 start OST, then it can be skipped, otherwise skip it only
31 if it is inactive/recovering/out-of-space.
35 Description: more cleanup in mds_lov
36 Details : not send LOV EA under replay, we can't know about they size at this
37 time. Don't allow client connect to mds before any ost connected,
38 for avoid problems with LOV EA size and returning EIO to client.
40 Severity : enhancement
42 Description: Interoperability at server side (Disk interoperability)
44 Severity : enhancement
46 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
48 Severity : enhancement
50 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
52 Severity : enhancement
54 Description: New client IO stack (CLIO).
56 Severity : enhancement
58 Description: Commit on sharing. Eliminate inter-client dependencies between
59 uncommitted transactions by doing transaction commits.
60 Thereby clients may recovery independently.
63 Frequency : Create a symlink file with a very long name
65 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
66 Details : If there is no extra space in the request for early cancels,
67 ldlm_req_handles_avail() returns 0 instead of a negative value.
69 Severity : enhancement
71 Description: Add /proc entry for import status
72 Details : The mdc, osc, and mgc import directories now have
73 an import directory that contains useful import data for debugging
76 Severity : enhancement
78 Description: Re-disable certain /proc logging
79 Details : Enable and disable client's offset_stats, extents_stats and
80 extents_stats_per_process stats logging on the fly.
83 Frequency : Only on FC kernels 2.6.22+
85 Description: oops in statahead
86 Details : Do not drop reference count for the dentry from VFS when lookup,
87 VFS will do that by itself.
89 Severity : enhancement
91 Description: Generic /proc file permissions
92 Details : Set /Proc file permissions in a more generic way to enable non-
93 root users operate on some /proc files.
97 Description: Hitting mdc_commit_close() ASSERTION
98 Details : Properly handle request reference release in
99 ll_release_openhandle().
103 Description: quota recovery deadlock during mds failover
104 Details : This patch includes att18982, att18236, att18237 in bz14840.
106 1. fix osts hang when mds does failover with quotaon
107 2. prevent watchdog storm when osts threads wait for the
112 Frequency : only patchless client
113 Description: add workaround for race between add/remove dentry from hash
115 Severity : enhancement
117 Description: Allow OST glimpses to return PW locks
121 Description: LBUG when llog conf file is full
122 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
126 Description: Prevent import from entering FULL state when server in recovery
130 Description: service mount cannot take device name with ":"
131 Details : Only when device name contains ":/" will mount treat it as
137 Description: replace ptlrpcd with the statahead thread to interpret the async
138 statahead RPC callback
142 Frequency : on recovery
143 Description: I/O failures after umount during fail back
144 Details : if client reconnected to restarted server we need join to recovery
145 instead of find server handler is changed and process self
146 eviction with cancel all locks.
148 Severity : enhancement
150 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
152 Severity : enhancement
154 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
156 Severity : enhancement
158 Description: Upcall on Lustre log has been dumped
159 Details : Allow for a user mode script to be called once a Lustre log has
160 been dumped. It passes the filename of the dumped log to the
161 script, the location of the script can be specified via
162 /proc/sys/lnet/debug_log_upcall.
167 Description: avoid idr_remove called for id which is not allocated.
168 Details : Move assigment s_dev for clustered nfs to end of initialization,
169 to avoid problem with error handling.
174 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
175 Details : When connection is reused this not moved from CONN_UNUSED_HASH
176 into CONN_USED_HASH and this prodice warning when put connection
177 again in unused hash.
180 Severity : enhancement
182 Description: File striping can now be set to use an arbitrary pool of OSTs.
184 Severity : enhancement
186 Description: Export bytes_read/bytes_write count on OSC/OST.
190 Description: Early reply size mismatch, MGC loses connection
191 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
192 the connect flags are properly negotiated.
197 Description: quotas are not honored with O_DIRECT
198 Details : all writes with the flag O_DIRECT will use grants which leads to
199 this problem. Now using OBD_BRW_SYNC to guard this.
203 Description: add quota statistics
204 Details : 1. sort out quota proc entries and proc code.
205 2. add quota statistics
207 Severity : enhancement
209 Description: enable quota support for HEAD.
213 Description: Properly propagate oinfo flags from lov to osc for statfs
214 Details : restore missing copy oi_flags to lov requests.
216 Severity : enhancement
218 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
219 : l_getgroups(8), lst(8), routerstat(8)
220 Details : included man pages for llobdstat(8), llstat(8),
221 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
223 Severity : enhancement
225 Description: configure's --enable-quota should check the
226 : kernel .config for CONFIG_QUOTA
227 Details : configure is terminated if --enable-quota is passed but
228 : no quota support is in kernel
232 Description: Remove portals compatibility
233 Details : Remove portals compatibility, not interoperable with releases
238 Description: Resolve device initialization race
239 Details : Prevent proc handler from accessing devices added to the
240 obd_devs array but yet be intialized.
242 Severity : enhancement
244 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
246 Severity : enhancement
248 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
253 Description: Using wrong pointer in osc_brw_prep_request
254 Details : Access to array[-1] can produce panic if kernel compiled with
255 CONFIG_PAGE_ALLOC enabled
259 Description: Client runs out of low memory
260 Details : Consider only lowmem when counting initial number of llap pages
264 Description: *optional* service tags registration
265 Details : if the "service tags" package is installed on a Lustre node
266 When the filesystem is mounted, a local-node service tag will
267 be created. See http://inventory.sun.com/ for more information
268 about the Service Tags asset management system.
272 Description: Kernel BUG tries to release flock
273 Details : Lustre does not destroy flock lock before last reference goes
274 away. So always drop flock locks when client is evicted and
275 perform unlock regardless of successfulness of speaking to MDS.
279 Description: add refcount protection for osc callbacks, avoid panic on shutdown
283 Description: sanity test 65a fails if stripecount of -1 is set
284 Details : handle -1 striping on filesystem in ll_dirstripe_verify
289 Description: ASSERTION(CheckWriteback(page,cmd)) failed
290 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
293 Severity : enhancement
295 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
299 Description: do not process already freed flock
300 Details : flock can possibly be freed by another thread before it reaches
301 to ldlm_flock_completion_ast.
305 Description: LBUG during stress test
306 Details : Need properly lock accesses the flock deadlock detection list.
310 Description: oops in page fault handler
311 Details : kernel page fault handler can return two special 'pages' in error
312 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
316 Description: timeout with invalidate import.
317 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
318 should be handled by ptlrpcd. This produce long age waiting and
319 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
321 Severity : enhancement
323 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
327 Description: enable MGS and MDT services start separately
328 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
329 a co-located MGS without starting the MGS, which is a complement
330 to 'nosvc' mount option.
334 Frequency : after recovery
335 Description: precreate to many object's after del orphan.
336 Details : del orphan st in oscc last_id == next_id and this triger growing
337 count of precreated objects. Set flag LOW to skip increase count
338 of precreated objects.
342 Frequency : rare, on clear nid stats
343 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
344 Details : when clean nid stats sometimes try destroy live entry,
345 and this produce panic in free.
349 Description: Stack overflow during MDS log replay
350 ease stack pressure by using a thread dealing llog_process.
354 Description: wait until IO finished before start new when do lock cancel.
355 Details : VM protocol want old IO finished before start new, in this case
356 need wait until PG_writeback is cleared until check dirty flag and
357 call writepages in lock cancel callback.
359 Severity : enhancement
361 Description: using special macro for print time and cleanup in includes.
365 Description: mds_mfd_close() ASSERTION(rc == 0)
366 Details : In mds_mfd_close(), we need protect inode's writecount change
367 within its orphan write semaphore to prevent possible races.
371 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
375 Frequency : rare, on shutdown ost
376 Description: don't hit live lock with umount ost.
377 Details : shrink_dcache_parent can be in long loop with destroy dentries,
378 use shrink_dcache_sb instead.
382 Description: don't panic with use echo client
383 Details : echo client pass NULL as client nid pointer and this produce null
388 Description: fix build on ppc32
389 Details : compile code with -m64 flag produce wrong object file for ppc32.
393 Description: add message levels for liblustreapi
397 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
398 Details : Move the 'good_osts' check before the 'total_bavail' check. This
399 will result in an -EAGAIN and in the exit call path we call
400 alloc_rr() which will with increasing aggressiveness attempt to
401 aquire precreated objects on the minimum number of required OSCs.
405 Description: Use old size assignment to avoid deadlock
406 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
407 the scheduling while holding a spinlock. We do not need locking
408 for size in ll_update_inode() because size is only updated from
409 the MDS for directories or files without objects, so there is no
410 other place to do the update, and concurrent access to such inodes
411 are protected by the inode lock.
415 Description: resolve "_IOWR redefined" build error on SLES10
419 Description: dump the memory debugging after all modules are unloaded to
420 suppress false negative in conf_sanity test 39
422 Severity : enhancement
424 Description: build kernel-ib packages for OFED 1.3 in our release cycle
429 Description: fix SLES kernel versioning
430 Details : the kernel version for our SLES 10 kernel did not include a "-"
431 before the "smp" at the end. while this was not a problem in
432 general, it did mean that software trying to use the kernel
433 version to try to detect a vendor specific kernel would fail.
434 this was most evident by the OFED build scripts.
438 Description: Don't update lov_desc members until making sure they are valid
439 Details : When updating lov_desc members via proc fs, need fix their
440 validities before doing the real update.
444 Description: don't put request into delay list while invalidate in flight.
445 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
446 import in flight. this produce timeout for invalidate and sometimes
447 can cause stale data.
451 Frequency : on ppc only
452 Description: not convert ost objects for directory because it's not exist.
453 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
457 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
459 Severity : enhancement
461 Description: Update to support 2.6.22.14 vanilla kernel.
465 Frequency : rare, on recovery
466 Description: read procfs can produce deadlock in some situation
467 Details : Holding lprocfs lock which send rpc can produce block for destroy
468 obd objects and this also block reconnect with -EALREADY. This
469 isn't fix all lprocfs bugs - but make it rare.
471 Severity : enhancement
473 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
476 Frequency : frequent on X2 node
478 Description: mdc_set_open_replay_data LBUG
479 Details : Set replay data for requests that are eligible for replay.
483 Description: lustre_mgs: operation 101 on unconnected MGS
484 Details : When MGC is disconnected from MGS long enough, MGS will evict the
485 MGC, and late on MGC cannot successfully connect to MGS and a lot
486 of the error messages complaining that MGS is not connected.
490 Frequency : on network error
491 Description: panic with double free request if network error
492 Details : mdc_finish_enqueue is finish request if any network error ocuring,
493 but it's true only for synchronus enqueue, for async enqueue
494 (via ptlrpcd) this incorrect and ptlrpcd want finish request
497 Severity : enhancement
499 Description: client-side metadata stat-ahead during readdir(directory readahead)
500 Details : perform client-side metadata stat-ahead when the client detects
501 readdir and sequential stat of dir entries therein
504 Frequency : on start mds
506 Description: Implement get_info(last_id) in obdfilter.
509 Frequency : occasional
511 Description: Correctly check stale fid, not start epoch if ost not support SOM
512 Details : open with flag O_CREATE need set old fid in op_fid3 because
513 op_fid2 was overwritten with new generated fid, but mds can answer
514 with one of these two fids and both is not stale. Setattr
515 incorrectly started an epoch and assume will be called
516 done_writing, but without SOM done_writing ever being called.
519 Frequency : rare, depends on device drivers and load
521 Description: MDS or OSS nodes crash due to stack overflow
522 Details : Code changes in 1.8.0 increased the stack usage of some functions.
523 In some cases, in conjunction with device drivers that use a lot
524 of stack the MDS (or possibly OSS) service threads could overflow
525 the stack. One change which was identified to consume additional
526 stack has been reworked to avoid the extra stack usage.
529 Frequency : occasional
531 Description: Do not fail import if osc_interpret_create gets -EAGAIN
532 Details : If osc_interpret_create got -EAGAIN it immediately exits and
533 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
534 oscc_has_objects and see OSC has no objests and call
535 oscc_internal_create to resend create request.
537 Severity : enhancement
539 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
541 Severity : enhancement
543 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
546 Frequency : very rare
548 Description: Fix replay if there is an un-replied request and open
549 Details : In some cases, older replay request will revert the
550 mcd->mcd_last_xid on MDS which is used to record the client's
553 Severity : enhancement
555 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
557 Severity : enhancement
559 Description: Add rhel5 support to HEAD.
561 Serverity : enhancement
563 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
568 Description: Don't allow skipping OSTs if index has been specified.
569 Details : Don't allow skipping OSTs if index has been specified, make
570 locking in internal create lots better.
574 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
575 Details : access to bitfield in structure is always rounded to long
576 and this produce problem with not atomic change any bit.
580 Description: Lustre make rpms failed.
581 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
582 builds Lustre rpms from tarball.
585 Frequency : rare on shutdown ost
587 Description: If llog cancel was not send before clean_exports phase, this can
588 produce deadlock in llog code.
589 Details : If llog thread has last reference to obd and call class_import_put
590 this produce deadlock because llog_cleanup_commit_master wait when
591 last llog_commit_thread exited, but this never success because was
592 called from llog_commit_thread.
596 Description: allow userland application know is lost one of stripes.
597 Details : fill lvb_blocks with error code on ost and return it to
598 application if error flag found.
602 Description: NULL lov_tgts causing MDS oops
603 Details : more safe checks for NULL lov_tgts for avoid oops.
605 Severity : enhancement
607 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
611 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
612 Details : If an OST is down the MDS will hang indefinitely in
613 obd_statfs_rqset() waiting for the statfs data. While for
614 MDS QOS usage of statfs, it should not stuck in waiting.
616 Severity : enhancement
618 Description: remote_acl support
619 Details : Support ACL-based permission check for remote user.
620 Support setfacl/getfacl for remote user with the utils
621 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
622 the system "{s,g}etfacl" utils.
624 Severity : enhancement
626 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
628 Severity : enhancement
630 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
634 Description: make mgs_setparam() handle fsname containing dash
635 Details : fsname containing a dash does not work with lctl conf_param
637 Severity : enhancement
639 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
641 Severity : enhancement
643 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
647 Frequency : rare, at shutdown
648 Description: access already free / zero obd_namespace.
649 Details : if client_disconnect_export was called without force flag set,
650 and exist connect request in flight, this can produce access to
651 NULL pointer (or already free pointer) when connect_interpret
652 store ocd flags in obd_namespace.
656 Frequency : only at startup
657 Description: not alloc memory with spinlock held.
658 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
659 if any spinlock held.
661 Severity : enhancement
663 Description: make lustre randomly fail allocating memory
664 Details : Make lustre randomly failed allocating memory for testing purpose.
666 Severity : enhancement
668 Description: lost problems with lov objid file
669 Details : Fixes some scability and access to not inited memory problems
670 in work with lov objdid file.
675 Description: lfs find does not continue on file error
676 Details : Continue other files processing when a file/dir is absent.
680 Description: Inconsistent usage of lustre_pack_reply()
681 Details : Standardize the usage of lustre_pack_reply() such that it
682 always generate a CERROR on failure.
685 Frequency : occasional
687 Description: MDS hang or stay in waiting lock
688 Details : If client receive lock with CBPENDING flag ldlm need send lock
689 cancel as separate rpc, to avoid situation when cancel request
690 can't processed due all i/o threads stay in wait lock.
694 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
698 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
702 Description: Update to SLES9 kernel-2.6.5-7.287.3.
706 Description: Update to RHEL4 latest kernel.
708 Severity : enhancement
710 Description: Build SLES10 patchless client fails
711 Details : The configure was broken by run ./configure with
712 --with-linux-obj=.... argument for patchless client. When the
713 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
714 can't search header adequately. Use absolute path such as
715 -I($LINUX)/include instead.
719 Description: interrupt oig_wait produce painc on resend.
720 Details : brw_redo_request can be used for resend requests from ptlrpcd and
721 private set, and this produce situation when rq_ptlrpcd_data not
722 copyed to new allocated request and triggered LBUG on assert
723 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
724 ptlrpcd set if request is changed and can be safety changed to use
729 Description: LASSERT_{REQ,REP}SWAB macros are buggy
730 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
731 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
732 these macros and replace them with their difinitions instead.
736 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
737 Details : Update which_patch & target file for SLES10 latest kernel.
740 Description: Remove client patches from SLES 10 kernel.
741 Details : This causes SLES 10 clients to behave as patchless clients
742 even on a Lustre-patched (server) kernel.
744 Severity : enhancement
746 Description: self-adjustable client's lru lists
747 Details : use adaptive algorithm for managing client cached locks lru
748 lists according to current server load, other client's work
749 pattern, memory activities, etc. Both, server and client
750 side namespaces provide number of proc tunables for controlling
753 Severity : enhancement
755 Description: light-weight GSS support
756 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
757 and reduce performance overhead.
759 Severity : enhancement
761 Description: Linux keyring support
762 Details : Support using service of Linux keyring for Lustre GSS internal
763 context refresh/cache mechanism.
767 Description: Fix errors in lfs documentation
768 Details : Fixes man pages
772 Description: don't use GFP_* in generic Lustre code.
773 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
777 Description: obdclass is limited by single OBD_ALLOC(idarray)
778 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
782 Description: warnings with build patchless client with vanila 2.6.19 and up
783 Details : change old ctl_table style and replace ctl_table/ctl_table_header
784 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
788 Frequency : Only for SLES
789 Description: sanity_quota fail test_1
790 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
791 file, which makes TSTID[2] inunique.
795 Frequency : Only for SLES9
796 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
797 Details : Update target/ChangeLog/which_patch .
801 Frequency : Only for test_10 in sanity-quota.sh
802 Description: a bug in quota test code
803 Details : checking a test flag in a wrong place
807 Frequency : Only for test_13 in sanity-quota.sh
808 Description: a bug in quota test script
809 Details : shouldn't check overfull allocated quota
813 Description: jbd statistics
814 Details : Port older jbd statistics patch for sles10
816 Severity : enhancement
818 Description: Add printing inode info into message about error in writepage.
823 Description: reply_lock_interpret crash due to race with it and lock cancel.
824 Details : Do not replay locks that are being cancelled. Do not reference
825 locks by their address during replay, just by their handle.
829 Frequency : When flocks are used.
830 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
831 Details : Flock locks might destroy just granted lock if it could be merged
832 with another existing flock, this is done in completion handler,
833 so teach ldlm_cli_enquque_fini that this is a valid case for
839 Description: Oops in read and write path when failing to allocate lock.
840 Details : Check if lock allocation failed and return error back.
844 Description: lstripe command fails for valid OST index
845 Details : The stripe offset is compared to lov->desc.ld_tgt_count
846 instead of lov->desc.ld_active_tgt_count.
850 Description: sanity.sh failed test 103
851 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
856 Description: df doesn't work properly if diskfs blocksize != 4K
857 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
861 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
862 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
864 Severity : enhancement
866 Description: Adaptive timeouts
867 Details : RPC timeouts adapt to changing server load and network
868 conditions to reduce resend attempts and improve recovery time.
872 Description: llapi_file_create() does not allow some changes
873 Details : add llapi_file_open() that allows specifying the mode and
874 open flags, and also returns an open file handle.
878 Description: merge and cleanup kernel patches.
879 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
883 Description: Add journal checksum support.(Kernel part)
884 Details : The journal checksum feature adds two new flags i.e
885 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
886 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
887 indicates that the commit block contains the checksum for
888 the blocks described by the descriptor blocks. Now commit
889 record can be sent to disk without waiting for descriptor
890 blocks to be written to disk. This behavior is controlled
891 using JBD2_FEATURE_ASYNC_COMMIT flag.
895 Description: OSS needs mutliple precreate threads
896 Details : Add ability to start more than one create thread per OSS.
900 Description: Sanity.sh test_65k failed: setstripe should have succeeded
901 Details : execute lfs setstripe on client
905 Description: mds_obd_create error creating tmp object
906 Details : When the user sets quota on root, llog will be affected and can't
907 create files and write files.
910 Frequency : Always on ia64 patchless client, and possibly others.
912 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
913 Details : This allows the patchless client to be loaded on architectures
918 Description: RedHat Update kernel for RHEL5
919 Details : Add the kernel config file for RHEL5.
923 Description: RedHat Update kernel for RHEL5
924 Details : Modify the kernel config file more closer RHEL5.
928 Description: Build failure against Centos5 (RHEL5)
929 Details : Define PAGE_SIZE when it isn't present.
933 Description: test_7 on sanity-quota.sh can't be used on separate machines
934 Details : add facet to handle it
938 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
939 Details : with NFS, the anon dentry's parent was set to itself in
940 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
941 corresponding dentry other than use rec->ur_name.
943 Severity : enhancement
945 Description: lfs setstripe enhancement
946 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
950 Description: enable data checksumming by default
951 Details : enable checksum by default, allow --disable-checksum
952 configure option and "-o nochecksum" mount option. Checksums
953 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
957 Description: lustre support for RHEL5
958 Details : Add support for RHEL5.
962 Description: Client eviction due to failover config
963 Details : after a connection loss, the lustre client should attempt to
964 reconnect to the last active server first before trying the
965 other potential connections.
969 Description: when mds and osts use different quota unit(32bit and 64bit),
970 quota will be released repeatly.
971 Details : void sending multiple quota reqs to mds, which will keep the
972 status between the reqs.
976 Description: osts not allocated evenly to files
977 Details : change the condition to increase offset_idx
980 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
982 Description: Data corruption for OSTs that are formatted larger than 2TB
984 Details : When generating the bio request for lustre file writes the
985 sector number would overflow a temporary variable before being
986 used for the IO. The data reads correctly from Lustre (which
987 will overflow in a similar manner) but other file data or
988 filesystem metadata may be corrupted in some cases.
992 Description: Tune the kernel for good SCSI performance.
993 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
994 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
999 Description: rewrite ext2-derived code in obdclass/uuid.c
1000 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1001 from scratch preserving functionality.
1005 Description: "lfs find -obd UUID" prints directories
1006 Details : "lfs find -obd UUID" will return all directory names instead
1007 of just file names. It is incorrect because the directories
1008 do not reside on the OSTs.
1012 Description: use i_size_read and i_size_write in 2.6 port
1013 Details : replace inode->i_size access with i_size_read/write()
1016 Frequency : when removing large files
1018 Description: scheduling issue during removal of large Lustre files
1019 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1020 It causes scheduling issues when removing large files (17TB in the
1024 Frequency : only with liblustre clients on XT3
1026 Description: evictions taking too long
1027 Details : allow llrd to evict clients directly on OSTs
1030 Frequency : only on ppc
1032 Description: /proc/fs/lustre/devices broken on ppc
1033 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1034 We should make sure the type of 'index' is loff_t and then cast
1035 explicitly as needed below. Do not assign an explicitly cast
1039 Frequency : only for rhel5
1041 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1042 Details : Modify the target file & which_kernel.
1044 Severity : enhancement
1046 Description: omit set fsid for export NFS
1047 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1048 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1052 Frequency : Always, for kernels after 2.6.16
1053 Description: Fix warning idr_remove called for id=.. which is not allocated.
1054 Details : Last kernels save old s_dev before kill super and not allow
1055 to restore from callback - restore it before call kill_anon_super.
1059 Description: buffer overruns could theoretically occur
1060 Details : llapi_semantic_traverse() modifies the "path" argument by
1061 appending values to the end of the origin string, and a
1062 overrun may occur. Adding buffer overrun check in liblustreapi.
1066 Description: Fix error on 'ls .' at the top of the Lustre mount.
1067 Details : Don't revalidate dentry if it is a root dentry.
1071 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1072 Details : Modify vm-tunables-rhel4.patch.
1076 Description: Kernel config for 2.6.18-vanilla.
1077 Details : Modify targets/2.6-vanilla.target.in.
1078 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1079 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1080 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1081 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1085 Description: improve handling recoverable errors
1086 Details : if request processig with error which can be recoverable on server
1087 request should be resend, otherwise page released from cache and
1092 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1093 Details : Modify target file & which_patch.
1094 A flaw was found in the IA32 system call emulation provided
1095 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1096 value could be stored in the %RAX register, which could trigger an
1097 out-of-bounds system call table access. An untrusted local user
1098 could exploit this flaw to run code in the kernel
1099 (ie a root privilege escalation). (CVE-2007-4573).
1103 Description: O_DIRECT bypasses client statistics.
1104 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1105 read_ahead_stats were updated. copied the stats section from
1106 osc_send_oap_rpc() into async_internal().
1110 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1114 Description: change order of libsysio includes
1115 Details : '#include sysio.h' should always come before '#include xtio.h'
1119 Description: handle "serious error: objid * already exists" more gracefully
1120 Details : If LAST_ID value on disk is smaller than the objects existing in
1121 the O/0/d* directories, it indicates disk corruption and causes an
1122 LBUG(). If the object is 0-length, then we should use the existing
1123 object. This will help to avoid a full fsck in most cases.
1125 Severity : enhancement
1127 Description: adapt the lustre_config script to support the upgrade case
1128 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1129 server targets to 1.6 in parallel.
1133 Description: To avoid grant space > avaible space when the disk is almost
1134 full. Without this patch you might see the error "grant XXXX >
1135 available" or some LBUG about grant, when the disk is almost
1137 Details : In filter_check_grant, for non_grant cache write, we should
1138 check the left space by if (*left > ungranted + bytes), instead
1139 of (*left > ungranted), because only we are sure the left space
1140 is enough for another "bytes", then the ungrant space should be
1141 increase. In client, we should update cl_avail_grant only there
1142 is OBD_MD_FLGRANT in the reply.
1146 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1147 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1148 described in CVE-2007-4573. This problem could allow untrusted
1149 local users to gain root access.
1152 Frequency : when using O_DIRECT and quotas
1154 Description: Incorrect file ownership on O_DIRECT output files
1155 Details : block usage reported by 'lfs quota' does not take into account
1156 files that have been written with O_DIRECT.
1161 Description: touch file failed when fs is not full
1162 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1163 otherwise we can get ENOSP while fs is not full.
1167 Description: parallel lock callbacks
1168 Details : Instead of sending blocking and completion callbacks as separated
1169 requests, adding them to a set and sending in parallel.
1172 Frequency : only for Cray XT3
1173 Bugzilla : 12829/13455
1174 Description: Changing primary group doesn't change the group lustre assigns to
1176 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1177 overridden with the primary group provided by the group upcall,
1178 whereas the supplied fsgid can be trusted if it is in the list of
1179 supplementary groups returned by the group upcall.
1181 Severity : enhancement
1183 Description: Allow masking D_WARNING, D_ERROR messages from console
1184 Details : Console messages can now be disabled via lnet.printk.
1188 Description: User code with malformed file open parameter crashes client node
1189 Details : Before packing join_file req, all the related reference should be
1190 checked carefully in case some malformed flags cause fake
1191 join_file req on client.
1195 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1196 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1197 mode is equal to the granted mode. However, it should also check
1198 that LDLM_FL_CP_REQD is not set.
1202 Description: Heavy nfs access might result in deadlocks
1203 Details : After ELC code landed, it is now improper to enqueue any mds
1204 locks under och_sem, because enqueue might want to decide to
1205 cancel open locks for same inode we are holding och_sem for.
1209 Description: Client eviction while running blogbench
1210 Details : A lot of unlink operations with concurrent I/O can lead to a
1211 deadlock causing evictions. To address the problem, the number of
1212 oustanding OST_DESTROY requests is now throttled to
1213 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1214 callbacks are processed in priority.
1218 Description: enable ACLs on MDS by default
1219 Details : ACLs must be enabled on MDS by default.
1222 Frequency : PPC/PPC64 only
1224 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1225 Details : fix duplicated definitions between asm-ppc64/types.h and
1226 lustre_types.h on PPC.
1229 Frequency : PPC/PPC64 only
1231 Description: asm-ppc/segment.h does not exist
1232 Details : fix compile issue on PPC.
1236 Description: better handle error messages in extents code
1239 Frequency : RHEL4 only
1241 Description: mkfs is very slow on IA64/RHEL4
1242 Details : A performance regression has been discovered in the MPT Fusion
1243 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1244 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1245 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1247 Severity : enhancement
1249 Description: SNMP support enhancement
1250 Details : Adding total number of sampled request for an MDS node in snmp
1253 Severity : enhancement
1255 Description: Optimize ldlm waiting list processing for PR extent locks
1256 Details : When processing waiting list for read extent lock and meeting read
1257 lock that is same or wider to it that is not contended, skip
1258 processing rest of the list and immediatelly return current
1259 status of conflictness, since we are guaranteed there are no
1260 conflicting locks in the rest of the list.
1264 Description: Time out and refuse to reconnect
1265 Details : When the failover node is the primary node, it is possible
1266 to have two identical connections in imp_conn_list. We must
1267 compare not conn's pointers but NIDs, otherwise we can defeat
1268 connection throttling.
1272 Description: port llog fixes from b1_6 into HEAD
1273 Details : Port llog reference couting and some llog cleanups from b1_6
1274 (bug 10800) into HEAD, for protect from panic and access to already
1275 free llog structures.
1279 Description: Detect stride IO mode in read-ahead
1280 Details : When a client does stride read, read-ahead should detect that and
1281 read-ahead pages according to the detected stride pattern.
1285 Description: data checksumming impacts single node performance
1286 Details : add support for several checksum algorithm. Currently, only CRC32
1287 and Adler-32 are supported. The checksum type can be changed on
1288 the fly via /proc/fs/lustre/osc/*/checksum_type.
1292 Description: use adler32 for page checksums
1293 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1298 Description: build for x2 fails
1299 Details : fix compile issue on Cray systems.
1303 Description: Properly match for duplicate locks
1304 Details : Due to different lock order from skiplists code, we need to
1305 traverse entire list for now
1308 Frequency : only on PPC/SLES10
1310 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1311 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1312 be defined. Add a hack in mkfs_lustre.c to work around this compile
1317 Description: LASSERT on MDS when client holding flock lock dies
1318 Details : ldlm pool logic depends on number of granted locks equal to
1319 number of released locks which is not true for flock locks, so
1320 just exclude such locks from consideration.
1324 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1325 Details : Use fsfilt_sync() for both the whole filesystem sync and
1326 individual file sync to eliminate dangerous inode locking
1327 with I_LOCK that can lead to a deadlock.
1331 Description: performance in 1.6.3
1332 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1333 sure that 1MB requests are not fragmented by the block layer.
1335 Severity : enhancement
1337 Description: organize the server-side client stats on per-nid basis
1338 Details : Change the structure of stats under obdfilter and mds to
1346 The "uuid"s file would list the uuids of _active_ exports.
1347 And the clear entry is to clear all stats and stale nids.
1349 Severity : enhancement
1351 Description: eliminate client locks in face of contention
1352 Details : file contention detection and lockless i/o implementation
1353 for contended files.
1357 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1360 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1362 Description: MDS may lose file striping (and hence file data) in some cases
1363 Details : If there are additional extended attributes stored on the MDS,
1364 in particular ACLs, SELinux, or user attributes (if user_xattr
1365 is specified for the client mount options) then there is a risk
1366 of attribute loss. Additionally, the Lustre file striping
1367 needs to be larger than default (e.g. striped over all OSTs),
1368 and an additional attribute must be stored initially in the
1369 inode and then increase in size enough to be moved to the
1370 external attribute block (e.g. ACL growing in size) for file
1375 Description: skiplist implementation simplification
1376 Details : skiplists are used to group compatible locks on granted list
1377 that was implemented as tracking first and last lock of each
1378 lock group the patch changes that to using doubly linked lists
1382 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1383 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1384 because LASSERT on both the data supplied by a client, and the
1385 data on disk is dangerous and incorrect.
1387 Severity : enhancement
1389 Description: Slow truncate/writes to huge files at high offsets.
1390 Details : Directly associate cached pages to lock that protect those pages,
1391 this allows us to quickly find what pages to write and remove
1392 once lock callback is received.
1396 Description: more ldlm soft lockups
1397 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1398 starve other threads from the resource lock for a long time in
1399 case of long waiting queue, so change the debug level from
1400 D_OTHER to the less frequently used D_INFO.
1402 Severity : enhancement
1404 Description: add -gid, -group, -uid, -user options to lfs find
1408 Description: Hung threads in invalidate_inode_pages2_range
1409 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1410 once one is completed. As a result, some RPCs are stuck in the
1411 queue and are never sent.
1415 Description: filter threads hungs on waiting journal commit
1416 Details : Cleanup filter group llog code, then only filter group llog will
1417 be only created in the MDS/OST syncing process.
1421 Description: Procfs and llog threads access destoryed import sometimes.
1422 Details : Sync the import destoryed process with procfs and llog threads by
1423 the import refcount and semaphore.
1425 Severity : enhancement
1427 Description: openlock cache of b1_6 port to HEAD
1432 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1433 Details : If insertion of an extent fails, then discard the inode
1434 preallocation and free data blocks else it can lead to duplicate
1439 Description: don't always update ctime in ext3_xattr_set_handle()
1440 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1441 In some cases the ctime should not be updated, for example for
1442 2.0->1.8 compatibility it is necessary to delete an xattr and it
1443 should not update the ctime.
1447 Bugzilla : 15713/16362
1448 Description: Assertion in iopen_connect_dentry in 1.6.3
1449 Details : looking up an inode via iopen with the wrong generation number can
1450 populate the dcache with a disconneced dentry while the inode
1451 number is in the process of being reallocated. This causes an
1452 assertion failure in iopen since the inode's dentry list contains
1453 both a connected and disconnected dentry.
1457 Description: assertion failure in ldlm_handle2lock()
1458 Details : fix a race between class_handle_unhash() and class_handle2object()
1459 introduced in lustre 1.6.5 by bug 13622.
1464 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1465 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1470 Description: oops in sd_iostats_seq_show()
1471 Details : unloading/reloading the scsi low level driver triggers a kernel
1472 bug when trying to access the sd iostat file.
1477 Description: Kernel panics during QLogic driver reload
1478 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1479 patch, causing memory corruption.
1484 Description: journal_dev option does not work in b1_6
1485 Details : pass mount option during pre-mount.
1487 Severity : enhancement
1489 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1490 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1491 extent information of a file. It can be used to map logical blocks
1492 in a file to physical blocks in the block device.
1496 Description: LDLM soft lockups - improvement
1497 Details : It is be possible to send the lock handle along with each read
1498 or write request because the client is already doing a lock match
1499 itself so there isn't any reason the OST should have to re-do that
1503 Frequency : only on Cray X2
1505 Description: X2 build failures
1506 Details : fix build failures on Cray X2.
1510 Description: xid & resent requests
1511 Details : Initialize RPC XID from clock at startup (randomly if clock is
1514 Severity : enhancement
1516 Description: Add lustre_start utility to start or stop multiple Lustre servers
1521 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1522 Details : In case of memory pressure, list_del() can be called twice on
1523 req->rq_history_list, causing a kernel oops.
1527 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1528 Details : fix stack overflow in the distributed lock manager by defering
1529 export eviction after a failed AST to the elt thread instead of
1530 handling it in the dlm interpret routine.
1534 Description: Convert some comments to new format.
1535 Details : Update documenting comments to match doxygen conventions.
1539 Description: Grammar fixes.
1540 Details : A couple of trivial spelling fixes.
1544 Description: OSD_COUNTERS-mandatory
1545 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1546 to simplify the code.
1550 Description: simplify cmm_device freeing logic.
1551 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1555 Description: Add lockdep support to dt_object_operations locking interface.
1556 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1557 parameter indicating lock ordering. Update mdd code to use new
1562 Description: Introduce failloc constants for lockless IO tests.
1563 Details : Add two new failloc constants to test lockless IO. Only one of
1564 them in implemented---another is checked in yet to be landed
1569 Description: Add lockdep support for inode mutex.
1570 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1571 in the situations where Lustre has to lock more than one inode
1576 Description: Add optional invariants checking support.
1577 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1578 configure switch. This macro is to be used for consistency and
1579 sanity checks that are too expensive to be left in `production'
1584 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1585 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1590 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1591 Details : Use spin_lock_nested() in (the only) situation where more than
1592 one ldlm_lock is locked simultaneously. Also, fix possible
1593 dead-lock in ldlm_lock_change_resource() by enforcing particular
1598 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1599 Details : Instead of specifying each ldlm_lock call-back through separate
1600 parameter, wrap them into struct ldlm_callback_suite.
1604 Description: Kill join_lru obd method and its callers.
1605 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1606 in memory---a requirement implemented through ->o_join_lru() obd
1607 method in HEAD. Remove this method and its users.
1611 Description: Add asynchronous ldlm ENQUEUE completion handler.
1612 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1613 case that stock ldlm_completion_ast() cannot handle as it waits
1614 until lock is granted. Introduce new ldlm_completion_ast_async()
1615 for this. Also comment ldlm_completion_ast().
1619 Description: ldlm_error <-> errno conversion.
1620 Details : Add functions to map (rather arbitrary) between LDLM error codes
1621 and standard errno values. CLIO needs this to prevent LDLM specific
1622 constants from escaping ldlm and osc.
1626 Description: Kill unused ldlm_handle2lock_ns() function.
1627 Details : Kill unused ldlm_handle2lock_ns() function.
1631 Description: Add lu_ref support to ldlm_lock
1632 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1633 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1634 and ldlm_resource. LDLM interface has to be changed, because code that
1635 releases a reference on a lock, has to "know" what reference this is.
1636 In the most frequent case
1638 lock = ldlm_handle2lock(handle);
1640 LDLM_LOCK_PUT(lock);
1642 no changes are required. When any other reference (received _not_ from
1643 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1644 instead of LDLM_LOCK_PUT().
1646 Arguably, changes are pervasive, and interface requires some discipline
1647 for proper use. On the other hand, it was very instrumental in finding
1648 a few leaked lock references.
1652 Description: Add ldlm_lock_addref_try().
1653 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1654 attempts to addref a lock that might be being canceled
1659 Description: Add ldlm_weigh_callback().
1660 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1661 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1662 hard-coded `number of pages' logic.
1666 Description: Add lockdep annotations to llog code.
1667 Details : Use appropriately tagged _nested() locking calls in the places
1668 where llog takes more than one ->lgh_lock lock.
1672 Description: Add loi_kms_set().
1673 Details : Wrap kms updates into a helper function.
1677 Description: Constify instances of struct lsm_operations.
1678 Details : Constify instances of struct lsm_operations.
1682 Description: lu_conf support.
1683 Details : On a server, a file system object is uniquely identified
1684 by a fid, which is sufficient to locate and load all object
1685 state (inode). On a client, on the other hand, more data are
1686 necessary instantiate an object. Change lu_object_find() and
1687 friends to take additional `lu_conf' argument describing object.
1688 Typically this includes layout information.
1692 Description: lu_context fixes.
1693 Details : Introduce new lu_context functions that are needed on the client
1694 side, where some system threads (ptlrpcd) are shared by multiple
1695 modules, and so cannot be stopped during module shutdown.
1699 Description: Add start and stop methods to lu_device_type_operations.
1700 Details : Introduce two new methods in lu_device_type_operations, that are
1701 invoked when first instance of a given type is created and last one
1702 is destroyed respectively. This is need by CLIO.
1706 Description: Add lu_ref support to struct lu_device.
1707 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1708 track leaked references.
1712 Description: Introduce lu_kmem_descr.
1713 Details : lu_kmem_descr and its companion interface allow to create
1714 and destroy a number of kmem caches at once.
1718 Description: Fix lu_object finalization race.
1719 Details : Fix a race between lu_object_find() finding an object and its
1720 concurrent finalization. This race is (most likely) not possible
1721 on the server, but might happen on the client.
1725 Description: Introduce lu_ref interface.
1726 Details : lu_ref is a debugging module allowing to track references to
1727 a given object. It is quite cpu expensive, and has to be
1728 explicitly enabled with --enable-lu_ref. See usage description
1733 Description: Factor lu_site procfs stats into a separate function.
1734 Details : Separate lu_site stats printing code into a separate function
1735 that can be reused on a client.
1739 Description: Constify instances of struct {lu,dt,md}_device_operations.
1740 Details : Constify instances of struct {lu,dt,md}_device_operations.
1744 Description: Introduce struct md_site and move meta-data specific parts of
1745 struct lu_site here.
1746 Details : Move md-specific fields out of struct lu_site into special struct
1747 md_site, so that lu_site can be used on a client.
1751 Description: Kill mdd_lov_destroy().
1752 Details : Remove unused mdd code.
1756 Description: Add st_block checking to multistat.c.
1757 Details : Add st_block checking to multistat.c.
1761 Description: Add lu_ref support to struct obd_device.
1762 Details : Add lu_ref tracking to obd_device.
1766 Description: Kill obd_set_fail_loc().
1767 Details : Remove unused code.
1771 Description: Add special type for ptlrpc_request interpret functions.
1772 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
1773 there. Actual usage will be in CLIO.
1777 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
1778 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
1779 doesn't work with lockdep.
1783 Description: Add rwv.c test program.
1784 Details : New testing program exercising readv(2) and writev(2) (Qian).
1788 Description: Add sendfile.c test program.
1789 Details : New testing program exercising sendfile(2) (Jay).
1793 Description: Ratelimit a message that can be very frequent.
1794 Details : Ratelimit a memory allocation failure message that can
1799 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1800 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
1804 Description: Liblustre build fixes.
1805 Details : Liblustre build fixes.
1809 Description: libcfs: add cfs_{need,cond}_resched() interface.
1810 Details : libcfs: add cfs_{need,cond}_resched() definition and
1811 implementations for Linux, NT, and liblustre.
1813 Severity : enhancement
1815 Description: More exported tunables for mballoc
1816 Details : Add support for tunable preallocation window and new tunables for
1817 large/small requests
1821 Description: Detect corruption of block bitmap and checking for preallocations
1822 Details : Checks validity of on-disk block bitmap. Also it does better
1823 checking of number of applied preallocations. When corruption is
1824 found, it turns filesystem readonly to prevent further corruptions.
1828 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1829 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1830 window to 0. If it does hit cache miss, and read-pattern is still
1831 stride-io mode, does not reset the stride window, but also does not
1832 increase the stride window length in this case.
1836 Frequency : only for big-endian servers
1837 Description: Check if system is big-endian while mounting fs with extents feature
1838 Details : Mounting a filesystem with extents feature will fail on big-endian
1839 systems since ext3-based ldiskfs is not supported on big-endian
1840 systems. This can be over-riden with "bigendian_extents" mount option.
1842 Severity : enhancement
1844 Description: The root squash functionality
1845 Details : A security feature, which is to prevent users from being able
1846 to mount lustre on their desktop, run as root, and delete
1847 all of the files in the filesystem. The goal is accomplished by
1848 remapping user id (UID) and group id (GID) of the root user to
1849 a UID and GID specified by the system administartor via Lustre
1850 configuration management server (MGS). The functionality also
1851 allows to specify sets of clients for which the remapping does
1856 Description: Excessive recovery window
1857 Details : With AT enabled, the recovery window can be excessively long (6000+
1858 seconds). To address this problem, we no longer use
1859 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
1860 timeout no longer depends on the service time, it is set to
1861 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
1862 time via pb_service_time.
1864 --------------------------------------------------------------------------------
1866 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1868 * Support for kernels:
1869 2.6.5-7.283 (SLES 9),
1870 2.6.9-55.EL (RHEL 4),
1871 2.6.16.46-0.14 (SLES 10),
1872 2.6.18.8 vanilla (kernel.org)
1873 * Client support for unpatched kernels:
1874 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1875 2.6.16 - 2.6.22 vanilla (kernel.org)
1876 * Due to recently discovered recovery problems, we do not recommend
1877 using patchless RHEL 4 clients with this or any earlier release.
1878 * Recommended e2fsprogs version: 1.39.cfs8
1879 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1880 * Starting with this release, the ldiskfs backing filesystem required
1881 by Lustre is now in its own package, lustre-ldiskfs. This package
1882 should be installed. It is versioned separately from Lustre and
1883 may be released separately in future.
1887 Description: block reactivating mgc import until all deactivates complete
1888 Details : Fix race when failing back MDT/MGS to itself (testing)
1891 Frequency : at statup only
1893 Description: mds_lov_synchronize race leads to various problems
1894 Details : simultaneous MDT->OST connections at startup can cause the
1895 sync to abort, leaving the OSC in a bad state.
1897 Severity : enhancement
1899 Description: add optional extra BUILD_VERSION info
1900 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1901 to override the lustre version.
1904 Frequency : 2.6.18 servers only
1906 Description: ll_kern_mount() doesn't release the module reference
1907 Details : The ldiskfs module reference count never drops down to 0
1908 because ll_kern_mount() doesn't release the module reference.
1913 Description: server LBUG when using old ost_num_threads parameter
1914 Details : Accept the old ost_num_threads parameter but warn that it
1915 is deprecated, and fix an off-by-one error that caused an LBUG.
1920 Description: Transient SCSI error results in persistent IO issue
1921 Details : iobuf->dr_error is not reinitialized to 0 between two
1925 Frequency : sometimes when underlying device returns I/O errors
1927 Description: OSTs not going read-only during write failures
1928 Details : OSTs are not remounted read-only when the journal commit threads
1929 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1930 instead of the ext3 wrappers.
1934 Description: poor connect scaling with increasing client count
1935 Details : Don't run filter_grant_sanity_check for more than 100 exports
1936 to improve scaling for large numbers of clients.
1939 Frequency : SLES10 only
1941 Description: sanity-quota.sh quotacheck failed: rc = -22
1942 Details : Quotas cannot be enabled on SLES10.
1945 Frequency : liblustre clients only
1947 Description: getdirentries does not give error when run on compute nodes
1948 Details : getdirentries does not fail when the size specified as an argument
1949 is too small to contain at least one entry
1951 Severity : enhancement
1953 Description: Add LNET router traceability for debug purposes
1954 Details : If a checksum failure occurs with a router as part of the
1955 IO path, the NID of the last router that forwarded the bulk data
1956 is printed so it can be identified.
1961 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1962 Details : Due to a race condition, liblustre clients were occasionally
1963 evicted incorrectly.
1965 Severity : enhancement
1967 Description: lfs setstripe use optional parameters instead of postional
1970 Severity : enhancement
1972 Description: Nanosecond timestamp support for ldiskfs
1973 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1974 resolution timestamps. There is not yet support for this at
1975 the Lustre filesystem level.
1978 Frequency : during server recovery
1980 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1981 Details : request with rq_no_resend flag not awake l_wait_event if they get
1985 Frequency : nfs export on patchless client
1987 Description: connectathon hang when test nfs export over patchless client
1988 Details : Disconnected dentry cannot be found with lookup, so we do not need
1989 to unhash it or make it invalid
1992 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1993 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1994 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1999 Description: open req refcounting wrong on reconnect
2000 Details : If reconnect happened between getting open reply from server and
2001 call to mdc_set_replay_data in ll_file_open, we will schedule
2002 replay for unreferenced request that we are about to free.
2003 Subsequent close will crash in variety of ways.
2004 Check that request is still eligible for replay in
2005 mdc_set_replay_data().
2010 Description: disable writes to filesystem when reading health_check file
2011 Details : the default for reading the health_check proc file has changed
2012 to NOT do a journal transaction and write to disk, because this
2013 can cause reads of the /proc file to hang and block HA state
2014 checking on a healthy but otherwise heavily loaded system. It
2015 is possible to return to the previous behaviour during configure
2016 with --enable-health-write.
2018 Severity : enhancement
2020 Description: 64-bit inode version
2021 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2022 to the inode. This will be required for version-based recovery.
2027 Description: MDS fails to start if a duplicate client export is detected
2028 Details : in some rare cases it was possible for a client to connect to
2029 an MDS multiple times. Upon recovery the MDS would detect this
2030 and fail during startup. Handle this more gracefully.
2032 Severity : enhancement
2034 Description: Add -o localflock option to simulate old noflock
2036 Details : This will achieve local-only flock/fcntl locks
2042 Description: log_commit_thread vs filter_destroy race leads to crash
2043 Details : Take import reference before releasing llog record semaphore
2048 Description: Wrong request locking in request set processing
2049 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2050 imp_delayed_list, in this place should be used imp_lock.
2053 Frequency : when reconnection
2055 Description: Grant Leak when osc reconnect to OST
2056 Details : When osc reconnect ost, OST(filter) should check whether it
2057 should grant more space to client by comparing fed_grant and
2058 cl_avail_grant, and return the granted space to client instead
2059 of "new granted" space, because client will call osc_init_grant
2060 to update the client grant space info.
2063 Frequency : when client reconnect to OST
2065 Description: Grant Leak when osc do resend and replay bulk write
2066 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2067 bulk write request, because the grant info while be sync between
2068 OSC and OST when reconnect, and we should ignore the grant info
2069 these of resend/replay write req.
2074 Description: Grant space more than avaiable left space sometimes.
2075 Details : When then OST is about to be full, if two bulk writing from
2076 different clients came to OST. Accord the avaliable space of the
2077 OST, the first req should be permitted, and the second one
2078 should be denied by ENOSPC. But if the seconde arrived before
2079 the first one is commited. The OST might wrongly permit second
2080 writing, which will cause grant space > avaiable space.
2083 Frequency : when client is evicted
2085 Description: Grant might be wrongly erased when osc is evicted by OST
2086 Details : when the import is evicted by server, it will fork another
2087 thread ptlrpc_invalidate_import_thread to invalidate the
2088 import, where the grant will be set to 0. While the original
2089 thread will update the grant it got when connecting. So if
2090 the former happened latter, the grant will be wrongly errased
2091 because of this race.
2096 Description: Checking Stale with correct fid
2097 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2098 to check whether it is stale, because sometimes, we want the
2099 enqueue happened anyway, and op_data.fid2 will not be initialized.
2101 Severity : enhancement
2103 Description: update patchless client
2104 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2107 Frequency : only with 2.4 kernel
2109 Description: random memory corruption
2110 Details : size of struct ll_inode_info is to big for union inode.u and this
2111 can be cause of random memory corruption.
2116 Description: Memory leak in recovery
2117 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2118 It should also check obd_fail before fsfilt_start, otherwise if
2119 fsfilt_start return -EROFS,(failover mds during mds recovery).
2120 then the req will return with repmsg->transno = 0 and rc = EROFS.
2121 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2122 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2123 be freed no matter whether fsfilt_commit success or not.
2126 Frequency : only with huge count clients
2128 Description: Prevents from taking the superblock lock in llap_from_page for
2130 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2131 for llap_from_page call in ll_removepage prevents from taking the
2132 superblock lock for a soon died page.
2137 Description: Not check open intent error before release open handle
2138 Details : in some rare cases, the open intent error is not checked before
2139 release open handle, which may cause
2140 ASSERTION(open_req->rq_transno != 0), because it tries to release
2141 the failed open handle.
2146 Description: Set cat log bitmap only after create log success.
2147 Details : in some rare cases, the cat log bitmap is set too early. and it
2148 should be set only after create log success.
2152 Description: Accessing a block bevice can re-enable I/O when Lustre is
2153 tearing down a device.
2154 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2159 Description: service threads may hog cpus when there are a lot of requests
2160 Details : Insert cond_resched to give other threads a chance to use some CPU
2165 Description: the cat log was not initialized in recovery
2166 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2167 unlink log on mds will not be initialized until mds post
2168 recovery. And also in mds post recovery, the unlink log will
2169 initialization will be done asynchronausly, so there will be race
2170 between add unlink log and unlink log initialization.
2174 Description: brw_stats were being printed incorrectly
2175 Details : brw_stats were being printed as log2 but all of them were not
2176 recorded as log2. Also remove some code duplication arising from
2177 filter_tally_{read,write}.
2181 Frequency : rare, only in recovery.
2182 Description: ASSERTION(req->rq_type != LI_POISON) failed
2183 Details : imp_lock should be held while iterating over imp_sending_list for
2184 prevent destroy request after get timeout in ptlrpc_queue_wait.
2188 Description: replay-single.sh test 52 fails
2189 Details : A lock's skiplist need to be cleanup when it being unlinked
2190 from its resource list.
2194 Description: Short directio read returns full requested size rather than
2196 Details : Direct I/O operations should return actual amount of bytes
2197 transferred rather than requested size.
2199 Severity : enhancement
2201 Description: metadata RPC reduction (e.g. for rm performance)
2202 Details : decrease the amount of synchronous RPC between clients and servers
2203 by canceling conflicing lock before the operation on the client
2204 and packing thier handles into the main operation RPC to server.
2206 Severity : enhancement
2208 Description: Async OSC create to avoid the blocking unnecessarily.
2209 Details : If a OST has no remain object, system will block on the creating
2210 when need to create a new object on this OST. Now, ways use
2211 pre-created objects when available, instead of blocking on an
2212 empty osc while others are not empty. If we must block, we block
2213 for the shortest possible period of time.
2217 Description: improve handling recoverable errors
2218 Details : if request processig with error which can be recoverable on server
2219 request should be resend, otherwise page released from cache and
2222 Severity : enhancement
2224 Description: refine locking for avoid write wrong info into lov_objid file
2225 Details : fix possible races with add new target and write/update data in
2228 --------------------------------------------------------------------------------
2230 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2235 Frequency : on some architectures
2237 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2238 Details : Uninitialized flags sometimes cause configuration commands to
2242 Frequency : patchless clients only
2244 Description: missing __iget() symbol export
2245 Details : The __iget() symbol export is missing. To avoid the need for
2246 this on patchless clients the deathrow inode reaper is turned
2247 off, and we depend on the VM to clean up old inodes. This
2248 dependency was during via the fix for bug 12181.
2250 --------------------------------------------------------------------------------
2252 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2254 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2255 INTEROPERATE with older versions automatically. In many cases a
2256 special upgrade step is needed. Please read the
2257 user documentation before upgrading any part of a live system.
2258 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2259 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2260 * WARNING: Lustre configuration and startup changes are required with
2261 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2265 Severity : enhancement
2267 Description: MountConf
2268 Details : Lustre configuration is now managed via mkfs and mount
2269 commands instead of lmc and lconf. New obd types (MGS, MGC)
2270 are added for dynamic configuration management. See
2271 https://mail.clusterfs.com/wikis/lustre/MountConf for
2274 Severity : enhancement
2276 Description: dynamic OST addition
2277 Details : OSTs can now be added to a live filesystem
2279 Severity : enhancement
2281 Description: startup order invariance
2282 Details : MDTs and OSTs can be started in any order. Clients only
2283 require the MDT to complete startup.
2285 Severity : enhancement
2287 Description: parallel, asynchronous orphan cleanup
2288 Details : orphan cleanup is now performed in separate threads for each
2289 OST, allowing parallel non-blocking operation.
2291 Severity : enhancement
2293 Description: optimized stripe assignment
2294 Details : stripe assignments are now made based on ost space available,
2295 ost previous usage, and OSS previous usage, in order to try
2296 to optimize storage space and networking resources.
2298 Severity : enhancement
2300 Description: Permanently set tunables
2301 Details : All writable /proc/fs/lustre tunables can now be permanently
2302 set on a per-server basis, at mkfs time or on a live system.
2304 Severity : enhancement
2306 Description: Lustre message v2
2307 Details : Add lustre message format v2.
2309 Severity : enhancement
2311 Description: client OST exclusion list
2312 Details : Clients can be started with a list of OSTs that should be
2313 declared "inactive" for known non-responsive OSTs.
2317 Description: SPEC SFS validation failure on NFS v2 over lustre.
2318 Details : Changes the blocksize for regular files to be 2x RPC size,
2319 and not depend on stripe size.
2321 Severity : enhancement
2323 Description: Multiple MD RPCs in flight.
2324 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2325 To avoid overly-overloading MDS, introduce a limit on number of
2326 MDS RPCs in flight for a single client and add /proc controls
2327 to adjust this limit.
2329 Severity : enhancement
2331 Description: client read/write statistics
2332 Details : Add client read/write call usage stats for performance
2333 analysis of user processes.
2334 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2335 file access. extents_stats shows chunk size distribution.
2336 extents_stats_per_process show chunk size distribution per
2339 Severity : enhancement
2341 Description: mds statistics
2342 Details : Add detailed mds operations statistics in
2343 /proc/fs/lustre/mds/*/stats.
2347 Description: Failure of copying files with lustre special EAs.
2348 Details : Client side always return success for setxattr call for lustre
2349 special xattr (currently only "trusted.lov").
2354 Description: Refcount LNET uuids
2355 Details : The global LNET uuid list grew linearly with every startup;
2356 refcount repeated list entries instead of always adding to
2359 Severity : enhancement
2361 Description: Dynamic service threads
2362 Details : Within a small range, start extra service threads
2363 automatically when the request queue builds up.
2366 Frequency : mixed-endian client/server environments
2368 Description: mixed-endian crashes
2369 Details : The new msg_v2 system had some failures in mixed-endian
2372 Severity : enhancement
2374 Description: Easy OST removal
2375 Details : OSTs can be permanently deactivated with e.g. 'lctl
2376 conf_param lustre-OST0001.osc.active=0'
2378 Severity : enhancement
2380 Description: MGS proc entries
2381 Details : Added basic proc entries for the MGS showing what filesystems
2384 Severity : enhancement
2386 Description: provide MGS failover
2387 Details : Added config lock reacquisition after MGS server failover.
2389 Severity : enhancement
2391 Description: add Linux 2.4 support
2392 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2396 Description: a large application tries to do I/O to the same resource and dies
2397 in the middle of it.
2398 Details : Check the req->rq_arrival time after the call to
2399 ost_brw_lock_get(), but before we do anything about
2400 processing it & sending the BULK transfer request. This
2401 should help move old stale pending locks off the queue as
2402 quickly as obd_timeout.
2405 Frequency : when an incorrect nid is specified during startup
2407 Description: ptlrpc connect to non-existant node causes kernel crash
2408 Details : LNET can't be re-entered from an event callback, which
2409 happened when we expire a message after the export has been
2410 cleaned up. Instead, hand the zombie cleanup off to another
2413 Severity : enhancement
2415 Description: plain/inodebits lock performance improvement
2416 Details : Grouping plain/inodebits in granted list by their request modes
2417 and bits policy, thus improving the performance of search through
2421 Frequency : only if OST filesystem is corrupted
2423 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2424 Details : for a short time RPCs with bulk IO are in the replay list,
2425 but replay of bulk IOs is unimplemented. If the OST filesystem
2426 is corrupted due to disk cache incoherency and then replay is
2427 started it is possible to trip an assertion. Avoid putting
2428 committed RPCs into the replay list at all to avoid this issue.
2431 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2433 Description: System hang on startup
2434 Details : This bug allowed the liblustre (e.g. catamount) client to
2435 return to the app before handling all startup RPCs. This
2436 could leave the node unresponsive to lustre network traffic
2437 and manifested as a server ptllnd timeout.
2439 Severity : enhancement
2441 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2442 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2443 boolean to control whether to print peer debug info when a
2444 client's RPC times out.
2447 Frequency : only for kernels with patches from Lustre below 1.4.3
2449 Description: Remove old rdonly API
2450 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2453 Frequency : only for devices with external journals
2455 Description: Set external device read-only also
2456 Details : During a commanded failover stop, we set the disk device
2457 read-only while the server shuts down. We now also set any
2458 external journal device read-only at the same time.
2461 Frequency : when upgrading from 1.4 while trying to change parameters
2463 Description: The wrong (new) MDC name was used when setting parameters for
2464 upgraded MDT's. Also allows changing of OSC (and MDC)
2465 parameters if --writeconf is specified at tunefs upgrade time.
2468 Frequency : when setting specific ost indicies
2470 Description: QOS code breaks on skipped indicies
2471 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2472 created with --index need not be sequential.
2474 Severity : enhancement
2476 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2477 Details : The uninit_groups feature works in conjunction with the kernel
2478 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2479 up the pass1 processing of e2fsck. This is a read-only feature
2480 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2481 mount filesystems that have had this feature enabled.
2483 Severity : enhancement
2485 Description: Improve multi-block allocation algorithm to avoid fragmentation
2486 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2487 allocation locality and avoid filesystem fragmentation.
2489 ------------------------------------------------------------------------------
2491 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2493 * Support for kernels:
2494 2.4.21-47.0.1.EL (RHEL 3)
2495 2.6.5-7.283 (SLES 9)
2496 2.6.9-42.0.10.EL (RHEL 4)
2497 2.6.12.6 vanilla (kernel.org)
2498 2.6.16.27-0.9 (SLES 10)
2499 * Recommended e2fsprogs version: 1.39.cfs5
2501 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2505 Frequency : occasional, depends on client load and configuration
2506 Bugzilla : 12181, 12203
2507 Description: data loss for recently-modified files
2509 Details : In some cases it is possible that recently written or created
2510 files may not be written to disk in a timely manner (this should
2511 normally be within 30s unless client IO load is very high).
2512 The problem appears as zero-length files or files that are a
2513 multiple of 1MB in size after a client crash or client eviction
2514 that are missing data at the end of the file.
2516 This problem is more likely to be hit on clients where files are
2517 repeatedly created and unlinked in the same directory, clients
2518 have a large amount of RAM, have many CPUs, the filesystem has
2519 many OSTs, the clients are rebooted frequently, and/or the files
2520 are not accessed by other nodes after being written.
2522 The presence of the problem can be detected by looking at
2523 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2524 smaller than the second (nr_unused) then dirty files will not
2525 be flushed automatically to disk. "sync; sleep 10" should be
2526 run several times on the node before unmounting it to update
2527 Lustre (this is also safe to run on nodes without this problem).
2529 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2530 that can cause this same problem, so customers using that kernel
2531 also need to update the kernel in addition to Lustre. In order
2532 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2534 It is normal that files written just before a client crash (less
2535 than 30s) may not yet have been flushed to disk, even for local
2539 Frequency : frequent on thin XT3 nodes
2541 Description: UUID collision on thin XT3 Linux nodes
2542 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2543 randomly, since we relied on an insufficiently-seeded PRNG.
2548 Description: OSS hangs after "All ost request buffers busy"
2549 Details : A deadlock between quota and journal operations caused OSS
2550 hangs after printing "All ost request buffers busy."
2553 Frequency : always on liblustre builds
2555 Description: Cleanup compiler warnings on liblustre
2558 Frequency : always on liblustre builds on XT3
2560 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2564 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2565 > 32000 subdirectories
2566 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2567 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2568 subdirectory count crosses 32000. This will aid e2fsck to
2569 correctly handle more than 32000 subdirectories.
2572 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2574 Description: System hang on startup
2575 Details : This bug allowed the liblustre (e.g. catamount) client to
2576 return to the app before handling all startup RPCs. This
2577 could leave the node unresponsive to lustre network traffic
2578 and manifested as a server ptllnd timeout.
2580 Severity : enhancement
2582 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2583 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2584 boolean to control whether to print peer debug info when a
2585 client's RPC times out.
2590 Description: make O_SYNC working on 2.6 kernels
2591 Details : 2.6 kernels use different method for mark pages for write,
2592 so need add a code to lustre for O_SYNC work.
2597 Description: Failure to close file and release space on NFS
2598 Details : Put inode details into lock acquired in ll_intent_file_open.
2599 Use mdc_intent_lock in ll_intent_open to properly
2600 detect all kind of errors unhandled by mdc_enqueue.
2605 Description: proc file read during shutdown sometimes raced obd removal,
2607 Details : Add lock to prevent obd access after proc file removal.
2610 Frequency : Only for files larger than 4GB on 32-bit clients.
2612 Description: improperly doing page alignment of locks
2613 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2614 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2619 Description: RPCs being resent when they shouldn't be
2620 Details : Some RPCs that should not be resent are being resent. This
2621 can cause inconsistencies in the RPC state machine. Do not
2622 resend such requests.
2625 Frequency : rare, only with NFS export
2627 Description: Crash on NFS re-export node
2628 Details : under very unusual load conditions an assertion is hit in
2629 ll_intent_file_open()
2632 Frequency : only if OST filesystem is corrupted
2634 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2635 Details : for a short time RPCs with bulk IO are in the replay list,
2636 but replay of bulk IOs is unimplemented. If the OST filesystem
2637 is corrupted due to disk cache incoherency and then replay is
2638 started it is possible to trip an assertion. Avoid putting
2639 committed RPCs into the replay list at all to avoid this issue.
2644 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2645 Details : Large single O_DIRECT read and write calls can fail to allocate
2646 a sufficiently large buffer to process the request. In case of
2647 allocation failure the allocation is retried with a smaller
2648 buffer and broken into smaller requests.
2650 Severity : enhancement
2652 Description: Add -o localflock option to simulate old noflock behaviour.
2653 Details : This will achieve local-only flock/fcntl locks coherentness.
2658 Description: versioning check is incomplete
2659 Details : Checking the version difference of client vs. server, report
2660 error if the gap is too big.
2665 Description: add support PG_writeback bit
2666 Details : add support for PG_writeback bit for Lustre, for more carefull
2667 work with page cache in 2.6 kernel. This also fix some deadlocks
2668 and remove hack for work O_SYNC with 2.6 kernel.
2670 Severity : enhancement
2672 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2673 Details : The uninit_groups feature works in conjunction with the kernel
2674 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2675 up the pass1 processing of e2fsck. This is a read-only feature
2676 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2677 mount filesystems that have had this feature enabled.
2679 Severity : enhancement
2681 Description: Improve multi-block allocation algorithm to avoid fragmentation
2682 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2683 allocation locality and avoid filesystem fragmentation.
2685 ------------------------------------------------------------------------------
2687 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2689 * Support for kernels:
2690 2.6.9-42.0.3.EL (RHEL 4)
2691 2.6.5-7.276 (SLES 9)
2692 2.4.21-47.0.1.EL (RHEL 3)
2693 2.6.12.6 vanilla (kernel.org)
2694 2.6.16.21-0.8 (SLES10)
2695 * Recommended e2fsprogs version: 1.39.cfs2-0
2697 * The backwards-compatible /proc/sys/portals symlink has been removed
2698 in this release. Before upgrading, please ensure that you change
2699 any configuration scripts or /etc/sysctl.conf files that access
2700 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2701 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2702 in advance of the upgrade on any system running Lustre 1.4.6 or
2703 newer, since /proc/sys/lnet was added in that version.
2704 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2708 Frequency : only when quota is used
2710 Description: avoid scanning export list for quota master
2711 Details : Change the algorithms to avoid scanning export list in order
2712 to improve the efficiency.
2715 Frequency : MDS failover only, very rarely
2717 Description: "went back in time" messages on mds failover
2718 Details : The greatest transno may be lost when the current operation
2719 finishes with an error (transno==0) and the client's last_rcvd
2720 record is over-written. Save the greatest transno in the
2721 mds_last_transno for this case.
2724 Frequency : always for specific kernels and striping counts
2726 Description: client may get "Matching packet too big" without ACL support
2727 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2728 when trying to access files in certain configurations. The
2729 clients should in fact be denied when mounting because they do
2730 not understand ACLs.
2733 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2735 Description: many clients connecting with IO in progress causes connect timeouts
2736 Details : Avoid synchronous journal commits to avoid delays caused by many
2737 clients connecting/disconnecting when bulk IO is in progress.
2738 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2739 OST_IO_PORTAL to avoid delays behind potentially many pending
2743 Frequency : occasionally with multiple writers to a single file
2745 Description: shared writes to file may result in wrong size reported by stat()
2746 Details : Allow growing of kms when extent lock is cancelled
2749 Frequency : always with random mmap IO to multi-striped file
2751 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2752 Details : Only if the hole is at the end of OST object so that kms is too
2753 small. Fix is to increase kms accordingly in ll_nopage.
2756 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2758 Description: writes to a missing object would leak memory on the OST
2759 Details : If there is an inconsistency between the MDS and OST filesystems,
2760 such that the MDS references an object that doesn't exist, writes
2761 to that object will leak memory due to incorrect cleanup in the
2762 error handling path, eventually running out of memory on the OST.
2767 Description: Creating too long symlink causes lustre errors
2768 Details : Check symlink and name lengths before sending requests to MDS.
2771 Frequency : only if flock is enabled (not on by default)
2773 Description: posix locks not released on fd closure on 2.6.9+
2774 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2775 kernels, this caused such locks not to be released on fd close and
2776 then assertions on fs unmount about still used locks.
2779 Frequency : MDS failover only, very rarely
2781 Description: clients may get ASSERTION(granted_lock != NULL)
2782 Details : When request was taking a long time, and a client was resending
2783 a getattr by name lock request. The were multiple lock requests
2784 with the same client lock handle and
2785 mds_getattr_name->fixup_handle_for_resent_request found one of the
2786 lock handles but later failed with ASSERTION(granted_lock != NULL).
2791 Description: handle->h_buffer_credits > 0, assertion failure
2792 Details : h_buffer_credits is zero after truncate, causing assertion
2793 failure. This patch extends the transaction or creates a new
2797 Frequency : NFS re-export or patchless client
2798 Bugzilla : 11179, 10796
2799 Description: Crash on NFS re-export node (__d_move)
2800 Details : We do not want to hash the dentry if we don't have a lock.
2801 But if this dentry is later used in d_move, we'd hit uninitialised
2802 list head d_hash, so we just do this to init d_hash field but
2803 leave dentry unhashed.
2806 Frequency : NFS re-export or patchless client
2808 Description: NFS exports has problem with symbolic link
2809 Details : lustre client didn't properly install dentry when re-exported
2810 to NFS or running patchless client.
2813 Frequency : NFS re-export or patchless client
2815 Description: Various nfs/patchless fixes.
2816 Details : fixes reuse disconected alias for lookup process - this fixes
2817 warning "find_exported_dentry: npd != pd",
2818 fix permission error with open files at nfs.
2819 fix apply umask when do revalidate.
2822 Frequency : occasional
2824 Description: Crash on NFS re-export node
2825 Details : calling clear_page() on the wrong pointer triggered oops in
2826 generic_mapping_read().
2829 Frequency : rarely, using O_DIRECT IO
2831 Description: unaligned directio crashes client with LASSERT
2832 Details : check for unaligned buffers before trying any requests.
2835 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2837 Description: stale data returned from RAID cache
2838 Details : If only a small amount of IO is done to the RAID device before
2839 reading it again it is possible to get stale data from the RAID
2840 cache instead of reading it from disk.
2843 Frequency : always for sles10 kernel
2845 Description: sles10 support
2846 Details : ll_follow_link: compile fixes and using of nd_set_link
2847 under newer kernels.
2850 Frequency : depends on arch, kernel and compiler version, always on sles10
2853 Description: recursive or deep enough symlinks cause stack overflow
2854 Details : getting rid of large stack-allocated variable in
2858 Frequency : depends on hardware
2860 Description: lustre write performance loss in the SLES10 kernel
2861 Details : the performance loss is caused by using of write barriers in the
2862 ext3 code. The SLES10 kernel turns barrier support on by
2863 default. The fix is to undo that change for ldiskfs.
2865 ------------------------------------------------------------------------------
2867 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2869 * Support for kernels:
2870 2.6.9-42.0.3EL (RHEL 4)
2871 2.6.5-7.276 (SLES 9)
2872 2.4.21-47.0.1.EL (RHEL 3)
2873 2.6.12.6 vanilla (kernel.org)
2877 Frequency : quota enabled and large files being deleted
2879 Description: releasing more than 4GB of quota at once hangs OST
2880 Details : If a user deletes more than 4GB of files on a single OST it
2881 will cause the OST to spin in an infinite loop. Release
2882 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2887 Description: statfs data retrieved from /proc may be stale or zero
2888 Details : When reading per-device statfs data from /proc, in the
2889 {kbytes,files}_{total,free,avail} files, it may appear
2890 as zero or be out of date.
2893 Frequency : systems with MD RAID1 external journal devices
2895 Description: lconf's call to blkid is confused by RAID1 journal devices
2896 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2897 of returning all block devices that match the journal UUID.
2900 Frequency : always, for aggregate stripe size over 4GB
2902 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2903 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2904 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2907 Frequency : always if "lfs find" used on a local file/directory
2909 Description: "lfs find" segfaults if used on a local file/directory
2910 Details : The case where a directory component was not specified wasn't
2911 handled correctly. Handle this properly.
2914 Frequency : always on ppc64
2916 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2917 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2918 perform endianness conversion.
2921 Frequency : rarely (truncate to non-zero file size after write under load)
2922 Bugzilla : 10730, 10687
2923 Description: Files padded with zeros to next 4K multiple
2924 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2925 it is possible that files that are truncated to a non-zero size
2926 immediately after being written are filled with zero bytes beyond
2927 the truncated size. No file data is lost.
2929 Severity : enhancement
2931 Description: Allow recovery/failover for liblustre clients.
2932 Details : liblustre clients were unaware of failover configurations until
2935 Severity : enhancement
2937 Description: user file locks should fail when not mounting with flock option
2938 Details : Set up an error-returning stub in ll_file_operations.lock field
2939 to prevent incorrect behaviour when client is mounted without
2940 flock option. Also, set up properly f_op->flock field for
2944 Frequency : always on ia64
2946 Description: "lfs df" loops on printing out MDS statfs information
2947 Details : The obd_ioctl_data was not initialized and in some systems
2948 this caused a failure during the ioctl that did not return
2949 an error. Initialize the struct and return an error on failure.
2952 Frequency : SLES 9 only
2954 Description: Error of copying files with lustre special EAs as root
2955 Details : Client side always return success for setxattr call for lustre
2956 special xattr (currently only "trusted.lov").
2959 Frequency : rarely on clusters with both ia64+i386 clients
2961 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2962 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2963 clients are doing concurrent writes to the same file it is
2964 possible that the ia64 clients may LASSERT because the OST
2965 extent locks are not PAGE_SIZE aligned. Ensure that grown
2966 locks are always aligned on the request boundary.
2969 Frequency : specific use, occasional
2971 Description: Overwriting in use executable truncates on-disk binary image
2972 Details : If one node attempts to overwrite an executable in use by
2973 another node, we now correctly return ETXTBSY instead of
2974 truncating the file.
2976 Severity : enhancement
2978 Description: Async OSC create to avoid the blocking unnecessarily.
2979 Details : If a OST has no remain object, system will block on the creating
2980 when need to create a new object on this OST. Now, ways use
2981 pre-created objects when available, instead of blocking on an
2982 empty osc while others are not empty. If we must block, we block
2983 for the shortest possible period of time.
2988 Description: chmod on Lustre root is propagated to other clients
2989 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2990 invalid by the follow_mount time.
2993 Frequency : liblustre clients only
2995 Description: Race in 'instant cancel' lock handling could lead to such locks
2996 never to be granted in case of SMP MDS
2997 Details : Do not destroy not yet granted but cbpending locks in
3001 Frequency : replay/resend of open
3003 Description: non null lock assetion failure in mds_intent_policy
3004 Details : Trying to replay/resend lockless open requests resulted in
3005 mds_open() returning 0 with no lock. Now it sets a flag if
3006 a lock is going to be returned.
3008 Severity : enhancement
3010 Description: Checksum enhancements
3011 Details : New checksum enhancements allow for resending RPCs that failed
3014 Severity : enhancement
3016 Description: Tunables on number of dirty pages in cacche
3017 Details : Allow to set limit on number of dirty pages cached.
3022 Description: client crash on unmount - lock still has references
3023 Details : In some error handling cases it was possible to leak a lock
3024 reference on a client while accessing a file. This was not
3025 harmful to the client during operation, but would cause the
3026 client to crash when the filesystem is unmounted.
3029 Frequency : specific case, rare
3031 Description: ETXTBSY on mds though file not in use by client
3032 Details : ETXTBSY is no longer incorrectly returned when attempting to
3033 chmod or chown a directory that the user previously tried to
3034 execute or a currently-executing binary.
3037 Frequency : extremely rare except on liblustre-based clients
3039 Description: Lustre space not freed when files are deleted
3040 Details : Clean up open-unlinked files after client eviction. Previously
3041 the unlink was skipped and the files remained as orphans.
3046 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3047 Details : In some cases it was possible to send a blocking callback to
3048 a client doing a glimpse, even though that client didn't get
3049 a lock granted. When the glimpse lock is cancelled on the OST
3050 the freed lock is left on the waiting list and corrupted the list.
3053 Frequency : all core dumps
3055 Description: Broke core dumps to lustre
3056 Details : Negative dentry may be unhashed if parent does not have UPDATE
3057 lock, but some callers, e.g. do_coredump, expect dentry to be
3058 hashed after successful create, hash it in ll_create_it.
3060 ------------------------------------------------------------------------------
3062 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3064 * Support for kernels:
3065 2.6.9-42.0.2.EL (RHEL 4)
3066 2.6.5-7.276 (SLES 9)
3067 2.4.21-40.EL (RHEL 3)
3068 2.6.12.6 vanilla (kernel.org)
3072 Frequency : always on RHEL 3
3074 Description: Number of open files grows over time
3075 Details : The number of open files grows over time, whether or not
3076 Lustre is started. This was due to a filp leak introduced
3077 by one of our kernel patches.
3079 ------------------------------------------------------------------------------
3081 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3083 * Support for kernels:
3084 2.6.9-42.EL (RHEL 4)
3085 2.6.5-7.276 (SLES 9)
3086 2.4.21-40.EL (RHEL 3)
3087 2.6.12.6 vanilla (kernel.org)
3092 Bugzilla : 5719, 9635, 9792, 9684
3093 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3094 Details : If a server is under heavy load and cannot reply to new
3095 connection requests before the client resends the (re)connect,
3096 the connection handling code can behave badly if two service
3097 threads are concurrently handing separate (re)connections from
3098 the same client. Add better locking to the connection handling
3099 code, and ensure that only a single connection will be processed
3100 for a given client UUID, even if the lock is dropped.
3102 Severity : enhancement
3104 Description: add TCP zero-copy support to kernel
3105 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3106 sends if the hardware supports scatter-gather and checksumming.
3107 This allows socklnd to do client-write and server-read more
3108 efficiently and reduce CPU utilization from skbuf copying.
3111 Frequency : only if NFS exporting from client
3113 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3114 Details : If setting attributes on a file created under NFS that had
3115 never been opened it would be possible to oops the client
3116 if the file had no objects.
3119 Frequency : always for liblustre
3121 Description: liblustre client does MDS+OSTs setattr RPC for each write
3122 Details : When doing a write from a liblustre client, the client
3123 incorrectly issued an RPC to the MDS and each OST the file was
3124 striped over in order to update the timestamps. When writing
3125 with small chunks and many clients this could overwhelm the MDS
3126 with RPCs. In all cases it would slow down the write because
3127 these RPCs are unnecessary.
3129 Severity : enhancement
3131 Description: allow number of MDS service threads to be changed at module load
3132 Details : It is now possible to change the number of MDS service threads
3133 running. Adding "options mds mds_num_threads={N}" to the MDS's
3134 /etc/modprobe.conf will set the number of threads for the next
3135 time Lustre is restarted (assuming the "mds" module is also
3136 reloaded at that time). The default number of threads will
3137 stay the same, 32 for most systems.
3142 Description: OST crash if filesystem is unformatted or corrupt
3143 Details : If an OST is started on a device that has never been formatted
3144 or if the filesystem is corrupt and cannot even mount then the
3145 error handling cleanup routines would dereference a NULL pointer.
3150 Description: NULL pointer deref in llap_from_page.
3151 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3152 incorrect) mapping if the page was truncated/reclaimed while it was
3153 searched for. Check for this condition and skip such pages when
3154 doing readahead. Introduce extra check to llap_from_page() to
3155 verify page->mapping->host is non-NULL (so page is not anonymous).
3158 Frequency : Sometimes when using sys_sendfile
3160 Description: "page not covered by a lock" warnings from ll_readpage
3161 Details : sendfile called ll_readpage without right page locks present.
3162 Now we introduced ll_file_sendfile that does necessary locking
3163 around call to generic_file_sendfile() much like we do in
3167 Frequency : with certain MDS communication failures at client mount time
3169 Description: NULL pointer deref after failed client mount
3170 Details : a client connection request may delayed by the network layer
3171 and not be sent until after the PTLRPC layer has timed out the
3172 request. If the client fails the mount immediately it will try
3173 to clean up before the network times out the request. Add a
3174 reference from the request import to the obd device and delay
3175 the cleanup until the network drops the request.
3178 Frequency : occasionally during client (re)connect
3180 Description: assertion failure during client (re)connect
3181 Details : processing a client connection request may be delayed by the
3182 client or server longer than the client connect timeout. This
3183 causes the client to resend the connection request. If the
3184 original connection request is replied in this interval, the
3185 client may trip an assertion failure in ptlrpc_connect_interpret()
3186 which thought it would be the only running connect process.
3189 Frequency : only with obd_echo servers and clients that are rebooted
3191 Description: kernel BUG accessing uninitialized data structure
3192 Details : When running an obd_echo server it did not start the ping_evictor
3193 thread, and when a client was evicted an uninitialized data
3194 structure was accessed. Start the ping_evictor in the RPC
3195 service startup instead of the OBD startup.
3197 Severity : enhancement
3198 Bugzilla : 10193 (patchless)
3199 Description: Remove dependency on various unexported kernel interfaces.
3200 Details : No longer need reparent_to_init, exit_mm, exit_files,
3201 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3204 Frequency : rare (only users of deprecated and unsupported LDAP config)
3206 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3207 Details : LDAP apparently contains 'lustreName' attributes instead of
3208 'name'. A simple remapping of the name is sufficient.
3211 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3213 Description: waiting_locks_callback trips kernel BUG if client is evicted
3214 Details : Running with the dump_on_timeout debug flag turned on makes
3215 it possible that the waiting_locks_callback() can try to dump
3216 the Lustre kernel debug logs from an interrupt handler. Defer
3217 this log dumping to the expired_lock_main() thread.
3219 Severity : enhancement
3221 Description: Support NFS exporting on 2.6 kernels.
3222 Details : Implement non-rawops metadata methods for NFS server to use without
3223 changing NFS server code.
3226 Frequency : very rare (synthetic metadata workload only)
3228 Description: two racing renames might cause an MDS thread to deadlock
3229 Details : Running the "racer" program may cause one MDS thread to rename
3230 a file from being the source of a rename to being the target of
3231 a rename at exactly the same time that another thread is doing
3232 so, and the second thread has already enqueued these locks after
3233 doing a lookup of the target and is trying to relock them in
3234 order. Ensure that we don't try to re-lock the same resource.
3237 Frequency : only very large systems with liblustre clients
3239 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3240 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3241 all OSTs in parallel. This allows the eviction of stale liblustre
3242 clients to proceed much faster than if they were done in series,
3243 and also offers similar improvements for other set_info RPCs.
3248 Description: excessive CPU usage during initial read phase on client
3249 Details : During the initial read phase on a client, it would agressively
3250 retry readahead on the file, consuming too much CPU and impacting
3251 performance (since 1.4.5.8). Improve the readahead algorithm
3252 to avoid this, and also improve some other common cases (read
3253 of small files in particular, where "small" is files smaller than
3254 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3259 Description: MDS crash when receiving packet with unknown intent.
3260 Details : Do not LBUG in unknown intent case, just return -EFAULT
3262 Severity : enhancement
3263 Bugzilla : 9293, 9385
3264 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3265 Details : Do not serialize getattr (non-intent version) and statfs.
3268 Frequency : occasional, when OST network is overloaded/intermittent
3270 Description: client evicted by OST after bulk IO timeout
3271 Details : If a client sends a bulk IO request (read or write) the OST
3272 may evict the client if it is unresposive to its data GET/PUT
3273 request. This is incorrect if the network is overloaded (takes
3274 too long to transfer the RPC data) or dropped the OST GET/PUT
3275 request. There is no need to evict the client at all, since
3276 the pinger and/or lock callbacks will handle this, and the
3277 client can restart the bulk request.
3280 Frequency : Always when mmapping file with no objects
3282 Description: client crashes when mmapping file with no objects
3283 Details : Check that we actually have objects in a file before doing any
3284 operations on objects in ll_vm_open, ll_vm_close and
3290 Description: Request leak when working with deleted CWD
3291 Details : Introduce advanced request refcount tracking for requests
3292 referenced from lustre intent.
3294 Severity : Enhancement
3296 Description: Cache open file handles on client.
3297 Details : MDS now will return special lock along with openhandle, if
3298 requested and client is allowed to hold openhandle, even if unused,
3299 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3300 closing files for every read/write openration.
3302 Severity : Enhancement
3304 Description: Cache open negative dentries on client when possible.
3305 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3306 negative dentries on lock revocation.
3311 Description: Remounting a client read-only wasn't possible with a zconf mount
3312 Details : It wasn't possible to remount a client read-only with llmount.
3314 Severity : enhancement
3315 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3316 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3317 MPICH is included to add Lustre-specific ADIO interfaces.
3318 This is based closely on the UFS ADIO layer and only differs
3319 in file creation, in order to allow the OST striping to be set.
3320 This is user-contributed code and not supported by CFS.
3325 Description: extended inode attributes (immutable, append-only) work improperly
3326 when 2.4 and 2.6 kernels are used on client/server or vice versa
3327 Details : Introduce kernel-independent values for these flags.
3329 Severity : enhancement
3332 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3333 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3334 granularity. Now, Fractional MB granularity is be supported,
3335 this is very useful for low memory system.
3337 Severity : enhancement
3339 Description: Getattr by fid
3340 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3341 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3342 have locks and avoids getattr rpc requests for it on every stat.
3345 Frequency : Always, for filesystems larger than 2TB
3347 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3348 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3349 filesystems larger than 2TB to be mounted with mballoc enabled.
3352 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3354 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3355 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3356 filesystems larger than 2TB is mounted then the kernel will
3357 silently corrupt the start of the filesystem. CONFIG_LBD is
3358 enabled for all CFS-supported kernels, but the possibility of
3359 this happening with a modified kernel config exists.
3361 Severity : enhancement
3363 Description: add client O_DIRECT support for 2.6 kernels
3364 Details : It is now possible to do O_DIRECT reads and writes to files
3365 in the Lustre client mountpoint on 2.6 kernel clients.
3367 Severity : enhancement
3369 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3370 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3371 parallel, not waiting for response from every OST before sending
3372 a rpc to the next OST.
3377 Description: setattr vs write race when updating file timestamps
3378 Details : Client processes that update a file timestamp into the past
3379 right after writing to the file (e.g. tar) it is possible that
3380 the updated file modification time can be reset to the current
3381 time due to a race between processing the setattr and write RPC.
3383 Severity : enhancement
3385 Description: Bring 'lfs find' closer in line with regular Linux find.
3386 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3387 -print0 options and obtains all the needed info through the lustre
3390 Severity : enhancement
3392 Description: support up to 1024 configured devices on one node
3393 Details : change obd_dev array from statically allocated to dynamically
3394 allocated structs as they are first used to reduce memory usage
3399 Description: Flush dirty partially truncated pages during truncate
3400 Details : Immediatelly flush partially truncated pages in filter_setattr,
3401 this way we completely avoid having any pages in page cache on OST
3402 and can retire ugly workarounds during writes to flush such pages.
3407 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3408 Details : protect inode from truncation within vfs_unlink() context
3409 just take a reference before calling vfs_unlink() and release it
3410 when parent's i_sem is free.
3415 Description: last_id value checked outside lock on OST caused LASSERT failure
3416 Details : If there were multiple MDS->OST object precreate requests in
3417 flight, it was possible that the OST's last object id was checked
3418 outside a lock and incorrectly tripped an assertion. Move checks
3419 inside locks, and discard old precreate requests.
3422 Frequency : always, if extents are used on OSTs
3424 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3425 Details : OSTs using the extents format would not zero the high 16 bits of
3426 the index physical block number. This is not a problem for any
3427 OST filesystems smaller than 16TB, and no kernels support ext3
3428 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3429 new/modified files) and can be fixed for existing filesystems
3430 with e2fsprogs-1.39-cfs1.
3435 Description: import connection selection may be incorrect if timer wraps
3436 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3437 import connections to be ignored if the 32-bit jiffies counter
3438 wraps. Use a 64-bit jiffies counter.
3441 Frequency : very large clusters immediately after boot
3443 Description: LNET request buffers exhausted under heavy short-term load
3444 Details : If a large number of client requests are generated on a service
3445 that has previously never seen so many requests it is possible
3446 that the request buffer growth cannot keep up with the spike in
3447 demand. Instead of dropping incoming requests, they are held in
3448 the LND until the RPC service can accept more requests.
3451 Frequency : Sometimes during replay
3453 Description: Assertion failure in ll_local_open after replay.
3454 Details : If replay happened on an open request reply before we were able
3455 to set replay handler, reply will become not swabbed tripping the
3456 assertion in ll_local_open. Now we set the handler right after
3457 recognising of open request
3460 Frequency : very rare
3462 Description: kernel reports "badness in vsnprintf"
3463 Details : Reading from the "recovery_status" /proc file in small chunks
3464 may cause a negative length in lprocfs_obd_rd_recovery_status()
3465 call to vsnprintf() (which is otherwise harmless). Exit early
3466 if there is no more space in the output buffer.
3468 Severity : enhancement
3470 Description: clear OBD RPC statistics by writing to them
3471 Details : It is now possible to clear the OBD RPC statistics by writing
3472 to the "stats" file.
3477 Description: Client mtime is not the same on different clients after utimes
3478 Details : In some cases, the client was using the utimes() syscall on
3479 a file cached on another node. The clients now validate the
3480 ctime from the MDS + OSTs to determine which one is right.
3485 Description: Inability to activate failout mode
3486 Details : lconf script incorrectly assumed that in pythong string's numeric
3487 value is used in comparisons.
3490 Frequency : always with multiple stripes per file
3492 Description: Inefficient object allocation for mutli-stripe files
3493 Details : When selecting which OSTs to stripe files over, for files with
3494 a stripe count that divides evenly into the number of OSTs,
3495 the MDS is always picking the same starting OST for each file.
3496 Return the OST selection heuristic to the original design.
3501 Description: mount failures may take full timeout to return an error
3502 Details : Under some heavy load conditions it is possible that a
3503 failed mount can wait for the full obd_timeout interval,
3504 possibly several minutes, before reporting an error.
3505 Instead return an error as soon as the status is known.
3507 Frequency : quota enabled and large files being deleted
3509 Description: releasing more than 4GB of quota at once hangs OST
3510 Details : If a user deletes more than 4GB of files on a single OST it
3511 will cause the OST to spin in an infinite loop. Release
3512 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3517 Description: statfs data retrieved from /proc may be stale or zero
3518 Details : When reading per-device statfs data from /proc, in the
3519 {kbytes,files}_{total,free,avail} files, it may appear
3520 as zero or be out of date.
3523 Frequency : systems with MD RAID1 external journal devices
3525 Description: lconf's call to blkid is confused by RAID1 journal devices
3526 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3527 of returning all block devices that match the journal UUID.
3530 Frequency : always, for aggregate stripe size over 4GB
3532 Description: assertion fails when trying to use 4GB stripe size
3533 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3534 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3537 Frequency : always on ppc64
3539 Description: the first write on an ext3 filesystem with mballoc got stuck
3540 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3541 perform endianness conversion.
3543 ------------------------------------------------------------------------------
3545 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3547 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3548 INTEROPERATE with older versions automatically. Please read the
3549 user documentation before upgrading any part of a live system.
3550 * WARNING: Lustre networking configuration changes are required with
3551 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3554 * Support for kernels:
3555 2.6.9-22.0.2.EL (RHEL 4)
3556 2.6.5-7.244 (SLES 9)
3557 2.6.12.6 vanilla (kernel.org)
3560 Severity : enhancement
3561 Bugzilla : 7981/8208
3562 Description: Introduced Lustre Networking (LNET)
3563 Details : LNET is new networking infrastructure for Lustre, it includes
3564 a reorganized network configuration mode (see the user
3565 documentation for full details) as well as support for routing
3566 between different network fabrics. Lustre Networking Devices
3567 (LNDs) for the supported network fabrics have also been
3568 created for this new infrastructure.
3570 Severity : enhancement
3571 Description: Introduced Access control lists
3572 Details : clients can set ACLs on files and directories in order to have
3573 more fine-grained permissions than the standard Unix UGO+RWX.
3574 The MDS must be started with the "-o acl" mount option.
3576 Severity : enhancement
3577 Description: Introduced filesystem quotas
3578 Details : Administrators may now establish per-user quotas on the
3581 Severity : enhancement
3583 Description: Configuration change for the XT3
3584 The PTLLND is now used to run Lustre over Portals on the XT3
3585 The configure option(s) --with-cray-portals are no longer used.
3586 Rather --with-portals=<path-to-portals-includes> is used to
3587 enable building on the XT3. In addition to enable XT3 specific
3588 features the option --enable-cray-xt3 must be used.
3593 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3594 Details : A race between allocating a new llcd and re-getting the llcd_lock
3595 allowed another thread to grab newly-allocated llcd.
3597 Severity : enhancement
3599 Description: 2.6 OST async journal commit and locking fix to improve performance
3600 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3601 2.6 kernels are now async as they already were in 2.4 kernels so
3602 that they can commit concurrently with the network bulk transfer.
3603 For block-allocated files the filter allocation semaphore is held
3604 to avoid filesystem fragmentation during allocation. BKL lock
3605 removed for 2.6 xattr operations where it is no longer needed.
3610 Description: lconf incorrectly determined whether two IP networks could talk
3611 Details : In some more complicated routing and multiple-network
3612 configurations, lconf will avoid trying to make a network
3613 connection to a disjoint part of the IP space. It was doing the
3614 math incorrectly for one set of cases.
3619 Description: Fix for potential infinite loop processing records in an llog.
3620 Details : If an llog record is corrupted/zeroed, it is possible to loop
3621 forever in llog_process(). Validate the llog record length
3622 and skip the remainder of the block on error.
3625 Frequency : occasional (liblustre only)
3627 Description: liblustre could not open files whose last component is a symlink
3628 Details : sysio_path_walk() would incorrectly pass the open intent to
3629 intermediate path components.
3632 Frequency : rare (liblustre only with non-standard tuning)
3633 Bugzilla : 7201 (7350)
3634 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3635 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3636 after it has been released, only do this for referenced locks
3638 Severity : enhancement
3640 Description: specifying an (invalid) directory default stripe_size of -1
3641 would reset the directory default striping
3642 Details : stripe_size -1 was used internally to signal directory stripe
3643 removal, now use "all default" to signal dir stripe removal
3644 as a directory striping of "all default" is not useful
3647 Frequency : common for large clusters running liblustre clients
3649 Description: doing an ls when liblustre clients are running is slow
3650 Details : sending a glimpse AST to a liblustre client waits for every AST
3651 to time out, as liblustre clients will not respond. Since they
3652 cannot cache data we refresh the OST lock LVB from disk instead.
3654 Severity : enhancement
3656 Description: doing an ls at the same time as file IO can be slow
3657 Details : enqueue and other "small" requests can be blocked behind many
3658 large IO requests. Create a new OST IO portal for non-IO
3659 requests so they can be processed faster.
3662 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3664 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3665 created file with mode 000
3666 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3667 allow an ftruncate() as a non-root user to a file with mode 000.
3668 Lustre now respects this flag to disable mode checks when
3669 truncating a file owned by the user
3672 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3675 Description: Revoking locks from clients that went dead or catatonic might take
3677 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3678 cancellation of such locks instant on servers without waiting for
3679 any reply from clients. Clients drops these locks when cancel
3680 notification from server is received without replying.
3683 Frequency : liblustre-only, when liblustre client dies or becomes busy
3685 Description: Doing ls on Linux clients can take a long time with active
3687 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3688 granting such locks to it in the first place if possible. Locks
3689 are taken by proxy on the OST during the read or write and
3690 dropped immediately afterward. Add connect flags handling, do
3691 not grant locks to liblustre clients for glimpse ASTs.
3693 Severity : enhancement
3695 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3696 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3697 sequential reads and behaves badly with applications doing random
3698 reads. Improve it by reading ahead at least read region, and
3699 avoiding excessive large RPC for small reads.
3701 Severity : enhancement
3703 Description: Creating more than 1000 files for a single job may cause a load
3704 imbalance on the OSTs if there are also a large number of OSTs.
3705 Details : qos_prep_create() uses an OST index reseed value that is an
3706 even multiple of the number of available OSTs so that if the
3707 reseed happens in the middle of the object allocation it will
3708 still utilize the OSTs as uniformly as possible.
3713 Description: OST or MDS may oops in ping_evictor_main()
3714 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3715 but doesn't restart at beginning of obd_exports_timed list
3718 Severity : enhancement
3720 Description: improve by-nid export eviction on the MDS and OST
3721 Details : allow multiple exports with the same NID to be evicted at one
3722 time without re-searching the exports list.
3725 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3727 Description: MDS may oops in groups_free()
3728 Details : in rare race conditions a newly allocated group_info struct is
3729 freed again, and this can be NULL. The 2.4 compatibility code
3730 for groups_free() checked for a NULL pointer, but 2.6 did not.
3733 Frequency : common for liblustre clients doing little filesystem IO
3734 Bugzilla : 9352, 7313
3735 Description: server may evict liblustre clients accessing contended locks
3736 Details : if a client is granted a lock or receives a completion AST
3737 with a blocking AST already set it would not reply to the AST
3738 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3741 Frequency : lfs setstripe, only systems with more than 160 OSTs
3743 Description: unable to set striping with a starting offset beyond OST 160
3744 Details : llapi_create_file() incorrectly limited the starting stripe
3745 index to the maximum single-file stripe count.
3748 Frequency : LDAP users only
3750 Description: lconf did not handle in-kernel recovery with LDAP properly
3751 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3753 Severity : enhancement
3755 Description: bind OST threads to NUMA nodes to improve performance
3756 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3757 node and do their allocations there to localize memory access
3759 Severity : enhancement
3761 Description: llmount can determine client NID directly from Myrinet (GM)
3762 Details : the client NID code from gmnalnid was moved directly into
3763 llmount, removing the need to use this or specifying the
3764 client NID explicitly when mounting GM clients with zeroconf
3767 Frequency : if client is started with down MDS
3769 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3770 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3771 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3772 ptlrpc_import_delayed_req() because we didn't check if the
3773 request was interrupted, and we also didn't break out of the
3774 event loop if there was no timeout
3779 Description: data loss during non-page-aligned writes to a single file from
3780 both multiple nodes and multiple threads on one node at same time
3781 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3782 inconsistency led to false short-reads, that were cached and later
3783 used by ->prepare_write() to fill in partially written page,
3784 leading to data loss.
3787 Frequency : always, if lconf --abort_recovery used
3789 Description: lconf --abort_recovery fails with 'Operation not supported'
3790 Details : lconf was attempting to abort recovery on the MDT device and not
3793 Severity : enhancement
3795 Description: remove cleanup logs
3796 Details : replace lconf-generated cleanup logs with lustre internal
3797 cleanup routines. Eliminates the need for client-cleanup and
3800 Severity : enhancement
3802 Description: add support for EAs (user and system) on lustre filesystems
3803 Details : it is now possible to store extended attributes in the Lustre
3804 client filesystem, and with the user_xattr mount option it
3805 is possible to allow users to store EAs on their files also
3807 Severity : enhancement
3809 Description: Add possibility (config option) to show minimal available OST free
3811 Details : When compiled with --enable-mindf configure option, statfs(2)
3812 (and so, df) will return least minimal free space available from
3813 all OSTs as amount of free space on FS, instead of summary of
3814 free spaces of all OSTs.
3816 Severity : enhancement
3818 Description: do not expand extent locks acquired on OST-side
3819 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3820 by server: they are not cached anyway.
3823 Frequency : when mmap is used/binaries executed from Lustre
3825 Description: Unmmap pages before throwing them away from read cache.
3826 Details : llap_shrink cache now attempts to unmap pages before discarding
3827 them (if unmapping failed - do not discard). SLES9 kernel has
3828 extra checks that trigger if this unmapping is not done first.
3833 Description: lconf didn't resolve symlinks before checking to see whether a
3834 given mountpoint was already in use
3837 Frequency : when migrating failover services
3838 Bugzilla : 6395, 9514
3839 Description: When migrating a subset of services from a node (e.g. failback
3840 from a failover service node) the remaining services would
3841 time out and evict clients.
3842 Details : lconf --force (implied by --failover) sets the global obd_timeout
3843 to 5 seconds in order to quickly disconnect, but this caused
3844 other RPCs to time out too quickly. Do not change the global
3845 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3847 Severity : enhancement
3848 Frequency : if MDS is started with down OST
3849 Bugzilla : 9439,5706
3850 Description: Allow startup/shutdown of an MDS without depending on the
3851 availability of the OSTs.
3852 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3853 Add appropriate locking and lov-osc refcounts for safe
3854 cleaning. Add osc abort_inflight calls in case the
3855 synchronize never started.
3858 Frequency : occasional (Cray XT3 only)
3860 Description: root not authorized to access files in CRAY_PORTALS environment
3861 Details : The client process capabilities were not honoured on the MDS in
3862 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3863 been accessed by an authorized user then root was able to access
3864 the file on the local client also. The root user capabilities
3865 are now allowed on the MDS, as this environment has secure UID.
3868 Frequency : occasional
3870 Description: ldiskfs "too long searching" message happens too often
3871 Details : A debugging message (otherwise harmless) prints too often on
3872 the OST console. This has been reduced to only happen when
3873 there are fragmentation problems on the filesystem.
3878 Description: Division by zero in statfs when all OSCs are inactive
3879 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3880 lov_statfs divides by value returned by lov_get_stripecnt().
3884 Bugzilla : 9489, 3273
3885 Description: First write from each client to each OST was only 4kB in size,
3886 to initialize client writeback cache, which caused sub-optimal
3887 RPCs and poor layout on disk for the first writen file.
3888 Details : Clients now request an initial cache grant at (re)connect time
3889 and so that they can start streaming writes to the cache right
3890 away and always do full-sized RPCs if there is enough data.
3891 If the OST is rebooted the client also re-establishes its grant
3892 so that client cached writes will be honoured under the grant.
3897 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3898 Details : Now I/O RPCs go to different portal number and (presumably) fast
3899 lock requests (and glimses) and other RPCs get their own service
3900 threads pool that should be able to service those RPCs
3903 Severity : enhancement
3905 Description: Ability to exchange lustre version between client and servers and
3906 issue warnings at client side if client is too old. Also for
3907 liblustre clients there is ability to refuse connection of too old
3909 Details : New 'version' field is added to connect data structure that is
3910 filled with version info. That info is later checked by server and
3914 Frequency : rare, liblustre only.
3915 Bugzilla : 9296, 9581
3916 Description: Two simultaneous writes from liblustre at offset within same page
3917 might proceed at the same time overwriting eachother with stale
3919 Details : I/O lock withing llu_file_prwv was released too early, before data
3920 actually was hitting the wire. Extended lock-holding time until
3921 server acknowledges receiving data.
3924 Frequency : extremely rare. Never observed in practice.
3926 Description: avoid generating lustre_handle cookie of 0.
3927 Details : class_handle_hash() generates handle cookies by incrementing
3928 global counter, and can hit 0 occasionaly (this is unlikely, but
3929 not impossible, because initial value of cookie counter is
3930 selected randonly). Value of 0 is used as a sentinel meaning
3931 "unassigned handle" --- avoid it. Also coalesce two critical
3932 sections in this function into one.
3934 Severity : enhancement
3936 Description: allow liblustre clients to delegate truncate locking to OST
3937 Details : To avoid overhead of locking, liblustre client instructs OST to
3938 take extent lock in ost_punch() on client's behalf. New connection
3939 flag is added to handle backward compatibility.
3941 Severity : enhancement
3942 Bugzilla : 4928, 7341, 9758
3943 Description: allow number of OST service threads to be specified
3944 Details : a module parameter allows the number of OST service threads
3945 to be specified via "options ost ost_num_threads={N}" in the
3946 OSS's /etc/modules.conf or /etc/modprobe.conf.
3950 Bugzilla : 6146, 9635, 9895
3951 Description: servers crash with bad pointer in target_handle_connect()
3952 Details : In rare cases when a client is reconnecting it was possible that
3953 the connection request was the last reference for that export.
3954 We would temporarily drop the export reference and get a new
3955 one, but this may have been the last reference and the export
3956 was just destroyed. Get new reference before dropping old one.
3958 Severity : enhancement
3959 Frequency : if client is started with failover MDS
3961 Description: Allow multiple MDS hostnames in the mount command
3962 Details : Try to read the configuration from all specified MDS
3963 hostnames during a client mount in case the "primary"
3966 Severity : enhancement
3968 Description: Stop sending data to evicted clients as soon as possible.
3969 Details : Check if the client we are about to send or are sending data to
3970 was evicted already. (Check is done every second of waiting,
3971 for which l_wait_event interface was extended to allow checking
3972 of exit condition at specified intervals).
3975 Frequency : rare, normally only when NFS exporting is done from client
3977 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3979 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3982 Frequency : always, if srand() is called before liblustre initialization
3984 Description: Liblustre uses system PRNG disturbing its usage by user application
3985 Details : Introduce internal to lustre fast and high-quality PRNG for
3986 lustre usage and make liblustre and some other places in generic
3987 lustre code to use it.
3989 Severity : enhancement
3990 Bugzilla : 9477, 9557, 9870
3991 Description: Verify that the MDS configuration logs are updated when xml is
3992 Details : Check if the .xml configuration logs are newer than the config
3993 logs stored on the MDS and report an error if this is the case.
3994 Request --write-conf, or allow starting with --old_conf.
3996 Severity : enhancement
3998 Description: Handle symlinks in the path when checking if Lustre is mounted.
3999 Details : Resolve intermediate symlinks when checking if a client has
4000 mounted a filesystem to avoid duplicate client mounts.
4005 Description: lconf can hit an error exception but still return success.
4006 Details : The lconf command catches the Command error exception at the top
4007 level script context and will exit with the associated exit
4008 status, but doesn't ensure that this exit status is non-zero.
4013 Description: failure of ptlrpc thread startup can cause oops
4014 Details : Starting a ptlrpc service thread can fail if there are a large
4015 number of threads or the server memory is very fragmented.
4016 Handle this without oopsing.
4019 Frequency : always, only if liblustre and non-default acceptor port was used
4021 Description: liblustre cannot connect to servers with non-default acceptor port
4022 Details : tcpnal_set_default_params() was not called and was therefore
4023 ignoring the environment varaible TCPNAL_PORT, as well as other
4024 TCPNAL_ environment variables
4029 Description: two objects could be created on the same OST for a single file
4030 Details : If an OST is down, in some cases it was possible to create two
4031 objects on a single OST for a single file. No problems other
4032 than potential performance impact and spurious error messages.
4036 Bugzilla : 5681, 9562
4037 Description: Client may oops in ll_unhash_aliases
4038 Details : Client dcache may become inconsistent in race condition.
4039 In some cases "getcwd" can fail if the current directory is
4045 Description: Inode refcounting problems in NFS export code
4046 Details : link_raw functions used to call d_instantiate without obtaining
4047 extra inode reference first.
4051 Bugzilla : 9942, 9903
4052 Description: Referencing freed requests leading to crash, memleaks with NFS.
4053 Details : We used to require that call to ll_revalidate_it was always
4054 followed by ll_lookup_it. Also with revalidate_special() it is
4055 possible to call ll_revalidate_it() twice for the same dentry
4056 even if first occurence returned success. This fix changes semantic
4057 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4058 reference on a request referred from the intent.
4059 ll_intent_release() then releases such a request.
4062 Frequency : rare, normally benchmark loads only
4064 Description: unlinked inodes were kept in memory on the client
4065 Details : If a client is repeatedly creating and unlinking files it
4066 can accumulate a lot of stale inodes in the inode slab cache.
4067 If there is no other client load running this can cause the
4068 client node to run out of memory. Instead flush old inodes
4069 from client cache that have the same inode number as a new inode.
4072 Frequency : SLES9 2.6.5 kernel and long filenames only
4073 Bugzilla : 9969, 10379
4074 Description: utime reports stale NFS file handle
4075 Details : SLES9 uses out-of-dentry names in some cases, which confused
4076 the lustre dentry revalidation. Change it to always use the
4080 Frequency : rare, unless heavy write-truncate concurrency is continuous
4081 Bugzilla : 4180, 6984, 7171, 9963, 9331
4082 Description: OST becomes very slow and/or deadlocked during object unlink
4083 Details : filter_destroy() was holding onto the parent directory lock
4084 while truncating+unlinking objects. For very large objects this
4085 may block other threads for a long time and slow overall OST
4086 responsiveness. It may also be possible to get a lock ordering
4087 deadlock in this case, or run out of journal credits because of
4088 the combined truncate+unlink. Solution is to do object truncate
4089 first in one transaction without parent lock, and then do the
4090 final unlink in a new transaction with the parent lock. This
4091 reduces the lock hold time dramatically.
4094 Frequency : rare, 2.4 kernels only
4096 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4097 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4098 before unmounting their filesystem in order to avoid deadlock.
4099 The kernel_locked() function in 2.4 kernels only checks whether
4100 the kernel lock is held, not whether it is this process that is
4101 holding it as 2.6 kernels do.
4106 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4107 Details : The client ptlrpc code may be trying to reconnect to a down
4108 server before a previous connection attempt has timed out.
4109 Increase the reconnect interval to be longer than the connection
4110 timeout interval to avoid sending duplicate connections to
4114 Frequency : echo_client brw_test command
4116 Description: fix echo_client to work with OST preallocated code
4117 Details : OST preallocation code (5137) didn't take echo_client IO path
4118 into account: echo_client calls filter methods outside of any
4119 OST thread and, hence, there is no per-thread preallocated
4120 pages and buffers to use. Solution: hijack pga pages for IO. As
4121 a byproduct, this avoids unnecessary data copying.
4125 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4126 Description: Client can oops in mdc_commit_close() after open replay
4127 Details : It was possible for the MDS to return an open request with no
4128 transaction number in mds_finish_transno() if the client was
4129 evicted, but without actually returning an error. Clients
4130 would later try to replay that open and may trip an assertion
4131 Simplify the client close codepath, and always return an error
4132 from the MDS in case the open is not successful.
4135 Frequency : rare, 2.6 OSTs only
4137 Description: OST may deadlock under high load on fragmented files
4138 Details : If there was a heavy load and highly-fragmented OST filesystems
4139 it was possible to have all the OST threads deadlock waiting on
4140 allocation of biovecs, because the biovecs were not released
4141 until the entire RPC IO was completed. Instead, release biovecs
4142 as soon as they are complete to ensure forward IO progress.
4144 Severity : enhancement
4146 Description: Support for specifying external journal device at mount
4147 Details : If an OST or MDS device is formatted with an external journal
4148 device, this device major/minor is stored in the ext3 superblock
4149 and may not be valid for failover. Allow detecting and
4150 specifying the external journal at mount time.
4155 Description: Mounting an MDS with pending unlinked files may cause oops
4156 Details : target_finish_recovery() calls mds_postrecov() which returned
4157 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4158 considers this an error and immediately begins cleaning up the
4159 lov, just after starting the mds_lov process
4161 Severity : enhancement
4163 Description: Implement 'lfs df' to report actual free space on per-OST basis
4164 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4165 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4166 usage of objects; '-h' to report in human readable format.
4168 ------------------------------------------------------------------------------
4170 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4177 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4178 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4179 at startup during orphan unlinking. Instead, skip buddy update
4180 before setup, as it will be regenerated after recovery is complete.
4185 Description: If an OST is inactive, its locks might reference stale inodes.
4186 Details : lov_change_cbdata() must iterate over all namespaces, even if
4187 they are inactive to clear inode references from the lock.
4189 Severity : enhancement
4190 Frequency : occasional, if non-standard max_dirty_mb used
4192 Description: Client will block write RPCs if not enough grant
4193 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4194 then the client will block writes while waiting for another RPC
4195 to complete instead of consuming its dirty limit. With change
4196 we get improved performance when max_dirty_mb is small.
4198 Severity : enhancement
4199 Bugzilla : 3389, 6253
4200 Description: Add support for supplementary groups on the MDS.
4201 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4202 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4203 lookups for user supplementary groups into a cache.
4207 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4208 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4209 for liblustre O_EXCL is enforced by MDS.
4213 Description: readdir never returns NULL in liblustre.
4214 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4215 dentry in struct dirent.
4219 Frequency : liblustre only, depends on application IO pattern
4220 Description: liblustre clients evicted if not contacting servers
4221 Details : Don't put liblustre clients into the ping_evictor list, so
4222 they will not be evicted by the pinger ever.
4224 Severity : enhancement
4226 Description: Add ability to evict clients by NID from MDS.
4227 Details : By echoing "nid:$NID" string into
4228 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4229 $NID would be instantly evicted from this MDS and from all active
4230 OSTs connected to it.
4234 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4235 Details : lookup_it_finish() used to query file size from OSTs that was not
4240 Description: service threads change working directory to that of init
4241 Details : Starting lustre service threads may pin the working directory
4242 of the parent thread, making that filesystem busy. Threads
4243 now change to the working directory of init to avoid this.
4247 Frequency : during shutdown only
4248 Description: shutdown with a failed MDS or OST can cause unmount to hang
4249 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4254 Frequency : 2.6 only
4255 Description: chmod/chown may include an extra supplementary group
4256 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4257 supplementary group and if none is specified this is used.
4260 Bugzilla : 5479 (6816)
4262 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4263 Details : If lookup is in progress on a file that is unlinked we might try
4264 to revalidate the inode and fail in revalidate after lookup is
4265 complete and ll_file_open() enqueues the open again but
4266 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4269 Frequency : always, if lconf --abort_recovery used
4271 Description: lconf --abort_recovery fails with 'Operation not supported'
4272 Details : lconf was attempting to abort recovery on the MDT device and not
4275 ------------------------------------------------------------------------------
4277 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4282 Frequency : rare (only unsupported configurations with a node running as an
4284 Bugzilla : 6514, 5137
4285 Description: Mounting a Lustre file system on a node running as an OST could
4287 Details : OSTs now preallocates memory needed to write out data at
4288 startup, instead of when needed, to avoid having to
4289 allocate memory in possibly low memory situations.
4290 Specifically, if the file system is mounted on on OST,
4291 memory pressure could force it to try to write out data,
4292 which it needed to allocate memory to do. Due to the low
4293 memory, it would be unable to do so and the node would
4294 become unresponsive.
4296 Severity : enhancement
4298 Description: Addition of lconf --service command line option
4299 Details : lconf now accepts a '--service <arg>' option, which is
4300 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4302 Severity : enhancement
4304 Description: Failover mode is now the default for OSTs.
4305 Details : By default, OSTs will now run in failover mode. To return to
4306 the old behaviour, add '--failout' to the lmc line for OSTs.
4308 Severity : enhancement
4310 Description: Health checks are now provided for MDS and OSTs
4311 Details : Additional detailed health check information on MSD and OSTs
4312 is now provided through the procfs health_check value.
4315 Frequency : occasional, depends on IO load
4317 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4318 after numerous create/delete cycles
4319 Details : The ext3 inode allocation policy would not allocate new inodes
4320 very well on the OSTs because there are no new directories
4321 being created. Instead we look for groups with free space if
4322 the parent directories are nearly full.
4327 Description: Network or server problems during mount may cause partially
4328 mounted clients instead of returning an error.
4329 Details : The config llog parsing code may overwrite the error return
4330 code during mount error handling, returning success instead
4336 Description: MDS can fail to allocate large reply buffers
4337 Details : After long uptimes the MDS can fail to allocate large reply
4338 buffers (e.g. zconf client mount config records) due to memory
4339 fragmentation or consumption by the buffer cache. Preallocate
4340 some large reply buffers so that these replies can be sent even
4341 under memory pressure.
4345 Frequency : rare (liblustre)
4346 Description: fsx running with liblustre complained that using truncate() to
4347 extend the file doesn't work. This patch corrects that issue.
4348 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4349 ATTR_SIZE and lsm use in llu_setattr_raw.
4353 Frequency : rare, only 2.6 kernels
4354 Description: Unusual file access patterns on the MDS may result in inode
4355 data being lost in very rare circumstances.
4356 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4357 code in the 2.6 kernel caused the inode and/or EA data not to
4358 be read from disk, causing single-file corruption.
4362 Frequency : rare, only 2.6 filesystems using extents
4363 Description: Heavy concurrent write and delete load may cause data corruption.
4364 Details : It was possible under high-load situations to have an extent
4365 metadata block in the block device cache from a just-unlinked
4366 file overwrite a newly-allocated data block. We now unmap any
4367 metadata buffers that alias just-allocated data blocks.
4371 Frequency : filesystems with default stripe_count larger than 77
4372 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4373 Details : lconf specifies an inode size of 4096 bytes when the default
4374 stripe_count is larger than 77. This conflicts with the default
4375 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4376 this case to avoid pinning too much memory for large EAs.
4378 ------------------------------------------------------------------------------
4380 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4385 Frequency : rare (extremely heavy IO load with hundreds of clients)
4387 Description: Client is evicted, gets IO error writing to file
4388 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4389 caused clients to be evicted for AST timeouts. The fixes in
4390 bug 5192 mean we no longer need to have such short AST timeouts
4391 so ldlm_timeout has been increased.
4394 Frequency : occasional during --force or --failover shutdown under load
4395 Bugzilla : 5949, 4834
4396 Description: Server oops/LBUG if stopped with --force or --failover under load
4397 Details : a collection of import/export refcount and cleanup ordering
4398 issues fixed for safer force cleanup
4401 Frequency : only filesystems larger than 120 OSTs
4402 Bugzilla : 5990, 6223
4403 Description: lfs getstripe would oops on a very large filesystem
4404 Details : lov_getconfig used kfree on vmalloc'd memory
4407 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4408 Bugzilla : 6242, 6243
4409 Description: reading from files that had been truncated to a non-zero size
4410 but never opened returned no data
4411 Details : ll_file_read() reads zeros from no-object files to EOF
4416 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4417 Details : There's little harm in aborting MDS/OSS recovery and letting it
4418 try again, so I removed the LASSERT and return an error instead.
4420 Severity : enhancement
4422 Description: New debugging infrastructure for tracking down data corruption
4423 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4424 (b) cover more of the client-side code path, and (c) try to narrow
4425 down where problems occurred
4429 Bugzilla : 3819, 4364, 4397, 6313
4430 Description: Racing close and eviction MDS could cause assertion in mds_close
4431 Details : It was possible to get multiple mfd references during close and
4432 client eviction, leading to one thread referencing a freed mfd.
4434 Severity: : enhancement
4435 Bugzilla : 3262, 6359
4436 Description: Attempts to reconnect to servers are now more aggressive.
4437 Details : This builds on the enhanced upcall-less recovery that was added
4438 in 1.4.2. When trying to reconnect to servers, clients will
4439 now try each server in the failover group every 10 seconds. By
4440 default, clients would previously try one server every 25 seconds.
4445 Description: After recovery, certain operations trigger a failed
4446 assertion on a client.
4447 Details : Failing over an mds, using lconf -d --failover, while a
4448 client was doing a readdir() call would cause the client to
4449 LBUG after recovery completed and the readdir() was resent.
4451 Severity : enhancement
4453 Description: Default groups are now added by lconf
4454 Details : You can now run lconf --group <servicename> without having to
4455 manually add groups with lmc.
4458 Frequency : occasional
4460 Description: Nodes with an elan id of 0 trigger a failed assertion
4463 Frequency : always when accessing e.g. tty/console device nodes
4465 Description: tty and some other devices nodes cannot be used on lustre
4466 Details : file's private_data field is used by device data and lustre
4467 values in there got lost. New field was added to struct file to
4468 store fs-specific private data.
4471 Frequency : when exporting Lustre via NFS
4473 Description: NFSD failed occasionally when looking up a path component
4474 Details : NFSD is looking up ".." which was broken in ext3 directories
4475 that had grown large enough to become hashed.
4478 Frequency : Clusters with multiple interfaces not on the same subnet
4480 Description: Nodes will repeatedly try to reconnect to an interface which it
4481 cannot reach and report an error to the log.
4482 Details : Extra peer list entries will be created by lconf with some peers
4483 unreachable. lconf now validates the peer before adding it.
4486 Frequency : Only if a default stripe is set on the filesystem root.
4488 Description: Setting a default stripe on the filesystem root prevented the
4489 filesystem from being remounted.
4490 Details : The client was sending extra request flags in the root getattr
4491 request and did not allocate a reply buffer for the dir EA.
4494 Frequency : occasional, higher if lots of files are accessed by one client
4495 Bugzilla : 6159, 6097
4496 Description: Client trips assertion regarding lsm mismatch/magic
4497 Details : While revalidating inodes the VFS looks up inodes with ifind()
4498 and in rare cases can find an inode that is being freed.
4499 The ll_test_inode() code will free the lsm during ifind()
4500 when it finds an existing inode and then the VFS later attaches
4501 this free lsm to a new inode.
4505 Bugzilla : 6422, 7030
4506 Description: MDS deadlock between mkdir and client eviction
4507 Details : Creating a new file via mkdir or mknod (starting a transaction
4508 and getting the ns lock) can deadlock with client eviction
4509 (gets ns lock and trying to finish a synchronous transaction).
4512 Frequency : occasional
4513 Description: While starting a server, the fsfilt_ext3 module could not be
4515 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4516 kernels. Previously, lconf would still use the ext3 name
4517 when trying to load modules. Now, it will correctly use
4518 ext3 on 2.4 and ldiskfs on 2.6.
4520 Severity : enhancement
4521 Description: The default stripe count has been changed to 1
4522 Details : The interpretation of the default stripe count (0, to lfs
4523 or lmc) has been changed to mean striping across a single
4524 OST, rather than all available. For general usage we have
4525 found a stripe count of 1 or 2 works best.
4527 Severity : enhancement
4528 Description: Add support for compiling against Cray portals.
4529 Details : Conditional compiling for some areas that are different
4533 Frequency : occasional
4534 Bugzilla : 6409, 6834
4535 Description: Creating files with an explicit stripe count may lead to
4536 a failed assertion on the MDS
4537 Details : If some OSTs are full or unavailable, creating files may
4538 trigger a failed assertion on the MDS. Now, Lustre will
4539 try to use other servers or return an error to the
4543 Frequency : occasional
4545 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4546 caused lock timeout during readahead (since 1.4.2).
4547 Details : Processes doing readahead might match a lock that hasn't been
4548 granted yet if there are overlapping and conflicting lock
4549 requests. The readahead process waits on ungranted lock
4550 (original lock is CBPENDING), while OST waits for that process
4551 to cancel CBPENDING read lock and eventually evicts client.
4553 Severity : enhancement
4555 Description: Initial enabling of flock support for clients
4556 Details : Implements fcntl advisory locking and file status functions.
4557 This feature is provided as an optional mount flag (default
4558 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4559 locking are implemented yet, and those that are are not guaranteed
4560 to be completely correct in production environments.
4561 mount -t lustre -o [flock|noflock] ...
4564 Frequency : occasional
4566 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4567 assertion in the ext3 JBD (journaling) layer.
4568 Details : The b_committed_data struct is protected by the big kernel lock
4569 in 2.4 kernels, serializing journal_commit_transaction() and
4570 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4571 access to this struct. In 2.6 kernels there is finer grained
4572 locking to improve SMP performance of the JBD layer.
4576 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4581 Description: merge and cleanup kernel patches.
4584 -----------------------------------------------------------------------------
4586 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4588 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4589 versions of Lustre. Please update all servers and clients to
4590 version 1.4.2 or later at the same time. You must also run
4591 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4592 to update the configuration logs.
4594 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4595 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4596 - swab llog records of type '0' so we get proper header size/idx (5861)
4597 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4598 - fix rename of one directory over another leaking an inode (5953)
4599 - avoid SetPageDirty on 2.6 (5981)
4600 - don't re-add just-being-destroyed locks to the waiting list (5653)
4601 - when creating new directories, inherit the parent's custom
4602 striping settings if present parent (3048)
4603 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4604 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4605 - don't hold client locks on temporary worklist from l_lru (5666)
4606 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4607 - automatically evict dead clients (5921)
4608 - Update file size properly in create+truncate+fstat case (6196)
4609 - Do not unhash mountpoint dentries, do not allow removal of
4611 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4612 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4613 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4614 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4615 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4616 - Correctly handle reads of files with no objects (6243)
4617 - lctl recover will also mark a device active if deactivate used (5933)
4619 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4620 - allow --write-conf on an MDS with different nettype than client (5619)
4621 - don't write config llogs to MDS for mounts not from that MDS (5617)
4622 - lconf should create multiple TCP connections from a client (5201)
4623 - init scripts are now turned off by default; run chkconfig --on
4624 lustre and chkconfig --on lustrefs to use them
4625 - upcalls are no longer needed for clients to recover to failover
4627 - add --abort-recovery option to lconf to abort recovery on device
4629 - add support for an arbitrary number of OSTs (3026)
4630 - Quota support protocol changes.
4631 - forward compatibility changes to wire structs (6007)
4632 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4633 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4634 - improved "lustre status" script
4635 - initialize blocksize for non-regular files (6062)
4636 - added --disable-server and --disable-client configure options (5782)
4637 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4638 - Vanilla 2.4.29 support
4639 - increase maximum number of obd devices to 520 (6242)
4640 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4641 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4642 SLES 9 2.6.5 kernels
4643 - limit stripes per file to 160 (the maximum EA size) (6093)
4645 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4648 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4649 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4650 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4651 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4652 - small ext3 extents cleanups and fixes (5733)
4653 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4654 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4655 - avoid needless client->OST connect, fix handle mismatch (5317)
4656 - fix DLM error path that led to out-of-sync client, long delays (5779)
4657 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4658 - fix several locking issues related to i_size (5492,5624,5654,5672)
4659 - don't move pending lock onto export if it is already evicted (5683)
4660 - fix kernel oops when creating .foo in unlinked directory (5548)
4661 - fix deadlock in obdfilter statistics vs. object create (5811)
4662 - use time_{before,after} to avoid timer jiffies wrap (5882)
4663 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4664 - Do not leak request if server was not able to process it (5154)
4665 - If mds_open unable to find parent dir, make that negative lookup(5154)
4666 - don't create new directories with extent-mapping (5909, 5936)
4668 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4669 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4670 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4671 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4672 - support for RHEL4 kernel on i686 (5773)
4673 - provide error messages when incompatible logs are encountered (5898)
4675 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4676 * version 1.4.0.10 (1.4.1 release candidate 1)
4678 - don't keep a lock reference when lock is not granted (4238)
4679 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4680 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4681 - drop import inflight refcount on signal_completed_replay error (5255)
4682 - unlock page after async write error during send (3677)
4683 - handle missing objects in filter_preprw_read properly (5265)
4684 - no transno return for symlink open, don't save no-trasno open (3440)
4685 - don't try to complete elan receive that already failed (4012)
4686 - free RPC server reply state on error (5406)
4687 - clean up thread from ptlrpc_start_thread() on error (5160)
4688 - readahead could read extra page into cache that wasn't ejected (5388)
4689 - prevent races in class_attach/setup/cleanup/detach (5260)
4690 - don't dereference de->d_inode after l_dput of de (5458)
4691 - use "int" for stripe value returned from lock_to_stripe (5544)
4692 - mballoc allocation and error-checking fixes in 2.6 (5504)
4693 - block device patches to fix I/O request sizes in 2.6 (5482)
4694 - look up hostnames for IB nals (5602)
4695 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4696 - don't start multiple acceptors for the same port (5277)
4697 - fix incorrect LASSERT in mds_getattr_name (5635)
4698 - export a proc file for general "ping" checking (5628)
4699 - fix "lfs check" to not block when the MDS is down (5628)
4701 - service request history (4965)
4702 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4703 - create an "evict_client" /proc entry on OSTs, like the MDS has
4704 - fix mount usage message, return errors per mount(8) (5168)
4705 - change grep [] to grep "[]" in tests so they work in more UMLs
4706 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4707 - remove ancient LOV_MAGIC_V0 EA support (5047)
4708 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4709 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4710 - allow readahead allocations to fail when low on memory (5383)
4711 - mmap locking landed again, after considerable improvement (2828)
4712 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4714 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4717 - send OST transaction number in read/write reply to free req (4966)
4718 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4719 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4720 - the watchdog thread now runs as interruptible (5246)
4721 - flock/lockf fixes (but it's still disabled, pending 5135)
4722 - don't use EXT3 constants in llite code (5094)
4723 - memory shortage at startup could cause assertion (5176)
4725 - reorganization of lov code
4726 - single portals codebase
4728 - add extents/mballoc support (5025)
4729 - direct I/O reads in the obdfilter (4048)
4730 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4732 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4735 - send OST transaction number in read/write reply to free req (4966)
4736 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4737 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4738 - the watchdog thread now runs as interruptible (5246)
4739 - handle missing objects in filter_preprw_read properly (5265)
4740 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4741 - drop import inflight refcount on signal_completed_replay error (5255)
4742 - unlock page after async write error during send (3677)
4743 - return original error code on reconstructed replies (3761)
4744 - no transno return for symlink open, don't save no-trasno open (3440)
4746 - add pid to ldlm debugging output (4922)
4747 - bump the watchdog timeouts -- we can't handle 30sec yet
4748 - extra debugging for orphan dentry/inode bug (5259)
4750 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4753 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4754 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4755 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4756 - fix NULL dereference in /proc/sys/portals/routes (4827)
4757 - allow failed mdc_close() operations to be interrupted (4561)
4758 - stop precreate on OST before MDS would time out on it (4778)
4759 - don't send partial-page writes before EOF from client (4410)
4760 - discard client grant for sub-page writes on large-page clients (4520)
4761 - don't free dentries not owned by NFS code, check generation (4806)
4762 - fix lsm leak if mds_create_objects() fails (4801)
4763 - limit debug_daemon file size, always print CERROR messages (4789)
4764 - use transno after validating reply (3892)
4765 - process timed out requests if import state changes (3754)
4766 - update mtime on OST during writes, return in glimpse (4829)
4767 - add mkfsoptions to LDAP (4679)
4768 - use ->max_readahead method instead of zapping global ra (5039)
4769 - don't interrupt __l_wait_event() during strace
4771 - add software watchdogs to catch hung threads quickly (4941)
4772 - make lustrefs init script start after nfs is mounted
4773 - fix CWARN/ERROR duplication (4930)
4774 - return async write errors to application if possible (2248)
4775 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4776 - print NAL number in %x format (4645)
4777 - update barely-supported suse-2.4.21-171 series (4842)
4778 - support for sles 9 %post scripts
4779 - support for building 2.6 kernel-source packages
4780 - support for sles km_* packages
4782 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4785 - ignore -ENOENT errors in osc_destroy (3639)
4786 - notify osc create thread that OSC is being cleaned up (4600)
4787 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4788 - reconstruct ost_handle() like mds_handle() (4657)
4789 - create a new thread to do import eviction to avoid deadlock (3969)
4790 - let lconf resolve symlinked-to devices (4629)
4791 - don't unlink "objects" from directory with default EA (4554)
4792 - hold socknal file ref over connect in case target is down (4394)
4793 - allow more than 32000 subdirectories in a single directory (3244)
4794 - fix blocks count for O_DIRECT writes (3751)
4795 - OST returns ENOSPC from object create when no space left (4539)
4796 - don't send truncate RPC if file size isn't changing (4410)
4797 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4798 - bind to privileged port in socknal and tcpnal (3689)
4800 - rate limit CERROR/CWARN console message to avoid overload (4519)
4801 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4802 - basic mmap support (3918)
4803 - kernel patch series update from b1_4 (4711)
4805 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4808 - avoid crash during MDS cleanup with OST shut down (2775)
4809 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4810 - don't use bad inodes on the MDS (3744)
4811 - dynamic object preallocation to improve recovery speed (4236)
4812 - don't hold spinlock over lock dumping or change debug flags (4401)
4813 - don't zero obd_dev when it is force cleaned (3651)
4814 - print grants to console if they go negative (4431)
4815 - "lctl deactivate" will stop automatic recovery attempts (3406)
4816 - look for existing locks in ldlm_handle_enqueue() (3764)
4817 - don't resolve lock handle twice in recovery avoiding race (4401)
4818 - revalidate should check working dir is a directory (4134)
4820 - don't always mark "slow" obdfilter messages as errors (4418)
4822 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4825 - don't close LustreDB during write_conf until it is done (3860)
4826 - fix typo in lconf for_each_profile (3821)
4827 - allow dumping logs from multiple threads at one time (3820)
4828 - don't allow multiple threads in OSC recovery (3812)
4829 - fix debug_size parameters (3864)
4830 - fix mds_postrecov to initialize import for llog ctxt (3121)
4831 - replace config semaphore with spinlock (3306)
4832 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4833 - don't allow enqueue to complete on a destroyed export (3822)
4834 - down write_lock before checking llog header bitmap (3825)
4835 - recover from lock replay timeout (3764)
4836 - up llog sem before sending rpc (3652)
4837 - reduce ns lock hold times when setting kms (3267)
4838 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4839 - fix NULL deref and obd_dev leak on setup error (3312)
4840 - replace some LBUG about llog ops with error handling (3841)
4841 - don't match INVALID dentries from d_lookup and spin (3784)
4842 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4843 - fix invalid assertion in ptlrpc_set_wait (3880)
4845 - add libwrap support for the TCP acceptor (3996)
4846 - add /proc/sys/portals/routes for non-root route listing (3994)
4847 - allow setting MDS UUID in .xml (2580)
4848 - print the stack of a process that LBUGs (4228)
4850 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4853 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4854 - make sure to unset replay flag from failed open requests (3440)
4855 - if default stripe count is 0, use OST count for inode size (3636)
4856 - update parent mtime/ctime on client for create/unlink (2611)
4857 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4858 - free recovery state on server during a forced cleanup (3571)
4859 - unregister_reply for resent reqs (3063)
4860 - loop back devices mounting and status check on 2.6 (3563)
4861 - fix resource-creation race that can provoke i_size == 0 (3513)
4862 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4863 - more debugging for page-accounting assertion (3746)
4864 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4865 - avoid deadlock after precreation failure (3758)
4866 - fix race and lock order deadlock in orphan handling (3450, 3750)
4867 - add validity checks when grabbing inodes from l_ast_data (3599)
4869 - add /proc/.../recovery_status to obdfilter (3428)
4870 - lightweight CDEBUG infrastructure, debug daemon (3668)
4871 - change default OSC RPC parameters to be better on small clusters
4872 - turn off OST read cache for files smaller than 32MB
4873 - install man pages and include them in rpms (3100)
4874 - add new init script for (un)mounting lustre filesystems (2593)
4875 - run chkconfig in %post for init scripts (3701)
4876 - drop scimac NAL (unmaintained)
4878 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4881 - clean kiobufs before and after use (3485)
4882 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4883 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4884 - change init script to fail more clearly if not run as root (1528)
4885 - allow clients to reconnect during replay (1742)
4886 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4887 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4888 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4889 - fix cond_resched (3554)
4890 - only evict unfinished clients after recovery (3515)
4891 - allow bulk resend, prevent data loss (3570)
4892 - dynamic ptlrpc request buffer allocation (2102)
4893 - don't allow unlinking open directory if it isn't empty (2904)
4894 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4895 - remove extraneous obd dereference causing LASSERT failure (3334)
4896 - don't use get_cycles() when creating temp. files on the mds (3156)
4897 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4898 - handle EEXIST for set-stripe, set proper directory name (3336)
4900 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4901 - fix ksocknal_fmb_callback() error messages (2918)
4903 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4906 - don't copy lvb into (possibly NULL) reply on error (2983)
4907 - don't deref dentry after dput, don't free lvb on error (2922)
4908 - use the kms to determine writeback rpc length (2947)
4909 - increment oti_logcookies when osc is inactive (2948)
4910 - update client's i_blocks count via lvb messages (2543)
4911 - handle intent open/close of special files properly (1557)
4912 - mount MDS with errors=remount-ro, like obdfilter (2009)
4913 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4914 - don't use cancelling-locks' kms values (2947)
4915 - use highest lock extent for kms, not last one (2925)
4916 - don't dereference ERR_PTR() dentry in error handling path (3107)
4917 - fix thread race in portals_debug_dumplog() (3122)
4918 - create lprocfs device entries at setup instead of at attach (1519)
4919 - common AST error handler, don't evict client on completion race (3145)
4920 - zero nameidata in detach_mnt in 2.6 (3118)
4921 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4922 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4923 - initialize RPC timeout timer earlier for 2.6 (3219)
4924 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4925 - print nal/nid for unknown nid (3258)
4926 - additional checks for oscc recovery before doing precreate (3284)
4927 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4928 - don't crash in mdc_close for bad permissions on open (3285)
4929 - zero i_rdev for non-device files (3147)
4930 - clear page->private before handing to FS, better assertion (3119)
4931 - tune the read pipeline (3236)
4932 - fix incorrect decref of invalidated dentry (2350)
4933 - provide read-ahead stats and refine rpc in flight stats (3328)
4934 - don't hold journal transaction open across create RPC (3313)
4935 - update atime on MDS at close time (3265)
4936 - close LDAP connection when recovering to avoid server load (3315)
4937 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4938 - don't leak open file on MDS after open resend (3325)
4939 - serialize filter_precreate and filter_destroy_precreated (3329)
4940 - loop device shouldn't call sync_dev() for nul device (3092)
4941 - clear page cache after eviction (2766)
4942 - resynchronize MDS->OST in background (2824)
4943 - refuse to mount the same filesystem twice on same mountpoint (3394)
4944 - allow llmount to create routes for mounting behind routers (3320)
4945 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4946 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4947 - fix rare problem with rename on htree directories (3417)
4949 - allow default OST striping configuration per directory (1414)
4950 - fix compilation for qswnal for 2.6 kernels (3125)
4951 - increase maximum number of MDS request buffers for large systems
4952 - change liblustreapi to be useful for external progs like lfsck (3098)
4953 - increase local configuration timeout for slow disks (3353)
4954 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4956 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4959 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4960 - don't overwrite extent policy data in reply if lock was blocked (2901)
4961 - drop filter export grants atomically with removal from device (2663)
4962 - del obd_self_export from work_list in class_disconnect_exports (2908)
4963 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4964 - swab reply message in mdc_close, other PPC fixes (2464)
4965 - fix destroying of named logs (2325)
4966 - overwrite old logs when running lconf --write_conf (2264)
4967 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4968 - fix race in target_handle_connect (2898)
4969 - mds_reint_create() should take same inode create lock (2926)
4970 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4971 - don't close files for self_export to avoid uninitialized obd (2936)
4972 - allow MDS with the same name as client node (2939)
4973 - hold dentry reference for closed log files for unlink (2325)
4974 - reserve space for all logs during transactions (2059)
4975 - don't evict page beyond end of stripe extent (2925)
4976 - don't oops on a deleted current working directory (2399)
4977 - handle hard links to targets without a parent properly (2517)
4978 - don't dereference NULL lock when racing during eviction (2867)
4979 - don't grow lock extents when lots of conflicting locks (2919)
4981 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4984 - account for cache space usage on clients to avoid data loss (974)
4985 - lfsck support in lustre kernel code (2349)
4986 - reduce journal credits needed for BRW writes (2370)
4987 - orphan handling to avoid losing space on client/server crashes
4988 - ptlrpcd can be blocked, stopping ALL progress (2477)
4989 - use lock value blocks to assist in proper KMS, faster stat (1021)
4990 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4991 - recovery for initial connections (2355)
4992 - fixes for mds_cleanup_orphans (1934)
4993 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4994 - block all file creations until orphan recovery completes (1901)
4995 - client remove rq_connection from request struct (2423)
4996 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4997 - recovery timer race (2670)
4998 - mdc_close recovey bug (2532)
4999 - ptlrpc cleanup bug (2710)
5000 - mds timeout on local locks (2588)
5001 - namespace lock held during RPCs (2431)
5002 - handle interrupted sync write properly (2503)
5003 - don't try to handle a message that hasn't been replied to (2699)
5004 - client assert failure during cleanup after abort recovery (2701)
5005 - leak mdc device after failed mount (2712)
5006 - ptlrpc_check_set allows timedout requests to complete (2714)
5007 - wait for inflight reqs when ptlrpcd finishes (2710)
5008 - make sure unregistered services are removed from the srv_list
5009 - reset bulk XID's when resending them (caught by 1138 test)
5010 - unregister_bulk after timeout
5011 - fix lconf error (2694)
5012 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5013 - readahead locks pages, leaves pending causing memory pressure (2673)
5014 - increase OST request buffers to 4096 on large machines (2729)
5015 - fix up permission of existing directories in simple_mkdir (2661)
5016 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5017 - don't assign transno to errored transactions (2742)
5018 - don't delete objects on OST if given a bogus objid from MDS (2751)
5019 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5020 - if rq_no_resend, then timeout request after recovery (2432)
5021 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5022 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5023 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5024 - MDS recovery shouldn't delete objects that it has given out (2730)
5025 - if enqueue arrives after completion, don't clobber LVB (2819)
5026 - don't unlock pages twice when trigger_group_io returns error (2814)
5027 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5028 - don't write pages to disk if there was an error (1450)
5029 - don't ping imports that have recovery disabled (2676)
5030 - take buffered bytes into account when balancing socknal conn (2817)
5031 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5032 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5033 - remove little-endian swabbing of llog records (1987)
5034 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5035 - retry reposting request buffers if they fail (1191)
5036 - grow extent at grant time to avoid granting a revoked lock (2809)
5037 - lock revoke doesn't evict page if covered by a second lock (2765)
5038 - disable VM readahead to avoid reading outside lock extents (2805)
5040 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5041 - updated kernel patches for hp-2.4.20 kernel (2681)
5043 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5046 - fix truncated write corruption (2366)
5047 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5049 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5050 - ensure len(uuid) < 37 in lmc (1171)
5051 - fix ia64 OOPS in llog_test (2255)
5052 - zero end of page at obdfilter for partial page writes (2648)
5053 - don't leave stale dentries around after renames (bug 2428)
5054 - fix timeouts when evicting a client with a single lock held (2642)
5055 - set deadline for the initial HELLO message to drain (2634)
5056 - print out dotted-quad IP addresses in the socknal (2302)
5058 - additional debugging for MDS client eviction problem (2443)
5059 - fix mkfsoptions support for osts (2603, 2604)
5061 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5064 - add series for the vanilla 2.6.0 kernel
5065 - add series for the vanilla 2.4.24 kernel
5066 - add series for a cray x86/64 UL kernel drop
5067 - fix xattr patches for the vanilla 2.4.19 series
5069 - generate true UUIDs in lmc (1171)
5070 - have portals stack dumping break in UML (2466)
5071 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5072 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5073 - wake callback waiting threads on client eviction (2460)
5074 - Add --ptldebug and --subsystem to lmc (1719)
5075 - update assertion to allow safe interrupt allocation
5076 - set rq_no_resend for cancel requests (2432)
5077 - recalculate ptlrpcd timeout after resend (2494)
5078 - call vfs_rmdir when removing pending directories (2368)
5079 - fix renaming a file to itself (2429)
5080 - lmc creates a default one-stripe lov (2454)
5081 - expand procfs space to handle large clusters (2326)
5082 - increase UML stack to avoid overflow
5083 - update lconf's list of debug and subsystem masks
5084 - fix lfs find --obd (2510)
5085 - /proc tunable for disabling filter read caching (2591)
5086 - stop rpm packages from altering slapd.conf (2301)
5087 - disable nagle in the socknal under 0conf (2578)
5088 - choose mds inode size based on stripe count (2572)
5089 - fix kernel-source rpm problems (2516)
5091 - add --disable-doc to avoid pdf generation (2421)
5092 - update documentation, tests, type-os, comments
5093 - avoid format warnings on ia64
5094 - remove the TOE NAL
5095 - tiny code cleanups by removing unused fields
5097 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5100 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5101 - remove the most blatant lies from BUILDING file (2371)
5102 - change default debug level to reasonable production setting
5103 - reduce client side cache size to reduce cache flush time
5104 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5105 - make TCP zerocopy and pinger support enabled by default (2476)
5106 - sync writes completed after process exits caused crashes (2319)
5107 - maintain correct mount count on the MDS (2356)
5108 - backout 1557, because 2316 wasn't really fixed
5109 - better file I/O statistics gathering in /proc
5110 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5111 - another kernel patch to fix zero-copy TCP function export
5112 - don't take duplicate lock when processing re-sent getattr (2420)
5113 - lctl uses obd_self_export instead of creating new conn (2353)
5114 - MDS/OST recovery case which requires object creation asserted (2425)
5115 - move lfs from /usr/sbin to /usr/bin in packages
5116 - fix race between mds_client_add and mds_client_free (2417)
5117 - use kmalloc instead of slabs in portals (2430)
5118 - don't create duplicate records when a failover MDS is present (2442)
5119 - remove unnecessary mount age check (2332)
5120 - don't remove directory inodes from locks prematurely (2451)
5121 - don't break if MDS service name is the same as hostname (2103)
5122 - fix races in client write RPC generation when cache full (2482)
5124 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5127 - remove now-unused request->rq_obd (278)
5128 - if an allocation fails, print out how much memory we've used (1933)
5129 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5130 - add the "configurable stack size" patch to most series files (1256)
5131 - ability to write large log records, for 100+ OST configs (2306)
5132 - fix NULL deref when filter_prep fails (2314)
5133 - fix operator precedence error in filter_sync
5134 - dynamic allocation of socknal TX descriptors (2315)
5135 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5136 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5137 - ensure that CWARN messages in Portals always get to the syslog
5138 - __init/__exit are not for prototype decls (ldlm_init/exit)
5139 - x86-64 compile warning fixes
5140 - fix gateway LMC keyword conflict (2318)
5141 - fix MDS lock inversions in getattr/reint paths (1844)
5142 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5143 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5144 - DLM race condition prevented some lock evictions (2328)
5145 - ENOMEM detection and retry on socknal sends (2230)
5146 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5147 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5148 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5149 - better error messages when a client is rejected during recovery (1505)
5150 - avoid cancelling locks which were never granted, after failure (2330)
5151 - fix i_sem/journal inversion in mds_client_add (2333)
5152 - fix truncate/getattr lock cycle deadlock (2334)
5153 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5154 - fix two rare exit paths which could leak an l_lock() ref (2321)
5155 - fencepost error in MDS/OST orphan recovery (2226)
5156 - make log record alignment 8 bytes (1988)
5157 - lstripe now fails when requested offset > ost_count (2237)
5158 - ensure that all kernel series have a complete list.h (1607)
5159 - fix crashes in special-file operations (2316)
5160 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5162 - allow configurable automake binary, for testing new versions
5163 - small update to the lfs documentation
5165 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5167 * fix negative export reference count in fsfilt_sync (2312)
5169 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5170 * release candidate 0.9.1
5172 - orphans are moved into the PENDING directory for possible recovery
5173 - replayed opens now open by fid for orphan/rename safety (1042)
5174 - last close of an orphan inode generates a transno (683)
5175 - chdir() and mount() now pin the directory entry (1020)
5176 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5177 - discard very old requests without processing them (1502)
5178 - remove some common, well-understood CERRORs (1505)
5179 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5180 - clear "grant" flags in OST replies until OST grant code lands (1644)
5181 - fix read performance by not clobbering i_blksize on client (1598)
5182 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5183 - make LRU size a /proc tunable, clears locks when reduced (707)
5184 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5185 - server locks take references on exports now (1558)
5186 - build fixes for 2.4.20-rh trees (1663)
5187 - return an error from lov_create if all OSCs are inactive (1751)
5188 - fix import levels when a reconnect happens without a timeout (1597)
5189 - exit early from mds_open if we get a lookup error (1749)
5190 - partial page read at EOF wouldn't wait for disk before sending (1642)
5191 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5192 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5193 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5194 - write/truncate lock inversion (1639)
5195 - Don't auto-load obdclass, portals modules during cleanup (1495)
5196 - fix timestamps from jumping to "now" (1763)
5197 - extra journal assertions (1648)
5198 - add an extra multiunlink test (1771)
5199 - fix read_record/write_record API (1776)
5200 - fix leak of offset_extent, possible incorrect i_size later (1772)
5201 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5202 - Debugging for the kqswnal_get_idle_tx problems (1820)
5203 - Allow recovery to be attempted multiple times (1536)
5204 - Write out MDS last_rcvd file after it is first created (1600)
5205 - Fix tx_descriptor leak in failed transmit situations (1827)
5206 - ext3 journaling fixes for assertion failure after IO error (1871)
5207 - class_export_put() on freed export after completion AST error (1896)
5208 - Fix revalidate looping in VFS (1322)
5209 - Don't access a freed export during MDS_REINT timeout (1521)
5210 - Add open-unlink recovery support on the MDS (1673,1764)
5211 - Return an error if no MDS data was read from last_rcvd (1946)
5212 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5213 - Don't setup a disk device that doesn't match exported UUID (317)
5214 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5215 - avoid committing NULL handle in force close
5216 - local.sh is now a one-stripe LOV configuration
5217 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5218 - fix ext3 htree duplicate directory entry corruption (1516)
5219 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5220 - update to new LOV EA format (2097)
5221 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5223 2003-06-15 Phil Schwan <phil@clusterfs.com>
5226 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5227 - per-import/export recovery handling (958, 931, 959)
5228 - multiple last-rcvd slots, for serving multiple FSes (949)
5229 - connections are again shared between multiple imp/exports (963, 964)
5230 - "umount -f" would hang if any requests needed to be sent (393, 978)
5231 - avoid pinning large req buffer by copying for queued messages (989)
5232 - add "uuid" to "lctl device" command to help upcalls (991)
5233 - "open" RPCs with transnos would confuse recovery counters (1037)
5234 - do proper endian conversion of all wire messages (288, 340, 891)
5235 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5236 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5237 - fix ptlrpc_connection leak in target_handle_connect (1174)
5238 - fix import refcounting bug in OST and MDS cleanup (1134)
5239 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5240 - fix very unlikely obd_types race condition (501)
5241 - remove osc_open hack for echo_client (1187)
5242 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5243 - a failure in read_inode2 leads to deadlock (1139)
5244 - cancel ack-locks as soon as transaction is committed (1072)
5245 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5246 - make sure to commitrw after any preprw to avoid deadlock (1162)
5247 - failing to execute a file in a lustre FS would lock inode (1203)
5248 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5249 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5250 - obd_uuid2tgt would walk off the end of the list (1255)
5251 - on IA64 the osc would give portals incorrect bulk size (1258)
5252 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5253 - fix lock inversion caused by new llite matching code (1282)
5254 - limit the number of dirty pages on a client to 10MB (1286)
5255 - timed out locks were not being corrected cancelled (1289)
5256 - fix O_DIRECT above 4GB on IA-32 (1292)
5257 * major user-visible changes
5258 - fail out/fail over policy now controlled by the upcall (993)
5260 - add OBD_PING to check server availability and failure (954)
5261 - lustre messages are now sent in sending host order (288, 340, 891)
5262 - add eadatalen to MDS getattr reply (340)
5263 - OST read replies may contain second buffer, with per-page status (593)
5265 2003-03-11 Phil Schwan <phil@clusterfs.com>
5268 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5269 - failed open()s could cause deadlock; fixed (867, 869)
5270 - stop cancelling OST locks when files are closed (481)
5271 - overlapping XID spaces caused network corruption (851, 853)
5272 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5273 - setattr_raw added, to do single-RPC, server-side setattrs
5274 - lmc/lconf syntax change for OST UUIDs
5275 - fix crashy race condition between ptlrpc_free_req and osc_close
5276 - don't use request in mdc_enqueue if we hit a timeout (889)
5277 - don't set the inode i_size for regular files from the MDS (896)
5278 - handle out of order completion AST (842)
5279 - don't LBUG if a lock request times out after receiving AST (913)
5280 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5281 - if a bad lock AST arrives, send an error instead of dropping entirely
5282 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5283 - fix leak in bulk IO when only partially completed (899, 900, 926)
5284 - fix O_DIRECT for ia64 (55)
5285 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5286 - C-z after timeout could hang a process forever; fixed (977)
5288 - client-side I/O cache (678, 924, 929, 941, 970)
5290 - READPAGE and SETATTRs which don't take server-side locks get
5293 2003-02-11 Phil Schwan <phil@clusterfs.com>
5296 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5297 - Fix fsfilt_extN_readpage() to read a full page of directory
5298 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5299 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5300 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5301 - Most intent operations were removed, in favour of a new RPC mode
5302 that does a single RPC to the server and bypasses most of the VFS
5303 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5304 - Aggressively cancel local locks on DLM servers
5305 - mds_reint_unlink sends EA to the client if it's the last nlink.
5306 client uses that EA to unlink OST objects.
5307 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5308 - recursive symlinks were fixed (439)
5309 - fixed NULL deref in DEBUG_REQ
5310 - filter_update_lastobjid no longer calls sync, which annoyed extN
5311 - fixed multi-client small-writes to a single file problem (445)
5312 - fixed mtime updates during file writes (607)
5313 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5314 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5315 - fixed Linux OST size reporting problem (444, 656)
5316 - OST now updates object mtime with writes or setattr (607, 619)
5317 - client verifies file size before zeroing page past EOF (445)
5318 - OST now writes last allocated objid to disk with allocation (108)
5319 - LOV on echo now works (409)
5321 - mds_reint_unlink sends a new buffer, with the EA included. this
5322 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5323 set if a regular file was being unlinked, and it was the last link
5324 - use PtlGet from the target for bulk writes (315)
5325 - OST now updates object mtime with writes or setattr (607, 619)
5326 - LDLM now has a grant-time callback to revalidate locked items, if
5328 - Many MDS operations were reorganized to combat race conditions
5330 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5331 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5333 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5336 - Fully reactivate OST imports after reconnection (512, others)
5337 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5338 - More graceful error handling for truncating on dead OST (515)
5339 - Don't error out unless we're actually accessing dead stripes (474)
5340 - Fix garbage sizes when stripes are missing (410)
5341 - LRU counters were broken, causing constant lock purge (433, 432)
5342 - garbage on read from stripes with failed OSTs (441)
5343 - mark OSCs as active before reconnecting during recovery (438)
5344 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5345 - lfind did not preserve OST order in output (443)
5346 - symlinks cause hung clients, incorrect data (439)
5347 - stop dereferencing request after dropping refcount (457)
5348 - don't LASSERT(spin_is_locked) on non-SMP (455)
5349 - fixes for many rename() bugs
5350 - fstat didn't correctly synchronize attributes (399)
5351 - server must handle lock cancellation during blocking AST prep (487)
5352 - bulk descriptors were free()d too soon (511)
5353 - fix paths in lconf, which would load incorrect modules (451, 507)
5354 - fix confusing lconf 'host not found' error message (386)
5355 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5356 - fix race condition in mdc_blocking_ast() for inode access (526)
5357 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5358 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5359 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5360 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5361 - fix lstripe in several ways (kernel side) (527)
5362 - fix request leak in ldlm_cli_enqueue (262)
5363 - incorrect OSC was marked inactive after OST failure
5364 - call mds_fs_cleanup before unmounting filesystem (524)
5365 - fix races between taking ns_lock and ldlm_lock_change_resource
5366 - fix races updating LOV export open file list
5367 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5368 - fix recovery NULL deref in ldlm_cli_cancel_unused
5369 - fix some DLM races by using new hash table for lock handles (419)
5370 - permit the client to specify desired inodes, at replay
5371 - duplicate requests when we queue them for replay reintegration
5372 - fix last_rcvd offset calculation
5373 - sync after each recovered transaction, so we always make progress
5374 - never, not always, ERESTART requests without transnos
5375 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5377 - skip replay if the MDS says that the client is already connected
5378 - don't check for a recovery-enabled export to match lctl's UUID
5379 - don't INC_USE_COUNT for phantom exports
5380 - don't crash when cleaning up phantom exports (567)
5381 - don't double-finish or set replay data for errored mdc_open requests
5382 - abort requests when they time out, so we don't get old replies
5383 - send/receive replies for AST messages again
5384 - if the client says that it doesn't have the lock, cancel it on the
5386 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5387 mark it as destroyed, it will all work out when decref is called
5388 - fix module use counts (22, 581)
5390 - ASTs now expect a reply (server cancels lock on error reply)
5392 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5395 - fix many simultaneous client startup (392)
5396 - fix dentry->d_it clobbering
5397 - credentials weren't being shipped for readdir/getattr operations
5398 - remove invalid assertions triggered during some concurrent MD
5400 - proper Lustre versions added (336, 389)
5401 - fix memory leak for create error case (398)
5402 - fix LOV locking bug that would get cli/srv out of sync
5403 - fix echo client over LOV (409)
5404 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5405 - fix double-O_EXCL intent crash (424)
5406 - avoid sending multiple lock CANCELs (352)
5408 - MDS can do multi-client recovery (modulo bugs in new code)
5410 - many updates, edits, cleanups
5412 2002-11-18 Phil Schwan <phil@clusterfs.com>
5415 - fix null d_it dereference (346)
5416 - fix full OST/dbench hang (333)
5417 - fix permission problem with file removal (286)
5418 - fix removal of OSCs from LOV when they fail
5419 - fix NULL deref during bulk timeout (214)
5420 - fix problems related to multiple filesystems on one MDS (241)
5421 - fixed serious subtle metadata locking bugs
5422 - free locks on clients when inodes are removed due to memory
5424 - fix inode pointer in lock data (285)
5425 - partial support for multiple MDS on a single host (241)
5426 - data locks weren't cancelled at clear_inode time (290, 311)
5427 - intent locks could lead to unbounded lock growth (205)
5428 - added a maximum lock count, an LRU list, and a flusher
5429 - fix multiple rename (365)
5430 - properly abstracted the echo client
5431 - OSC locked 1 byte too many; fixed
5432 - rewrote brw callback code:
5433 - fixed recovery bugs related to LOVs (306)
5434 - fixed too-many-pages-in-one-write crash (191)
5435 - fixed (again) crash in sync_io_timeout (214)
5436 - probably fixed callback-related race (385)
5438 - Add capability to MDS protocol
5439 - LDLM cancellations and callbacks on different portals
5441 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5444 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5445 - defer open object destruction to close time (601981, 138)
5446 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5447 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5448 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5449 - don't call obd_setattr in ll_file_release for destroyed objects
5450 * protocol change to lustre_msg: move |version| and add |flags|
5451 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5452 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5453 * added replay of create, unlink, link and rename operations during
5454 MDS failover; recovery should be much more robust now
5455 * remove failed OSCs from LOVs (only lov_create uses this so far)
5456 * the lustre-HOWTO was brought (more) up to date (582544)
5458 2002-10-23 Phil Schwan <phil@clusterfs.com>
5461 - in-use dentries weren't being reused properly (617851)
5462 - prevent multiple LDLM setup (599178)
5463 - fix LOV size calculations for truncate (617853)
5464 - fix client handling of MDS intent errors (POSIX)
5465 - fix permission bug in lovstripe.c test (624321)
5466 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5467 - truncate past end of file could corrupt data
5468 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5469 - a race in recovery could return ETIMEDOUT to apps (623947)
5470 - building outside the source directory was fixed
5471 * the lustre-HOWTO was brought (more) up to date (582544)
5472 * major progress was made on recovery functionality
5474 2002-10-10 Phil Schwan <phil@clusterfs.com>
5477 - recovery deadlock fix
5478 - rm -rf causes LBUG fix (617817)
5479 - file open by multiple tasks fix (618962)
5480 - directory permissions bugs (602707 and 620007)
5481 - journal_stop fixed with locking (611313)
5482 - O_APPEND failures resolved (618273, perhaps 614459)
5483 - lconf PATH fix (619770)
5484 - IA64 build fix (621450)
5485 - RPC buffer sizes scale with amount of memory
5487 2002-10-01 Phil Schwan <phil@clusterfs.com>
5490 - locks would be cancelled without throwing away data pages,
5491 resulting in inconsistent data (605627)
5492 - inode attributes were not always being refreshed (605627, 612449)
5493 - lconf now continues to cleanup after lctl reports an error
5494 - MDS now enforces user permissions (602707)
5495 - lprocfs cleanup fixed, but not yet enabled (614157)
5496 - fixed infinite server hang, should a client not respond to an AST
5497 - avoid going into recovery if user calls readlink() with a buffer
5498 that's too small (613941)
5499 - AST RPCs no longer require replies (614867) -- this may be changed
5500 - don't crash server if client sends an IOV that's too big (611336)
5501 - fixed lock conversion deadlock (611892)
5502 - fixed the following of symlinks (614622)
5503 * recovery: the server can remove locks from a client that dies, other
5504 clients can make progress
5505 * more extN patch fixes
5506 * compile-time configurable ptlrpc buffer allocations
5508 - collaborative read cache document
5509 - Lustre Lite Performance CDR document-in-progress
5511 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5514 - fix typo in patch-2.4.18
5516 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5519 - clear ptlrpc request each time in handle_incoming_request()
5520 - unlink of files now destroys the object on the OST
5522 2002-09-19 Peter Braam <braam@clusterfs.com>
5524 * add hard link support
5525 * change obdfile creation method
5526 * kernel patch changed
5528 2002-09-19 Peter Braam <braam@clusterfs.com>
5531 - stack overflow bug in extN fixed
5533 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5535 * documentation updates
5536 - add man pages for config tools
5537 - update tests/README to describe testing with new config tools
5538 - finish metadata API descriptions
5539 * bug fixes and cleanups
5540 - statfs workaround for 16TB limit
5541 - LOV stripe allocation improved, can stripe on subset of OSTs
5542 - LOV file size/IO offset was wrong for files > 4GB in size
5543 - object EA data was being dropped, caused files to be unreadable
5544 - memory overflow with non-LOV OST caused memory corruption
5545 - fixed regression tests to work with new config tools, obdfilter
5546 - fixed bug when directory size became larger than 1 block
5547 - fixed bug (for single client case) when PWD was deleted
5548 - invalidate local directory pages when doing intent-based ops
5549 - avoid LDLM oops when lock callback contained bad data
5551 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5553 * documentation updates
5554 * bug fixes and cleanups
5555 - configuration tools
5558 - 64-bit compile warnings
5559 - 64-bit internal statfs data
5561 * test_brw on persistent OST devices
5563 * lprocfs (disabled)
5565 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5567 * documentation updates
5568 * bug fixes and cleanups
5569 * configuration tools
5571 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5574 * many small fixes to 0_5_4
5575 * io/network handling
5576 * thinkos in MDS operations
5578 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5581 * crucial basic fixes to 0.5.3
5582 * IOR, Iozone work over Elan
5585 2002-08-07 Phil Schwan <phil@clusterfs.com>
5586 * version 0_5_3, our first alpha
5587 * we use the new Portals iovs
5588 * documentation updates
5589 * bug fixes and cleanups
5590 * small changes in the DLM wire protocol
5592 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5593 * version 0_5_1 with some initial stability,
5594 * locking on MD and file I/O.
5595 * documentation updates
5596 * several bug fixes since 0.5.0
5597 * small changes in wire protocol
5599 2002-07-18 Phil Schwan <phil@clusterfs.com>
5601 * delivered as Lustre Light Alpha
5602 * fixed a crash after handling invalid MDS requests
5603 * fixed directory pages for architectures with non-4k pages sizes
5605 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5606 * release version v0_4_4
5607 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5609 * Fixes a number of interruption problems with OST operations.
5610 * Update documentation for portals header changes
5611 * Move all wire protocol structs/defines to lustre_idl.h
5612 * Fixes symlink length bug.
5613 * Add tcpdump to repository.
5615 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5616 * release version v0_4_3
5617 * Fixes statfs for inodes on extN.
5618 * Fixes bug in runtests which would delete /etc/hosts.
5619 * Use 64-bit object IDs wherever possible (not into VFS though)
5620 Remove ost_get_info, which is unused by lustre, and out of date.
5622 2002-07-03 Peter Braam <braam@clusterfs.com>
5623 * release version v0_4_2 Fixes a lookup error (type not passed)
5624 * move forward to head of Portals
5625 * move forward to latest Lustre kernel
5627 2002-06-25 Peter Braam <braam@clusterfs.com>
5628 * release version v0_4_1. Hopefully stable on single node use.