1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.27 (SLES 10),
5 2.6.18-92.1.6.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
15 Severity : enhancement
17 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
19 Severity : enhancement
21 Description: Upcall on Lustre log has been dumped
22 Details : Allow for a user mode script to be called once a Lustre log has
23 been dumped. It passes the filename of the dumped log to the
24 script, the location of the script can be specified via
25 /proc/sys/lnet/debug_log_upcall.
30 Description: avoid messages about idr_remove called for id which is not allocated.
31 Details : Move assigment s_dev for clustered nfs to end of initialization, for
32 avoid problem with error handling.
37 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
38 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
39 CONN_USED_HASH and this prodice warning when put connection again
42 Severity : enhancement
44 Description: Export bytes_read/bytes_write count on OSC/OST.
48 Description: Early reply size mismatch, MGC loses connection
49 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
50 the connect flags are properly negotiated.
54 Description: Properly propagate oinfo flags from lov to osc for statfs
55 Details : restore missing copy oi_flags to lov requests.
57 Severity : enhancement
59 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
60 : l_getgroups(8), lst(8), routerstat(8)
61 Details : included man pages for llobdstat(8), llstat(8),
62 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
64 Severity : enhancement
66 Description: configure's --enable-quota should check the
67 : kernel .config for CONFIG_QUOTA
68 Details : configure is terminated if --enable-quota is passed but
69 : no quota support is in kernel
73 Description: Remove portals compatibility
74 Details : Remove portals compatibility, not interoperable with releases
79 Description: Resolve device initialization race
80 Details : Prevent proc handler from accessing devices added to the
81 obd_devs array but yet be intialized.
83 Severity : enhancement
85 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
87 Severity : enhancement
89 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
94 Description: Using wrong pointer in osc_brw_prep_request
95 Details : Access to array[-1] can produce panic if kernel compiled with
96 CONFIG_PAGE_ALLOC enabled
100 Description: Client runs out of low memory
101 Details : Consider only lowmem when counting initial number of llap pages
105 Description: *optional* service tags registration
106 Details : if the "service tags" package is installed on a Lustre node
107 When the filesystem is mounted, a local-node service tag will
108 be created. See http://inventory.sun.com/ for more information
109 about the Service Tags asset management system.
113 Description: Kernel BUG tries to release flock
114 Details : Lustre does not destroy flock lock before last reference goes
115 away. So always drop flock locks when client is evicted and
116 perform unlock regardless of successfulness of speaking to MDS.
120 Description: add recount protection for osc callbacks, so avoid panic on shutdown
124 Description: sanity test 65a fails if stripecount of -1 is set
125 Details : handle -1 striping on filesystem in ll_dirstripe_verify
130 Description: ASSERTION(CheckWriteback(page,cmd)) failed
131 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
134 Severity : enhancement
136 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
140 Description: do not process already freed flock
141 Details : flock can possibly be freed by another thread before it reaches
142 to ldlm_flock_completion_ast.
146 Description: LBUG during stress test
147 Details : Need properly lock accesses the flock deadlock detection list.
151 Description: oops in page fault handler
152 Details : kernel page fault handler can return two special 'pages' in error case, don't
153 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
157 Description: timeout with invalidate import.
158 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
159 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
160 ptlrpc_invalidate_import and as result LASSERT.
162 Severity : enhancement
164 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
168 Description: enable MGS and MDT services start separately
169 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
170 a co-located MGS without starting the MGS, which is a complement
171 to 'nosvc' mount option.
175 Frequency : after recovery
176 Description: precreate to many object's after del orphan.
177 Details : del orphan st in oscc last_id == next_id and this triger growing
178 count of precreated objects. Set flag LOW to skip increase count
179 of precreated objects.
183 Frequency : rare, on clear nid stats
184 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
185 Details : when clean nid stats sometimes try destroy live entry,
186 and this produce panic in free.
190 Description: Stack overflow during MDS log replay
191 ease stack pressure by using a thread dealing llog_process.
195 Description: wait until IO finished before start new when do lock cancel.
196 Details : VM protocol want old IO finished before start new, in this case
197 need wait until PG_writeback is cleared until check dirty flag and
198 call writepages in lock cancel callback.
200 Severity : enhancement
202 Description: using special macro for print time and cleanup in includes.
206 Description: mds_mfd_close() ASSERTION(rc == 0)
207 Details : In mds_mfd_close(), we need protect inode's writecount change
208 within its orphan write semaphore to prevent possible races.
212 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
216 Frequency : rare, on shutdown ost
217 Description: don't hit live lock with umount ost.
218 Details : shrink_dcache_parent can be in long loop with destroy dentries,
219 use shrink_dcache_sb instead.
223 Description: don't panic with use echo client
224 Details : echo client pass NULL as client nid pointer and this produce null
229 Description: fix build on ppc32
230 Details : compile code with -m64 flag produce wrong object file for ppc32.
234 Description: add message levels for liblustreapi
238 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
239 Details : Move the 'good_osts' check before the 'total_bavail' check. This
240 will result in an -EAGAIN and in the exit call path we call
241 alloc_rr() which will with increasing aggressiveness attempt to
242 aquire precreated objects on the minimum number of required OSCs.
246 Description: Use old size assignment to avoid deadlock
247 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
248 the scheduling while holding a spinlock. We do not need locking
249 for size in ll_update_inode() because size is only updated from
250 the MDS for directories or files without objects, so there is no
251 other place to do the update, and concurrent access to such inodes
252 are protected by the inode lock.
256 Description: resolve "_IOWR redefined" build error on SLES10
260 Description: dump the memory debugging after all modules are unloaded to
261 suppress false negative in conf_sanity test 39
263 Severity : enhancement
265 Description: build kernel-ib packages for OFED 1.3 in our release cycle
270 Description: fix SLES kernel versioning
271 Details : the kernel version for our SLES 10 kernel did not include a "-"
272 before the "smp" at the end. while this was not a problem in
273 general, it did mean that software trying to use the kernel
274 version to try to detect a vendor specific kernel would fail.
275 this was most evident by the OFED build scripts.
279 Description: Don't update lov_desc members until making sure they are valid
280 Details : When updating lov_desc members via proc fs, need fix their
281 validities before doing the real update.
285 Description: don't put request into delay list while invalidate in flight.
286 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
287 import in flight. this produce timeout for invalidate and sometimes
288 can cause stale data.
292 Frequency : on ppc only
293 Description: not convert ost objects for directory because it's not exist.
294 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
298 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
300 Severity : enhancement
302 Description: Update to support 2.6.22.14 vanilla kernel.
306 Frequency : rare, on recovery
307 Description: read procfs can produce deadlock in some situation
308 Details : Holding lprocfs lock which send rpc can produce block for destroy
309 obd objects and this also block reconnect with -EALREADY. This isn't
310 fix all lprocfs bugs - but make it rare.
312 Severity : enhancement
314 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
317 Frequency : frequent on X2 node
319 Description: mdc_set_open_replay_data LBUG
320 Details : Set replay data for requests that are eligible for replay.
324 Description: lustre_mgs: operation 101 on unconnected MGS
325 Details : When MGC is disconnected from MGS long enough, MGS will evict the
326 MGC, and late on MGC cannot successfully connect to MGS and a lot
327 of the error messages complaining that MGS is not connected.
331 Frequency : on network error
332 Description: panic with double free request if network error
333 Details : mdc_finish_enqueue is finish request if any network error ocuring,
334 but it's true only for synchronus enqueue, for async enqueue
335 (via ptlrpcd) this incorrect and ptlrpcd want finish request
338 Severity : enhancement
340 Description: client-side metadata stat-ahead during readdir(directory readahead)
341 Details : perform client-side metadata stat-ahead when the client detects
342 readdir and sequential stat of dir entries therein
345 Frequency : on start mds
347 Description: Implement get_info(last_id) in obdfilter.
350 Frequency : occasional
352 Description: Correctly check stale fid, not start epoch if ost not support SOM
353 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
354 overwrited with new generated fid, but mds can anwer with one of these
355 two fids and both is not stale. setattr incorectly start epoch and
356 assume will be called done_writeting, but without SOM done_writing
360 Frequency : rare, depends on device drivers and load
362 Description: MDS or OSS nodes crash due to stack overflow
363 Details : Code changes in 1.8.0 increased the stack usage of some functions.
364 In some cases, in conjunction with device drivers that use a lot
365 of stack the MDS (or possibly OSS) service threads could overflow
366 the stack. One change which was identified to consume additional
367 stack has been reworked to avoid the extra stack usage.
370 Frequency : occasional
372 Description: Do not fail import if osc_interpret_create gets -EAGAIN
373 Details : If osc_interpret_create got -EAGAIN it immediately exits and
374 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
375 oscc_has_objects and see OSC has no objests and call
376 oscc_internal_create to resend create request.
378 Severity : enhancement
380 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
382 Severity : enhancement
384 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
387 Frequency : very rare
389 Description: Fix replay if there is an un-replied request and open
390 Details : In some cases, older replay request will revert the
391 mcd->mcd_last_xid on MDS which is used to record the client's
394 Severity : enhancement
396 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
398 Severity : enhancement
400 Description: Add rhel5 support to HEAD.
402 Serverity : enhancement
404 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
409 Description: Don't allow skipping OSTs if index has been specified.
410 Details : Don't allow skipping OSTs if index has been specified, make locking
411 in internal create lots better.
415 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
416 Details : access to bitfield in structure is always rounded to long
417 and this produce problem with not atomic change any bit.
421 Description: Lustre make rpms failed.
422 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
423 builds Lustre rpms from tarball.
426 Frequency : rare on shutdown ost
428 Description: If llog cancel was not send before clean_exports phase, this can
429 produce deadlock in llog code.
430 Details : If llog thread has last reference to obd and call class_import_put
431 this produce deadlock because llog_cleanup_commit_master wait when
432 last llog_commit_thread exited, but this never success because was
433 called from llog_commit_thread.
437 Description: allow userland application know is lost one of stripes.
438 Details : fill lvb_blocks with error code on ost and return it to
439 application if error flag found.
443 Description: NULL lov_tgts causing MDS oops
444 Details : more safe checks for NULL lov_tgts for avoid oops.
446 Severity : enhancement
448 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
452 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
453 Details : If an OST is down the MDS will hang indefinitely in
454 obd_statfs_rqset() waiting for the statfs data. While for
455 MDS QOS usage of statfs, it should not stuck in waiting.
457 Severity : enhancement
459 Description: remote_acl support
460 Details : Support ACL-based permission check for remote user.
461 Support setfacl/getfacl for remote user with the utils
462 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
463 the system "{s,g}etfacl" utils.
465 Severity : enhancement
467 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
469 Severity : enhancement
471 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
475 Description: make mgs_setparam() handle fsname containing dash
476 Details : fsname containing a dash does not work with lctl conf_param
478 Severity : enhancement
480 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
482 Severity : enhancement
484 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
488 Frequency : rare, at shutdown
489 Description: access already free / zero obd_namespace.
490 Details : if client_disconnect_export was called without force flag set,
491 and exist connect request in flight, this can produce access to
492 NULL pointer (or already free pointer) when connect_interpret
493 store ocd flags in obd_namespace.
497 Frequency : only at startup
498 Description: not alloc memory with spinlock held.
499 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
500 if any spinlock held.
502 Severity : enhancement
504 Description: make lustre randomly fail allocating memory
505 Details : Make lustre randomly failed allocating memory for testing purpose.
507 Severity : enhancement
509 Description: lost problems with lov objid file
510 Details : Fixes some scability and access to not inited memory problems
511 in work with lov objdid file.
516 Description: lfs find does not continue on file error
517 Details : Continue other files processing when a file/dir is absent.
521 Description: Inconsistent usage of lustre_pack_reply()
522 Details : Standardize the usage of lustre_pack_reply() such that it
523 always generate a CERROR on failure.
526 Frequency : occasional
528 Description: MDS hang or stay in waiting lock
529 Details : If client receive lock with CBPENDING flag ldlm need send lock
530 cancel as separate rpc, to avoid situation when cancel request
531 can't processed due all i/o threads stay in wait lock.
535 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
539 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
543 Description: Update to SLES9 kernel-2.6.5-7.287.3.
547 Description: Update to RHEL4 latest kernel.
549 Severity : enhancement
551 Description: Build SLES10 patchless client fails
552 Details : The configure was broken by run ./configure with
553 --with-linux-obj=.... argument for patchless client. When the
554 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
555 can't search header adequately. Use absolute path such as
556 -I($LINUX)/include instead.
560 Description: interrupt oig_wait produce painc on resend.
561 Details : brw_redo_request can be used for resend requests from ptlrpcd and
562 private set, and this produce situation when rq_ptlrpcd_data not
563 copyed to new allocated request and triggered LBUG on assert
564 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
565 ptlrpcd set if request is changed and can be safety changed to use
570 Description: LASSERT_{REQ,REP}SWAB macros are buggy
571 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
572 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
573 these macros and replace them with their difinitions instead.
577 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
578 Details : Update which_patch & target file for SLES10 latest kernel.
581 Description: Remove client patches from SLES 10 kernel.
582 Details : This causes SLES 10 clients to behave as patchless clients
583 even on a Lustre-patched (server) kernel.
585 Severity : enhancement
587 Description: self-adjustable client's lru lists
588 Details : use adaptive algorithm for managing client cached locks lru
589 lists according to current server load, other client's work
590 pattern, memory activities, etc. Both, server and client
591 side namespaces provide number of proc tunables for controlling
594 Severity : enhancement
596 Description: light-weight GSS support
597 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
598 and reduce performance overhead.
600 Severity : enhancement
602 Description: Linux keyring support
603 Details : Support using service of Linux keyring for Lustre GSS internal
604 context refresh/cache mechanism.
608 Description: Fix errors in lfs documentation
609 Details : Fixes man pages
613 Description: don't use GFP_* in generic Lustre code.
614 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
618 Description: obdclass is limited by single OBD_ALLOC(idarray)
619 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
623 Description: warnings with build patchless client with vanila 2.6.19 and up
624 Details : change old ctl_table style and replace ctl_table/ctl_table_header
625 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
629 Frequency : Only for SLES
630 Description: sanity_quota fail test_1
631 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
632 file, which makes TSTID[2] inunique.
636 Frequency : Only for SLES9
637 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
638 Details : Update target/ChangeLog/which_patch .
642 Frequency : Only for test_10 in sanity-quota.sh
643 Description: a bug in quota test code
644 Details : checking a test flag in a wrong place
648 Frequency : Only for test_13 in sanity-quota.sh
649 Description: a bug in quota test script
650 Details : shouldn't check overfull allocated quota
654 Description: jbd statistics
655 Details : Port older jbd statistics patch for sles10
657 Severity : enhancement
659 Description: Add printing inode info into message about error in writepage.
664 Description: reply_lock_interpret crash due to race with it and lock cancel.
665 Details : Do not replay locks that are being cancelled. Do not reference
666 locks by their address during replay, just by their handle.
670 Frequency : When flocks are used.
671 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
672 Details : Flock locks might destroy just granted lock if it could be merged
673 with another existing flock, this is done in completion handler,
674 so teach ldlm_cli_enquque_fini that this is a valid case for
680 Description: Oops in read and write path when failing to allocate lock.
681 Details : Check if lock allocation failed and return error back.
685 Description: lstripe command fails for valid OST index
686 Details : The stripe offset is compared to lov->desc.ld_tgt_count
687 instead of lov->desc.ld_active_tgt_count.
691 Description: sanity.sh failed test 103
692 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
697 Description: df doesn't work properly if diskfs blocksize != 4K
698 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
702 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
703 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
705 Severity : enhancement
707 Description: Adaptive timeouts
708 Details : RPC timeouts adapt to changing server load and network
709 conditions to reduce resend attempts and improve recovery time.
713 Description: llapi_file_create() does not allow some changes
714 Details : add llapi_file_open() that allows specifying the mode and
715 open flags, and also returns an open file handle.
719 Description: merge and cleanup kernel patches.
720 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
724 Description: Add journal checksum support.(Kernel part)
725 Details : The journal checksum feature adds two new flags i.e
726 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
727 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
728 indicates that the commit block contains the checksum for
729 the blocks described by the descriptor blocks. Now commit
730 record can be sent to disk without waiting for descriptor
731 blocks to be written to disk. This behavior is controlled
732 using JBD2_FEATURE_ASYNC_COMMIT flag.
736 Description: OSS needs mutliple precreate threads
737 Details : Add ability to start more than one create thread per OSS.
741 Description: Sanity.sh test_65k failed: setstripe should have succeeded
742 Details : execute lfs setstripe on client
746 Description: mds_obd_create error creating tmp object
747 Details : When the user sets quota on root, llog will be affected and can't
748 create files and write files.
751 Frequency : Always on ia64 patchless client, and possibly others.
753 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
754 Details : This allows the patchless client to be loaded on architectures
759 Description: RedHat Update kernel for RHEL5
760 Details : Add the kernel config file for RHEL5.
764 Description: RedHat Update kernel for RHEL5
765 Details : Modify the kernel config file more closer RHEL5.
769 Description: Build failure against Centos5 (RHEL5)
770 Details : Define PAGE_SIZE when it isn't present.
774 Description: test_7 on sanity-quota.sh can't be used on separate machines
775 Details : add facet to handle it
779 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
780 Details : with NFS, the anon dentry's parent was set to itself in
781 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
782 corresponding dentry other than use rec->ur_name.
784 Severity : enhancement
786 Description: lfs setstripe enhancement
787 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
791 Description: enable data checksumming by default
792 Details : enable checksum by default, allow --disable-checksum
793 configure option and "-o nochecksum" mount option. Checksums
794 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
798 Description: lustre support for RHEL5
799 Details : Add support for RHEL5.
803 Description: Client eviction due to failover config
804 Details : after a connection loss, the lustre client should attempt to
805 reconnect to the last active server first before trying the
806 other potential connections.
810 Description: when mds and osts use different quota unit(32bit and 64bit),
811 quota will be released repeatly.
812 Details : void sending multiple quota reqs to mds, which will keep the
813 status between the reqs.
817 Description: osts not allocated evenly to files
818 Details : change the condition to increase offset_idx
821 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
823 Description: Data corruption for OSTs that are formatted larger than 2TB
825 Details : When generating the bio request for lustre file writes the
826 sector number would overflow a temporary variable before being
827 used for the IO. The data reads correctly from Lustre (which
828 will overflow in a similar manner) but other file data or
829 filesystem metadata may be corrupted in some cases.
833 Description: Tune the kernel for good SCSI performance.
834 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
835 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
840 Description: rewrite ext2-derived code in obdclass/uuid.c
841 Details : rewrite inherited code (uuid parsing code from ext2 utils)
842 from scratch preserving functionality.
846 Description: "lfs find -obd UUID" prints directories
847 Details : "lfs find -obd UUID" will return all directory names instead
848 of just file names. It is incorrect because the directories
849 do not reside on the OSTs.
853 Description: use i_size_read and i_size_write in 2.6 port
854 Details : replace inode->i_size access with i_size_read/write()
857 Frequency : when removing large files
859 Description: scheduling issue during removal of large Lustre files
860 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
861 It causes scheduling issues when removing large files (17TB in the
865 Frequency : only with liblustre clients on XT3
867 Description: evictions taking too long
868 Details : allow llrd to evict clients directly on OSTs
871 Frequency : only on ppc
873 Description: /proc/fs/lustre/devices broken on ppc
874 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
875 We should make sure the type of 'index' is loff_t and then cast
876 explicitly as needed below. Do not assign an explicitly cast
880 Frequency : only for rhel5
882 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
883 Details : Modify the target file & which_kernel.
885 Severity : enhancement
887 Description: omit set fsid for export NFS
888 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
889 as FS_REQUIRES_DEV for avoid problems with generate fsid.
893 Frequency : Always, for kernels after 2.6.16
894 Description: Fix warning idr_remove called for id=.. which is not allocated.
895 Details : Last kernels save old s_dev before kill super and not allow
896 to restore from callback - restore it before call kill_anon_super.
900 Description: buffer overruns could theoretically occur
901 Details : llapi_semantic_traverse() modifies the "path" argument by
902 appending values to the end of the origin string, and a
903 overrun may occur. Adding buffer overrun check in liblustreapi.
907 Description: Fix error on 'ls .' at the top of the Lustre mount.
908 Details : Don't revalidate dentry if it is a root dentry.
912 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
913 Details : Modify vm-tunables-rhel4.patch.
917 Description: Kernel config for 2.6.18-vanilla.
918 Details : Modify targets/2.6-vanilla.target.in.
919 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
920 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
921 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
922 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
926 Description: improve handling recoverable errors
927 Details : if request processig with error which can be recoverable on server
928 request should be resend, otherwise page released from cache and
933 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
934 Details : Modify target file & which_patch.
935 A flaw was found in the IA32 system call emulation provided
936 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
937 value could be stored in the %RAX register, which could trigger an
938 out-of-bounds system call table access. An untrusted local user
939 could exploit this flaw to run code in the kernel
940 (ie a root privilege escalation). (CVE-2007-4573).
944 Description: O_DIRECT bypasses client statistics.
945 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
946 read_ahead_stats were updated. copied the stats section from
947 osc_send_oap_rpc() into async_internal().
951 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
955 Description: change order of libsysio includes
956 Details : '#include sysio.h' should always come before '#include xtio.h'
960 Description: handle "serious error: objid * already exists" more gracefully
961 Details : If LAST_ID value on disk is smaller than the objects existing in
962 the O/0/d* directories, it indicates disk corruption and causes an
963 LBUG(). If the object is 0-length, then we should use the existing
964 object. This will help to avoid a full fsck in most cases.
966 Severity : enhancement
968 Description: adapt the lustre_config script to support the upgrade case
969 Details : Add "-u" option for lustre_config script to support upgrading 1.4
970 server targets to 1.6 in parallel.
974 Description: To avoid grant space > avaible space when the disk is almost
975 full. Without this patch you might see the error "grant XXXX >
976 available" or some LBUG about grant, when the disk is almost
978 Details : In filter_check_grant, for non_grant cache write, we should
979 check the left space by if (*left > ungranted + bytes), instead
980 of (*left > ungranted), because only we are sure the left space
981 is enough for another "bytes", then the ungrant space should be
982 increase. In client, we should update cl_avail_grant only there
983 is OBD_MD_FLGRANT in the reply.
987 Description: Update RHEL 4 kernel to fix local root privilege escalation.
988 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
989 described in CVE-2007-4573. This problem could allow untrusted
990 local users to gain root access.
993 Frequency : when using O_DIRECT and quotas
995 Description: Incorrect file ownership on O_DIRECT output files
996 Details : block usage reported by 'lfs quota' does not take into account
997 files that have been written with O_DIRECT.
1002 Description: touch file failed when fs is not full
1003 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1004 otherwise we can get ENOSP while fs is not full.
1008 Description: parallel lock callbacks
1009 Details : Instead of sending blocking and completion callbacks as separated
1010 requests, adding them to a set and sending in parallel.
1013 Frequency : only for Cray XT3
1014 Bugzilla : 12829/13455
1015 Description: Changing primary group doesn't change the group lustre assigns to
1017 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1018 overridden with the primary group provided by the group upcall,
1019 whereas the supplied fsgid can be trusted if it is in the list of
1020 supplementary groups returned by the group upcall.
1022 Severity : enhancement
1024 Description: Allow masking D_WARNING, D_ERROR messages from console
1025 Details : Console messages can now be disabled via lnet.printk.
1029 Description: User code with malformed file open parameter crashes client node
1030 Details : Before packing join_file req, all the related reference should be
1031 checked carefully in case some malformed flags cause fake
1032 join_file req on client.
1036 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1037 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1038 mode is equal to the granted mode. However, it should also check
1039 that LDLM_FL_CP_REQD is not set.
1043 Description: Heavy nfs access might result in deadlocks
1044 Details : After ELC code landed, it is now improper to enqueue any mds
1045 locks under och_sem, because enqueue might want to decide to
1046 cancel open locks for same inode we are holding och_sem for.
1050 Description: Client eviction while running blogbench
1051 Details : A lot of unlink operations with concurrent I/O can lead to a
1052 deadlock causing evictions. To address the problem, the number of
1053 oustanding OST_DESTROY requests is now throttled to
1054 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1055 callbacks are processed in priority.
1059 Description: enable ACLs on MDS by default
1060 Details : ACLs must be enabled on MDS by default.
1063 Frequency : PPC/PPC64 only
1065 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1066 Details : fix duplicated definitions between asm-ppc64/types.h and
1067 lustre_types.h on PPC.
1070 Frequency : PPC/PPC64 only
1072 Description: asm-ppc/segment.h does not exist
1073 Details : fix compile issue on PPC.
1077 Description: better handle error messages in extents code
1080 Frequency : RHEL4 only
1082 Description: mkfs is very slow on IA64/RHEL4
1083 Details : A performance regression has been discovered in the MPT Fusion
1084 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1085 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1086 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1088 Severity : enhancement
1090 Description: SNMP support enhancement
1091 Details : Adding total number of sampled request for an MDS node in snmp
1094 Severity : enhancement
1096 Description: Optimize ldlm waiting list processing for PR extent locks
1097 Details : When processing waiting list for read extent lock and meeting read
1098 lock that is same or wider to it that is not contended, skip
1099 processing rest of the list and immediatelly return current
1100 status of conflictness, since we are guaranteed there are no
1101 conflicting locks in the rest of the list.
1105 Description: Time out and refuse to reconnect
1106 Details : When the failover node is the primary node, it is possible
1107 to have two identical connections in imp_conn_list. We must
1108 compare not conn's pointers but NIDs, otherwise we can defeat
1109 connection throttling.
1113 Description: port llog fixes from b1_6 into HEAD
1114 Details : Port llog reference couting and some llog cleanups from b1_6
1115 (bug 10800) into HEAD, for protect from panic and access to already
1116 free llog structures.
1120 Description: Detect stride IO mode in read-ahead
1121 Details : When a client does stride read, read-ahead should detect that and
1122 read-ahead pages according to the detected stride pattern.
1126 Description: data checksumming impacts single node performance
1127 Details : add support for several checksum algorithm. Currently, only CRC32
1128 and Adler-32 are supported. The checksum type can be changed on
1129 the fly via /proc/fs/lustre/osc/*/checksum_type.
1133 Description: use adler32 for page checksums
1134 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1139 Description: build for x2 fails
1140 Details : fix compile issue on Cray systems.
1144 Description: Properly match for duplicate locks
1145 Details : Due to different lock order from skiplists code, we need to
1146 traverse entire list for now
1149 Frequency : only on PPC/SLES10
1151 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1152 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1153 be defined. Add a hack in mkfs_lustre.c to work around this compile
1158 Description: LASSERT on MDS when client holding flock lock dies
1159 Details : ldlm pool logic depends on number of granted locks equal to
1160 number of released locks which is not true for flock locks, so
1161 just exclude such locks from consideration.
1165 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1166 Details : Use fsfilt_sync() for both the whole filesystem sync and
1167 individual file sync to eliminate dangerous inode locking
1168 with I_LOCK that can lead to a deadlock.
1172 Description: performance in 1.6.3
1173 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1174 sure that 1MB requests are not fragmented by the block layer.
1176 Severity : enhancement
1178 Description: organize the server-side client stats on per-nid basis
1179 Details : Change the structure of stats under obdfilter and mds to
1187 The "uuid"s file would list the uuids of _active_ exports.
1188 And the clear entry is to clear all stats and stale nids.
1190 Severity : enhancement
1192 Description: eliminate client locks in face of contention
1193 Details : file contention detection and lockless i/o implementation
1194 for contended files.
1198 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1201 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1203 Description: MDS may lose file striping (and hence file data) in some cases
1204 Details : If there are additional extended attributes stored on the MDS,
1205 in particular ACLs, SELinux, or user attributes (if user_xattr
1206 is specified for the client mount options) then there is a risk
1207 of attribute loss. Additionally, the Lustre file striping
1208 needs to be larger than default (e.g. striped over all OSTs),
1209 and an additional attribute must be stored initially in the
1210 inode and then increase in size enough to be moved to the
1211 external attribute block (e.g. ACL growing in size) for file
1216 Description: skiplist implementation simplification
1217 Details : skiplists are used to group compatible locks on granted list
1218 that was implemented as tracking first and last lock of each lock group
1219 the patch changes that to using doubly linked lists
1223 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1224 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1225 because LASSERT on both the data supplied by a client, and the data
1226 on disk is dangerous and incorrect.
1228 Severity : enhancement
1230 Description: Slow truncate/writes to huge files at high offsets.
1231 Details : Directly associate cached pages to lock that protect those pages,
1232 this allows us to quickly find what pages to write and remove
1233 once lock callback is received.
1237 Description: more ldlm soft lockups
1238 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1239 starve other threads from the resource lock for a long time in
1240 case of long waiting queue, so change the debug level from
1241 D_OTHER to the less frequently used D_INFO.
1243 Severity : enhancement
1245 Description: add -gid, -group, -uid, -user options to lfs find
1249 Description: Hung threads in invalidate_inode_pages2_range
1250 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1251 one is completed. As a result, some RPCs are stuck in the queue
1256 Description: filter threads hungs on waiting journal commit
1257 Details : Cleanup filter group llog code, then only filter group llog will
1258 be only created in the MDS/OST syncing process.
1262 Description: Procfs and llog threads access destoryed import sometimes.
1263 Details : Sync the import destoryed process with procfs and llog threads by
1264 the import refcount and semaphore.
1266 Severity : enhancement
1268 Description: openlock cache of b1_6 port to HEAD
1273 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1274 Details : If insertion of an extent fails, then discard the inode
1275 preallocation and free data blocks else it can lead to duplicate
1280 Description: don't always update ctime in ext3_xattr_set_handle()
1281 Details : Current xattr code updates the inode ctime in ext3_xattr_set_handle.
1282 In some cases the ctime should not be updated, for example for
1283 2.0->1.8 compatibility it is necessary to delete an xattr and it
1284 should not update the ctime.
1288 Bugzilla : 15713/16362
1289 Description: Assertion in iopen_connect_dentry in 1.6.3
1290 Details : looking up an inode via iopen with the wrong generation number can
1291 populate the dcache with a disconneced dentry while the inode
1292 number is in the process of being reallocated. This causes an
1293 assertion failure in iopen since the inode's dentry list contains
1294 both a connected and disconnected dentry.
1298 Description: assertion failure in ldlm_handle2lock()
1299 Details : fix a race between class_handle_unhash() and class_handle2object()
1300 introduced in lustre 1.6.5 by bug 13622.
1305 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1306 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1311 Description: oops in sd_iostats_seq_show()
1312 Details : unloading/reloading the scsi low level driver triggers a kernel
1313 bug when trying to access the sd iostat file.
1318 Description: Kernel panics during QLogic driver reload
1319 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1320 patch, causing memory corruption.
1325 Description: journal_dev option does not work in b1_6
1326 Details : pass mount option during pre-mount.
1328 Severity : enhancement
1330 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1331 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1332 extent information of a file. It can be used to map logical blocks
1333 in a file to physical blocks in the block device.
1337 Description: LDLM soft lockups - improvement
1338 Details : It is be possible to send the lock handle along with each read
1339 or write request because the client is already doing a lock match
1340 itself so there isn't any reason the OST should have to re-do that
1343 --------------------------------------------------------------------------------
1345 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1347 * Support for kernels:
1348 2.6.5-7.283 (SLES 9),
1349 2.6.9-55.EL (RHEL 4),
1350 2.6.16.46-0.14 (SLES 10),
1351 2.6.18.8 vanilla (kernel.org)
1352 * Client support for unpatched kernels:
1353 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1354 2.6.16 - 2.6.22 vanilla (kernel.org)
1355 * Due to recently discovered recovery problems, we do not recommend
1356 using patchless RHEL 4 clients with this or any earlier release.
1357 * Recommended e2fsprogs version: 1.39.cfs8
1358 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1359 * Starting with this release, the ldiskfs backing filesystem required
1360 by Lustre is now in its own package, lustre-ldiskfs. This package
1361 should be installed. It is versioned separately from Lustre and
1362 may be released separately in future.
1366 Description: block reactivating mgc import until all deactivates complete
1367 Details : Fix race when failing back MDT/MGS to itself (testing)
1370 Frequency : at statup only
1372 Description: mds_lov_synchronize race leads to various problems
1373 Details : simultaneous MDT->OST connections at startup can cause the
1374 sync to abort, leaving the OSC in a bad state.
1376 Severity : enhancement
1378 Description: add optional extra BUILD_VERSION info
1379 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1380 to override the lustre version.
1383 Frequency : 2.6.18 servers only
1385 Description: ll_kern_mount() doesn't release the module reference
1386 Details : The ldiskfs module reference count never drops down to 0
1387 because ll_kern_mount() doesn't release the module reference.
1392 Description: server LBUG when using old ost_num_threads parameter
1393 Details : Accept the old ost_num_threads parameter but warn that it
1394 is deprecated, and fix an off-by-one error that caused an LBUG.
1399 Description: Transient SCSI error results in persistent IO issue
1400 Details : iobuf->dr_error is not reinitialized to 0 between two
1404 Frequency : sometimes when underlying device returns I/O errors
1406 Description: OSTs not going read-only during write failures
1407 Details : OSTs are not remounted read-only when the journal commit threads
1408 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1409 instead of the ext3 wrappers.
1413 Description: poor connect scaling with increasing client count
1414 Details : Don't run filter_grant_sanity_check for more than 100 exports
1415 to improve scaling for large numbers of clients.
1418 Frequency : SLES10 only
1420 Description: sanity-quota.sh quotacheck failed: rc = -22
1421 Details : Quotas cannot be enabled on SLES10.
1424 Frequency : liblustre clients only
1426 Description: getdirentries does not give error when run on compute nodes
1427 Details : getdirentries does not fail when the size specified as an argument
1428 is too small to contain at least one entry
1430 Severity : enhancement
1432 Description: Add LNET router traceability for debug purposes
1433 Details : If a checksum failure occurs with a router as part of the
1434 IO path, the NID of the last router that forwarded the bulk data
1435 is printed so it can be identified.
1440 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1441 Details : Due to a race condition, liblustre clients were occasionally
1442 evicted incorrectly.
1444 Severity : enhancement
1446 Description: lfs setstripe use optional parameters instead of postional
1449 Severity : enhancement
1451 Description: Nanosecond timestamp support for ldiskfs
1452 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1453 resolution timestamps. There is not yet support for this at
1454 the Lustre filesystem level.
1457 Frequency : during server recovery
1459 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1460 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1464 Frequency : nfs export on patchless client
1466 Description: connectathon hang when test nfs export over patchless client
1467 Details : Disconnected dentry cannot be found with lookup, so we do not need
1468 to unhash it or make it invalid
1471 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1472 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1473 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1478 Description: open req refcounting wrong on reconnect
1479 Details : If reconnect happened between getting open reply from server and
1480 call to mdc_set_replay_data in ll_file_open, we will schedule
1481 replay for unreferenced request that we are about to free.
1482 Subsequent close will crash in variety of ways.
1483 Check that request is still eligible for replay in
1484 mdc_set_replay_data().
1489 Description: disable writes to filesystem when reading health_check file
1490 Details : the default for reading the health_check proc file has changed
1491 to NOT do a journal transaction and write to disk, because this
1492 can cause reads of the /proc file to hang and block HA state
1493 checking on a healthy but otherwise heavily loaded system. It
1494 is possible to return to the previous behaviour during configure
1495 with --enable-health-write.
1497 Severity : enhancement
1499 Description: 64-bit inode version
1500 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1501 to the inode. This will be required for version-based recovery.
1506 Description: MDS fails to start if a duplicate client export is detected
1507 Details : in some rare cases it was possible for a client to connect to
1508 an MDS multiple times. Upon recovery the MDS would detect this
1509 and fail during startup. Handle this more gracefully.
1511 Severity : enhancement
1513 Description: Add -o localflock option to simulate old noflock
1515 Details : This will achieve local-only flock/fcntl locks
1521 Description: log_commit_thread vs filter_destroy race leads to crash
1522 Details : Take import reference before releasing llog record semaphore
1527 Description: Wrong request locking in request set processing
1528 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1529 imp_delayed_list, in this place should be used imp_lock.
1532 Frequency : when reconnection
1534 Description: Grant Leak when osc reconnect to OST
1535 Details : When osc reconnect ost, OST(filter) should check whether it
1536 should grant more space to client by comparing fed_grant and
1537 cl_avail_grant, and return the granted space to client instead
1538 of "new granted" space, because client will call osc_init_grant
1539 to update the client grant space info.
1542 Frequency : when client reconnect to OST
1544 Description: Grant Leak when osc do resend and replay bulk write
1545 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1546 bulk write request, because the grant info while be sync between
1547 OSC and OST when reconnect, and we should ignore the grant info
1548 these of resend/replay write req.
1553 Description: Grant space more than avaiable left space sometimes.
1554 Details : When then OST is about to be full, if two bulk writing from
1555 different clients came to OST. Accord the avaliable space of the
1556 OST, the first req should be permitted, and the second one
1557 should be denied by ENOSPC. But if the seconde arrived before
1558 the first one is commited. The OST might wrongly permit second
1559 writing, which will cause grant space > avaiable space.
1562 Frequency : when client is evicted
1564 Description: Grant might be wrongly erased when osc is evicted by OST
1565 Details : when the import is evicted by server, it will fork another
1566 thread ptlrpc_invalidate_import_thread to invalidate the
1567 import, where the grant will be set to 0. While the original
1568 thread will update the grant it got when connecting. So if
1569 the former happened latter, the grant will be wrongly errased
1570 because of this race.
1575 Description: Checking Stale with correct fid
1576 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1577 to check whether it is stale, because sometimes, we want the
1578 enqueue happened anyway, and op_data.fid2 will not be initialized.
1580 Severity : enhancement
1582 Description: update patchless client
1583 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1586 Frequency : only with 2.4 kernel
1588 Description: random memory corruption
1589 Details : size of struct ll_inode_info is to big for union inode.u and this
1590 can be cause of random memory corruption.
1595 Description: Memory leak in recovery
1596 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1597 It should also check obd_fail before fsfilt_start, otherwise if
1598 fsfilt_start return -EROFS,(failover mds during mds recovery).
1599 then the req will return with repmsg->transno = 0 and rc = EROFS.
1600 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1601 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1602 be freed no matter whether fsfilt_commit success or not.
1605 Frequency : only with huge count clients
1607 Description: Prevents from taking the superblock lock in llap_from_page for
1609 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1610 for llap_from_page call in ll_removepage prevents from taking the
1611 superblock lock for a soon died page.
1616 Description: Not check open intent error before release open handle
1617 Details : in some rare cases, the open intent error is not checked before
1618 release open handle, which may cause
1619 ASSERTION(open_req->rq_transno != 0), because it tries to release
1620 the failed open handle.
1625 Description: Set cat log bitmap only after create log success.
1626 Details : in some rare cases, the cat log bitmap is set too early. and it
1627 should be set only after create log success.
1631 Description: Accessing a block bevice can re-enable I/O when Lustre is
1632 tearing down a device.
1633 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1638 Description: service threads may hog cpus when there are a lot of requests
1639 Details : Insert cond_resched to give other threads a chance to use some CPU
1644 Description: the cat log was not initialized in recovery
1645 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1646 unlink log on mds will not be initialized until mds post
1647 recovery. And also in mds post recovery, the unlink log will
1648 initialization will be done asynchronausly, so there will be race
1649 between add unlink log and unlink log initialization.
1653 Description: brw_stats were being printed incorrectly
1654 Details : brw_stats were being printed as log2 but all of them were not
1655 recorded as log2. Also remove some code duplication arising from
1656 filter_tally_{read,write}.
1660 Frequency : rare, only in recovery.
1661 Description: ASSERTION(req->rq_type != LI_POISON) failed
1662 Details : imp_lock should be held while iterating over imp_sending_list for
1663 prevent destroy request after get timeout in ptlrpc_queue_wait.
1667 Description: replay-single.sh test 52 fails
1668 Details : A lock's skiplist need to be cleanup when it being unlinked
1669 from its resource list.
1673 Description: Short directio read returns full requested size rather than
1675 Details : Direct I/O operations should return actual amount of bytes
1676 transferred rather than requested size.
1678 Severity : enhancement
1680 Description: metadata RPC reduction (e.g. for rm performance)
1681 Details : decrease the amount of synchronous RPC between clients and servers
1682 by canceling conflicing lock before the operation on the client
1683 and packing thier handles into the main operation RPC to server.
1685 Severity : enhancement
1687 Description: Async OSC create to avoid the blocking unnecessarily.
1688 Details : If a OST has no remain object, system will block on the creating
1689 when need to create a new object on this OST. Now, ways use
1690 pre-created objects when available, instead of blocking on an
1691 empty osc while others are not empty. If we must block, we block
1692 for the shortest possible period of time.
1694 Severity : enhancement
1696 Description: refine locking for avoid write wrong info into lov_objid file
1697 Details : fix possible races with add new target and write/update data in
1700 --------------------------------------------------------------------------------
1702 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1707 Frequency : on some architectures
1709 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1710 Details : Uninitialized flags sometimes cause configuration commands to
1714 Frequency : patchless clients only
1716 Description: missing __iget() symbol export
1717 Details : The __iget() symbol export is missing. To avoid the need for
1718 this on patchless clients the deathrow inode reaper is turned
1719 off, and we depend on the VM to clean up old inodes. This
1720 dependency was during via the fix for bug 12181.
1722 --------------------------------------------------------------------------------
1724 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1726 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1727 INTEROPERATE with older versions automatically. In many cases a
1728 special upgrade step is needed. Please read the
1729 user documentation before upgrading any part of a live system.
1730 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1731 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1732 * WARNING: Lustre configuration and startup changes are required with
1733 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1739 Severity : enhancement
1741 Description: MountConf
1742 Details : Lustre configuration is now managed via mkfs and mount
1743 commands instead of lmc and lconf. New obd types (MGS, MGC)
1744 are added for dynamic configuration management. See
1745 https://mail.clusterfs.com/wikis/lustre/MountConf for
1748 Severity : enhancement
1750 Description: dynamic OST addition
1751 Details : OSTs can now be added to a live filesystem
1753 Severity : enhancement
1755 Description: startup order invariance
1756 Details : MDTs and OSTs can be started in any order. Clients only
1757 require the MDT to complete startup.
1759 Severity : enhancement
1761 Description: parallel, asynchronous orphan cleanup
1762 Details : orphan cleanup is now performed in separate threads for each
1763 OST, allowing parallel non-blocking operation.
1765 Severity : enhancement
1767 Description: optimized stripe assignment
1768 Details : stripe assignments are now made based on ost space available,
1769 ost previous usage, and OSS previous usage, in order to try
1770 to optimize storage space and networking resources.
1772 Severity : enhancement
1774 Description: Permanently set tunables
1775 Details : All writable /proc/fs/lustre tunables can now be permanently
1776 set on a per-server basis, at mkfs time or on a live system.
1778 Severity : enhancement
1780 Description: Lustre message v2
1781 Details : Add lustre message format v2.
1783 Severity : enhancement
1785 Description: client OST exclusion list
1786 Details : Clients can be started with a list of OSTs that should be
1787 declared "inactive" for known non-responsive OSTs.
1791 Description: SPEC SFS validation failure on NFS v2 over lustre.
1792 Details : Changes the blocksize for regular files to be 2x RPC size,
1793 and not depend on stripe size.
1795 Severity : enhancement
1797 Description: Multiple MD RPCs in flight.
1798 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1799 To avoid overly-overloading MDS, introduce a limit on number of
1800 MDS RPCs in flight for a single client and add /proc controls
1801 to adjust this limit.
1803 Severity : enhancement
1805 Description: client read/write statistics
1806 Details : Add client read/write call usage stats for performance
1807 analysis of user processes.
1808 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1809 file access. extents_stats shows chunk size distribution.
1810 extents_stats_per_process show chunk size distribution per
1813 Severity : enhancement
1815 Description: mds statistics
1816 Details : Add detailed mds operations statistics in
1817 /proc/fs/lustre/mds/*/stats.
1821 Description: Failure of copying files with lustre special EAs.
1822 Details : Client side always return success for setxattr call for lustre
1823 special xattr (currently only "trusted.lov").
1828 Description: Refcount LNET uuids
1829 Details : The global LNET uuid list grew linearly with every startup;
1830 refcount repeated list entries instead of always adding to
1833 Severity : enhancement
1835 Description: Dynamic service threads
1836 Details : Within a small range, start extra service threads
1837 automatically when the request queue builds up.
1840 Frequency : mixed-endian client/server environments
1842 Description: mixed-endian crashes
1843 Details : The new msg_v2 system had some failures in mixed-endian
1846 Severity : enhancement
1848 Description: Easy OST removal
1849 Details : OSTs can be permanently deactivated with e.g. 'lctl
1850 conf_param lustre-OST0001.osc.active=0'
1852 Severity : enhancement
1854 Description: MGS proc entries
1855 Details : Added basic proc entries for the MGS showing what filesystems
1858 Severity : enhancement
1860 Description: provide MGS failover
1861 Details : Added config lock reacquisition after MGS server failover.
1863 Severity : enhancement
1865 Description: add Linux 2.4 support
1866 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1870 Description: a large application tries to do I/O to the same resource and dies
1871 in the middle of it.
1872 Details : Check the req->rq_arrival time after the call to
1873 ost_brw_lock_get(), but before we do anything about
1874 processing it & sending the BULK transfer request. This
1875 should help move old stale pending locks off the queue as
1876 quickly as obd_timeout.
1879 Frequency : when an incorrect nid is specified during startup
1881 Description: ptlrpc connect to non-existant node causes kernel crash
1882 Details : LNET can't be re-entered from an event callback, which
1883 happened when we expire a message after the export has been
1884 cleaned up. Instead, hand the zombie cleanup off to another
1887 Severity : enhancement
1889 Description: plain/inodebits lock performance improvement
1890 Details : Grouping plain/inodebits in granted list by their request modes
1891 and bits policy, thus improving the performance of search through
1895 Frequency : only if OST filesystem is corrupted
1897 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1898 Details : for a short time RPCs with bulk IO are in the replay list,
1899 but replay of bulk IOs is unimplemented. If the OST filesystem
1900 is corrupted due to disk cache incoherency and then replay is
1901 started it is possible to trip an assertion. Avoid putting
1902 committed RPCs into the replay list at all to avoid this issue.
1905 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1907 Description: System hang on startup
1908 Details : This bug allowed the liblustre (e.g. catamount) client to
1909 return to the app before handling all startup RPCs. This
1910 could leave the node unresponsive to lustre network traffic
1911 and manifested as a server ptllnd timeout.
1913 Severity : enhancement
1915 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1916 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1917 boolean to control whether to print peer debug info when a
1918 client's RPC times out.
1921 Frequency : only for kernels with patches from Lustre below 1.4.3
1923 Description: Remove old rdonly API
1924 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1927 Frequency : only for devices with external journals
1929 Description: Set external device read-only also
1930 Details : During a commanded failover stop, we set the disk device
1931 read-only while the server shuts down. We now also set any
1932 external journal device read-only at the same time.
1935 Frequency : when upgrading from 1.4 while trying to change parameters
1937 Description: The wrong (new) MDC name was used when setting parameters for
1938 upgraded MDT's. Also allows changing of OSC (and MDC)
1939 parameters if --writeconf is specified at tunefs upgrade time.
1942 Frequency : when setting specific ost indicies
1944 Description: QOS code breaks on skipped indicies
1945 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1946 created with --index need not be sequential.
1948 Severity : enhancement
1950 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1951 Details : The uninit_groups feature works in conjunction with the kernel
1952 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1953 up the pass1 processing of e2fsck. This is a read-only feature
1954 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1955 mount filesystems that have had this feature enabled.
1957 Severity : enhancement
1959 Description: Improve multi-block allocation algorithm to avoid fragmentation
1960 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1961 allocation locality and avoid filesystem fragmentation.
1963 ------------------------------------------------------------------------------
1965 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1967 * Support for kernels:
1968 2.4.21-47.0.1.EL (RHEL 3)
1969 2.6.5-7.283 (SLES 9)
1970 2.6.9-42.0.10.EL (RHEL 4)
1971 2.6.12.6 vanilla (kernel.org)
1972 2.6.16.27-0.9 (SLES 10)
1973 * Recommended e2fsprogs version: 1.39.cfs5
1975 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1979 Frequency : occasional, depends on client load and configuration
1980 Bugzilla : 12181, 12203
1981 Description: data loss for recently-modified files
1983 Details : In some cases it is possible that recently written or created
1984 files may not be written to disk in a timely manner (this should
1985 normally be within 30s unless client IO load is very high).
1986 The problem appears as zero-length files or files that are a
1987 multiple of 1MB in size after a client crash or client eviction
1988 that are missing data at the end of the file.
1990 This problem is more likely to be hit on clients where files are
1991 repeatedly created and unlinked in the same directory, clients
1992 have a large amount of RAM, have many CPUs, the filesystem has
1993 many OSTs, the clients are rebooted frequently, and/or the files
1994 are not accessed by other nodes after being written.
1996 The presence of the problem can be detected by looking at
1997 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1998 smaller than the second (nr_unused) then dirty files will not
1999 be flushed automatically to disk. "sync; sleep 10" should be
2000 run several times on the node before unmounting it to update
2001 Lustre (this is also safe to run on nodes without this problem).
2003 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2004 that can cause this same problem, so customers using that kernel
2005 also need to update the kernel in addition to Lustre. In order
2006 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2008 It is normal that files written just before a client crash (less
2009 than 30s) may not yet have been flushed to disk, even for local
2013 Frequency : frequent on thin XT3 nodes
2015 Description: UUID collision on thin XT3 Linux nodes
2016 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2017 randomly, since we relied on an insufficiently-seeded PRNG.
2022 Description: OSS hangs after "All ost request buffers busy"
2023 Details : A deadlock between quota and journal operations caused OSS
2024 hangs after printing "All ost request buffers busy."
2027 Frequency : always on liblustre builds
2029 Description: Cleanup compiler warnings on liblustre
2032 Frequency : always on liblustre builds on XT3
2034 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2038 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2039 > 32000 subdirectories
2040 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2041 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2042 subdirectory count crosses 32000. This will aid e2fsck to
2043 correctly handle more than 32000 subdirectories.
2046 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2048 Description: System hang on startup
2049 Details : This bug allowed the liblustre (e.g. catamount) client to
2050 return to the app before handling all startup RPCs. This
2051 could leave the node unresponsive to lustre network traffic
2052 and manifested as a server ptllnd timeout.
2054 Severity : enhancement
2056 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2057 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2058 boolean to control whether to print peer debug info when a
2059 client's RPC times out.
2064 Description: make O_SYNC working on 2.6 kernels
2065 Details : 2.6 kernels use different method for mark pages for write,
2066 so need add a code to lustre for O_SYNC work.
2071 Description: Failure to close file and release space on NFS
2072 Details : Put inode details into lock acquired in ll_intent_file_open.
2073 Use mdc_intent_lock in ll_intent_open to properly
2074 detect all kind of errors unhandled by mdc_enqueue.
2079 Description: proc file read during shutdown sometimes raced obd removal,
2081 Details : Add lock to prevent obd access after proc file removal.
2084 Frequency : Only for files larger than 4GB on 32-bit clients.
2086 Description: improperly doing page alignment of locks
2087 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2088 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2093 Description: RPCs being resent when they shouldn't be
2094 Details : Some RPCs that should not be resent are being resent. This
2095 can cause inconsistencies in the RPC state machine. Do not
2096 resend such requests.
2099 Frequency : rare, only with NFS export
2101 Description: Crash on NFS re-export node
2102 Details : under very unusual load conditions an assertion is hit in
2103 ll_intent_file_open()
2106 Frequency : only if OST filesystem is corrupted
2108 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2109 Details : for a short time RPCs with bulk IO are in the replay list,
2110 but replay of bulk IOs is unimplemented. If the OST filesystem
2111 is corrupted due to disk cache incoherency and then replay is
2112 started it is possible to trip an assertion. Avoid putting
2113 committed RPCs into the replay list at all to avoid this issue.
2118 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2119 Details : Large single O_DIRECT read and write calls can fail to allocate
2120 a sufficiently large buffer to process the request. In case of
2121 allocation failure the allocation is retried with a smaller
2122 buffer and broken into smaller requests.
2124 Severity : enhancement
2126 Description: Add -o localflock option to simulate old noflock behaviour.
2127 Details : This will achieve local-only flock/fcntl locks coherentness.
2132 Description: versioning check is incomplete
2133 Details : Checking the version difference of client vs. server, report
2134 error if the gap is too big.
2139 Description: add support PG_writeback bit
2140 Details : add support for PG_writeback bit for Lustre, for more carefull
2141 work with page cache in 2.6 kernel. This also fix some deadlocks
2142 and remove hack for work O_SYNC with 2.6 kernel.
2144 Severity : enhancement
2146 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2147 Details : The uninit_groups feature works in conjunction with the kernel
2148 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2149 up the pass1 processing of e2fsck. This is a read-only feature
2150 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2151 mount filesystems that have had this feature enabled.
2153 Severity : enhancement
2155 Description: Improve multi-block allocation algorithm to avoid fragmentation
2156 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2157 allocation locality and avoid filesystem fragmentation.
2159 ------------------------------------------------------------------------------
2161 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2163 * Support for kernels:
2164 2.6.9-42.0.3.EL (RHEL 4)
2165 2.6.5-7.276 (SLES 9)
2166 2.4.21-47.0.1.EL (RHEL 3)
2167 2.6.12.6 vanilla (kernel.org)
2168 2.6.16.21-0.8 (SLES10)
2169 * Recommended e2fsprogs version: 1.39.cfs2-0
2171 * The backwards-compatible /proc/sys/portals symlink has been removed
2172 in this release. Before upgrading, please ensure that you change
2173 any configuration scripts or /etc/sysctl.conf files that access
2174 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2175 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2176 in advance of the upgrade on any system running Lustre 1.4.6 or
2177 newer, since /proc/sys/lnet was added in that version.
2178 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2182 Frequency : only when quota is used
2184 Description: avoid scanning export list for quota master
2185 Details : Change the algorithms to avoid scanning export list in order
2186 to improve the efficiency.
2189 Frequency : MDS failover only, very rarely
2191 Description: "went back in time" messages on mds failover
2192 Details : The greatest transno may be lost when the current operation
2193 finishes with an error (transno==0) and the client's last_rcvd
2194 record is over-written. Save the greatest transno in the
2195 mds_last_transno for this case.
2198 Frequency : always for specific kernels and striping counts
2200 Description: client may get "Matching packet too big" without ACL support
2201 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2202 when trying to access files in certain configurations. The
2203 clients should in fact be denied when mounting because they do
2204 not understand ACLs.
2207 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2209 Description: many clients connecting with IO in progress causes connect timeouts
2210 Details : Avoid synchronous journal commits to avoid delays caused by many
2211 clients connecting/disconnecting when bulk IO is in progress.
2212 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2213 OST_IO_PORTAL to avoid delays behind potentially many pending
2217 Frequency : occasionally with multiple writers to a single file
2219 Description: shared writes to file may result in wrong size reported by stat()
2220 Details : Allow growing of kms when extent lock is cancelled
2223 Frequency : always with random mmap IO to multi-striped file
2225 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2226 Details : Only if the hole is at the end of OST object so that kms is too
2227 small. Fix is to increase kms accordingly in ll_nopage.
2230 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2232 Description: writes to a missing object would leak memory on the OST
2233 Details : If there is an inconsistency between the MDS and OST filesystems,
2234 such that the MDS references an object that doesn't exist, writes
2235 to that object will leak memory due to incorrect cleanup in the
2236 error handling path, eventually running out of memory on the OST.
2241 Description: Creating too long symlink causes lustre errors
2242 Details : Check symlink and name lengths before sending requests to MDS.
2245 Frequency : only if flock is enabled (not on by default)
2247 Description: posix locks not released on fd closure on 2.6.9+
2248 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2249 kernels, this caused such locks not to be released on fd close and
2250 then assertions on fs unmount about still used locks.
2253 Frequency : MDS failover only, very rarely
2255 Description: clients may get ASSERTION(granted_lock != NULL)
2256 Details : When request was taking a long time, and a client was resending
2257 a getattr by name lock request. The were multiple lock requests
2258 with the same client lock handle and
2259 mds_getattr_name->fixup_handle_for_resent_request found one of the
2260 lock handles but later failed with ASSERTION(granted_lock != NULL).
2265 Description: handle->h_buffer_credits > 0, assertion failure
2266 Details : h_buffer_credits is zero after truncate, causing assertion
2267 failure. This patch extends the transaction or creates a new
2271 Frequency : NFS re-export or patchless client
2272 Bugzilla : 11179, 10796
2273 Description: Crash on NFS re-export node (__d_move)
2274 Details : We do not want to hash the dentry if we don't have a lock.
2275 But if this dentry is later used in d_move, we'd hit uninitialised
2276 list head d_hash, so we just do this to init d_hash field but
2277 leave dentry unhashed.
2280 Frequency : NFS re-export or patchless client
2282 Description: NFS exports has problem with symbolic link
2283 Details : lustre client didn't properly install dentry when re-exported
2284 to NFS or running patchless client.
2287 Frequency : NFS re-export or patchless client
2289 Description: Various nfs/patchless fixes.
2290 Details : fixes reuse disconected alias for lookup process - this fixes
2291 warning "find_exported_dentry: npd != pd",
2292 fix permission error with open files at nfs.
2293 fix apply umask when do revalidate.
2296 Frequency : occasional
2298 Description: Crash on NFS re-export node
2299 Details : calling clear_page() on the wrong pointer triggered oops in
2300 generic_mapping_read().
2303 Frequency : rarely, using O_DIRECT IO
2305 Description: unaligned directio crashes client with LASSERT
2306 Details : check for unaligned buffers before trying any requests.
2309 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2311 Description: stale data returned from RAID cache
2312 Details : If only a small amount of IO is done to the RAID device before
2313 reading it again it is possible to get stale data from the RAID
2314 cache instead of reading it from disk.
2317 Frequency : always for sles10 kernel
2319 Description: sles10 support
2320 Details : ll_follow_link: compile fixes and using of nd_set_link
2321 under newer kernels.
2324 Frequency : depends on arch, kernel and compiler version, always on sles10
2327 Description: recursive or deep enough symlinks cause stack overflow
2328 Details : getting rid of large stack-allocated variable in
2332 Frequency : depends on hardware
2334 Description: lustre write performance loss in the SLES10 kernel
2335 Details : the performance loss is caused by using of write barriers in the
2336 ext3 code. The SLES10 kernel turns barrier support on by
2337 default. The fix is to undo that change for ldiskfs.
2339 ------------------------------------------------------------------------------
2341 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2343 * Support for kernels:
2344 2.6.9-42.0.3EL (RHEL 4)
2345 2.6.5-7.276 (SLES 9)
2346 2.4.21-47.0.1.EL (RHEL 3)
2347 2.6.12.6 vanilla (kernel.org)
2351 Frequency : quota enabled and large files being deleted
2353 Description: releasing more than 4GB of quota at once hangs OST
2354 Details : If a user deletes more than 4GB of files on a single OST it
2355 will cause the OST to spin in an infinite loop. Release
2356 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2361 Description: statfs data retrieved from /proc may be stale or zero
2362 Details : When reading per-device statfs data from /proc, in the
2363 {kbytes,files}_{total,free,avail} files, it may appear
2364 as zero or be out of date.
2367 Frequency : systems with MD RAID1 external journal devices
2369 Description: lconf's call to blkid is confused by RAID1 journal devices
2370 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2371 of returning all block devices that match the journal UUID.
2374 Frequency : always, for aggregate stripe size over 4GB
2376 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2377 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2378 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2381 Frequency : always if "lfs find" used on a local file/directory
2383 Description: "lfs find" segfaults if used on a local file/directory
2384 Details : The case where a directory component was not specified wasn't
2385 handled correctly. Handle this properly.
2388 Frequency : always on ppc64
2390 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2391 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2392 perform endianness conversion.
2395 Frequency : rarely (truncate to non-zero file size after write under load)
2396 Bugzilla : 10730, 10687
2397 Description: Files padded with zeros to next 4K multiple
2398 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2399 it is possible that files that are truncated to a non-zero size
2400 immediately after being written are filled with zero bytes beyond
2401 the truncated size. No file data is lost.
2403 Severity : enhancement
2405 Description: Allow recovery/failover for liblustre clients.
2406 Details : liblustre clients were unaware of failover configurations until
2409 Severity : enhancement
2411 Description: user file locks should fail when not mounting with flock option
2412 Details : Set up an error-returning stub in ll_file_operations.lock field
2413 to prevent incorrect behaviour when client is mounted without
2414 flock option. Also, set up properly f_op->flock field for
2418 Frequency : always on ia64
2420 Description: "lfs df" loops on printing out MDS statfs information
2421 Details : The obd_ioctl_data was not initialized and in some systems
2422 this caused a failure during the ioctl that did not return
2423 an error. Initialize the struct and return an error on failure.
2426 Frequency : SLES 9 only
2428 Description: Error of copying files with lustre special EAs as root
2429 Details : Client side always return success for setxattr call for lustre
2430 special xattr (currently only "trusted.lov").
2433 Frequency : rarely on clusters with both ia64+i386 clients
2435 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2436 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2437 clients are doing concurrent writes to the same file it is
2438 possible that the ia64 clients may LASSERT because the OST
2439 extent locks are not PAGE_SIZE aligned. Ensure that grown
2440 locks are always aligned on the request boundary.
2443 Frequency : specific use, occasional
2445 Description: Overwriting in use executable truncates on-disk binary image
2446 Details : If one node attempts to overwrite an executable in use by
2447 another node, we now correctly return ETXTBSY instead of
2448 truncating the file.
2450 Severity : enhancement
2452 Description: Async OSC create to avoid the blocking unnecessarily.
2453 Details : If a OST has no remain object, system will block on the creating
2454 when need to create a new object on this OST. Now, ways use
2455 pre-created objects when available, instead of blocking on an
2456 empty osc while others are not empty. If we must block, we block
2457 for the shortest possible period of time.
2462 Description: chmod on Lustre root is propagated to other clients
2463 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2464 invalid by the follow_mount time.
2467 Frequency : liblustre clients only
2469 Description: Race in 'instant cancel' lock handling could lead to such locks
2470 never to be granted in case of SMP MDS
2471 Details : Do not destroy not yet granted but cbpending locks in
2475 Frequency : replay/resend of open
2477 Description: non null lock assetion failure in mds_intent_policy
2478 Details : Trying to replay/resend lockless open requests resulted in
2479 mds_open() returning 0 with no lock. Now it sets a flag if
2480 a lock is going to be returned.
2482 Severity : enhancement
2484 Description: Checksum enhancements
2485 Details : New checksum enhancements allow for resending RPCs that failed
2488 Severity : enhancement
2490 Description: Tunables on number of dirty pages in cacche
2491 Details : Allow to set limit on number of dirty pages cached.
2496 Description: client crash on unmount - lock still has references
2497 Details : In some error handling cases it was possible to leak a lock
2498 reference on a client while accessing a file. This was not
2499 harmful to the client during operation, but would cause the
2500 client to crash when the filesystem is unmounted.
2503 Frequency : specific case, rare
2505 Description: ETXTBSY on mds though file not in use by client
2506 Details : ETXTBSY is no longer incorrectly returned when attempting to
2507 chmod or chown a directory that the user previously tried to
2508 execute or a currently-executing binary.
2511 Frequency : extremely rare except on liblustre-based clients
2513 Description: Lustre space not freed when files are deleted
2514 Details : Clean up open-unlinked files after client eviction. Previously
2515 the unlink was skipped and the files remained as orphans.
2520 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2521 Details : In some cases it was possible to send a blocking callback to
2522 a client doing a glimpse, even though that client didn't get
2523 a lock granted. When the glimpse lock is cancelled on the OST
2524 the freed lock is left on the waiting list and corrupted the list.
2527 Frequency : all core dumps
2529 Description: Broke core dumps to lustre
2530 Details : Negative dentry may be unhashed if parent does not have UPDATE
2531 lock, but some callers, e.g. do_coredump, expect dentry to be
2532 hashed after successful create, hash it in ll_create_it.
2534 ------------------------------------------------------------------------------
2536 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2538 * Support for kernels:
2539 2.6.9-42.0.2.EL (RHEL 4)
2540 2.6.5-7.276 (SLES 9)
2541 2.4.21-40.EL (RHEL 3)
2542 2.6.12.6 vanilla (kernel.org)
2546 Frequency : always on RHEL 3
2548 Description: Number of open files grows over time
2549 Details : The number of open files grows over time, whether or not
2550 Lustre is started. This was due to a filp leak introduced
2551 by one of our kernel patches.
2553 ------------------------------------------------------------------------------
2555 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2557 * Support for kernels:
2558 2.6.9-42.EL (RHEL 4)
2559 2.6.5-7.276 (SLES 9)
2560 2.4.21-40.EL (RHEL 3)
2561 2.6.12.6 vanilla (kernel.org)
2566 Bugzilla : 5719, 9635, 9792, 9684
2567 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2568 Details : If a server is under heavy load and cannot reply to new
2569 connection requests before the client resends the (re)connect,
2570 the connection handling code can behave badly if two service
2571 threads are concurrently handing separate (re)connections from
2572 the same client. Add better locking to the connection handling
2573 code, and ensure that only a single connection will be processed
2574 for a given client UUID, even if the lock is dropped.
2576 Severity : enhancement
2578 Description: add TCP zero-copy support to kernel
2579 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2580 sends if the hardware supports scatter-gather and checksumming.
2581 This allows socklnd to do client-write and server-read more
2582 efficiently and reduce CPU utilization from skbuf copying.
2585 Frequency : only if NFS exporting from client
2587 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2588 Details : If setting attributes on a file created under NFS that had
2589 never been opened it would be possible to oops the client
2590 if the file had no objects.
2593 Frequency : always for liblustre
2595 Description: liblustre client does MDS+OSTs setattr RPC for each write
2596 Details : When doing a write from a liblustre client, the client
2597 incorrectly issued an RPC to the MDS and each OST the file was
2598 striped over in order to update the timestamps. When writing
2599 with small chunks and many clients this could overwhelm the MDS
2600 with RPCs. In all cases it would slow down the write because
2601 these RPCs are unnecessary.
2603 Severity : enhancement
2605 Description: allow number of MDS service threads to be changed at module load
2606 Details : It is now possible to change the number of MDS service threads
2607 running. Adding "options mds mds_num_threads={N}" to the MDS's
2608 /etc/modprobe.conf will set the number of threads for the next
2609 time Lustre is restarted (assuming the "mds" module is also
2610 reloaded at that time). The default number of threads will
2611 stay the same, 32 for most systems.
2616 Description: OST crash if filesystem is unformatted or corrupt
2617 Details : If an OST is started on a device that has never been formatted
2618 or if the filesystem is corrupt and cannot even mount then the
2619 error handling cleanup routines would dereference a NULL pointer.
2624 Description: NULL pointer deref in llap_from_page.
2625 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2626 incorrect) mapping if the page was truncated/reclaimed while it was
2627 searched for. Check for this condition and skip such pages when
2628 doing readahead. Introduce extra check to llap_from_page() to
2629 verify page->mapping->host is non-NULL (so page is not anonymous).
2632 Frequency : Sometimes when using sys_sendfile
2634 Description: "page not covered by a lock" warnings from ll_readpage
2635 Details : sendfile called ll_readpage without right page locks present.
2636 Now we introduced ll_file_sendfile that does necessary locking
2637 around call to generic_file_sendfile() much like we do in
2641 Frequency : with certain MDS communication failures at client mount time
2643 Description: NULL pointer deref after failed client mount
2644 Details : a client connection request may delayed by the network layer
2645 and not be sent until after the PTLRPC layer has timed out the
2646 request. If the client fails the mount immediately it will try
2647 to clean up before the network times out the request. Add a
2648 reference from the request import to the obd device and delay
2649 the cleanup until the network drops the request.
2652 Frequency : occasionally during client (re)connect
2654 Description: assertion failure during client (re)connect
2655 Details : processing a client connection request may be delayed by the
2656 client or server longer than the client connect timeout. This
2657 causes the client to resend the connection request. If the
2658 original connection request is replied in this interval, the
2659 client may trip an assertion failure in ptlrpc_connect_interpret()
2660 which thought it would be the only running connect process.
2663 Frequency : only with obd_echo servers and clients that are rebooted
2665 Description: kernel BUG accessing uninitialized data structure
2666 Details : When running an obd_echo server it did not start the ping_evictor
2667 thread, and when a client was evicted an uninitialized data
2668 structure was accessed. Start the ping_evictor in the RPC
2669 service startup instead of the OBD startup.
2671 Severity : enhancement
2672 Bugzilla : 10193 (patchless)
2673 Description: Remove dependency on various unexported kernel interfaces.
2674 Details : No longer need reparent_to_init, exit_mm, exit_files,
2675 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2678 Frequency : rare (only users of deprecated and unsupported LDAP config)
2680 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2681 Details : LDAP apparently contains 'lustreName' attributes instead of
2682 'name'. A simple remapping of the name is sufficient.
2685 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2687 Description: waiting_locks_callback trips kernel BUG if client is evicted
2688 Details : Running with the dump_on_timeout debug flag turned on makes
2689 it possible that the waiting_locks_callback() can try to dump
2690 the Lustre kernel debug logs from an interrupt handler. Defer
2691 this log dumping to the expired_lock_main() thread.
2693 Severity : enhancement
2695 Description: Support NFS exporting on 2.6 kernels.
2696 Details : Implement non-rawops metadata methods for NFS server to use without
2697 changing NFS server code.
2700 Frequency : very rare (synthetic metadata workload only)
2702 Description: two racing renames might cause an MDS thread to deadlock
2703 Details : Running the "racer" program may cause one MDS thread to rename
2704 a file from being the source of a rename to being the target of
2705 a rename at exactly the same time that another thread is doing
2706 so, and the second thread has already enqueued these locks after
2707 doing a lookup of the target and is trying to relock them in
2708 order. Ensure that we don't try to re-lock the same resource.
2711 Frequency : only very large systems with liblustre clients
2713 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2714 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2715 all OSTs in parallel. This allows the eviction of stale liblustre
2716 clients to proceed much faster than if they were done in series,
2717 and also offers similar improvements for other set_info RPCs.
2722 Description: excessive CPU usage during initial read phase on client
2723 Details : During the initial read phase on a client, it would agressively
2724 retry readahead on the file, consuming too much CPU and impacting
2725 performance (since 1.4.5.8). Improve the readahead algorithm
2726 to avoid this, and also improve some other common cases (read
2727 of small files in particular, where "small" is files smaller than
2728 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2733 Description: MDS crash when receiving packet with unknown intent.
2734 Details : Do not LBUG in unknown intent case, just return -EFAULT
2736 Severity : enhancement
2737 Bugzilla : 9293, 9385
2738 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2739 Details : Do not serialize getattr (non-intent version) and statfs.
2742 Frequency : occasional, when OST network is overloaded/intermittent
2744 Description: client evicted by OST after bulk IO timeout
2745 Details : If a client sends a bulk IO request (read or write) the OST
2746 may evict the client if it is unresposive to its data GET/PUT
2747 request. This is incorrect if the network is overloaded (takes
2748 too long to transfer the RPC data) or dropped the OST GET/PUT
2749 request. There is no need to evict the client at all, since
2750 the pinger and/or lock callbacks will handle this, and the
2751 client can restart the bulk request.
2754 Frequency : Always when mmapping file with no objects
2756 Description: client crashes when mmapping file with no objects
2757 Details : Check that we actually have objects in a file before doing any
2758 operations on objects in ll_vm_open, ll_vm_close and
2764 Description: Request leak when working with deleted CWD
2765 Details : Introduce advanced request refcount tracking for requests
2766 referenced from lustre intent.
2768 Severity : Enhancement
2770 Description: Cache open file handles on client.
2771 Details : MDS now will return special lock along with openhandle, if
2772 requested and client is allowed to hold openhandle, even if unused,
2773 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2774 closing files for every read/write openration.
2776 Severity : Enhancement
2778 Description: Cache open negative dentries on client when possible.
2779 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2780 negative dentries on lock revocation.
2785 Description: Remounting a client read-only wasn't possible with a zconf mount
2786 Details : It wasn't possible to remount a client read-only with llmount.
2788 Severity : enhancement
2789 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2790 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2791 MPICH is included to add Lustre-specific ADIO interfaces.
2792 This is based closely on the UFS ADIO layer and only differs
2793 in file creation, in order to allow the OST striping to be set.
2794 This is user-contributed code and not supported by CFS.
2799 Description: extended inode attributes (immutable, append-only) work improperly
2800 when 2.4 and 2.6 kernels are used on client/server or vice versa
2801 Details : Introduce kernel-independent values for these flags.
2803 Severity : enhancement
2806 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2807 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2808 granularity. Now, Fractional MB granularity is be supported,
2809 this is very useful for low memory system.
2811 Severity : enhancement
2813 Description: Getattr by fid
2814 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2815 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2816 have locks and avoids getattr rpc requests for it on every stat.
2819 Frequency : Always, for filesystems larger than 2TB
2821 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2822 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2823 filesystems larger than 2TB to be mounted with mballoc enabled.
2826 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2828 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2829 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2830 filesystems larger than 2TB is mounted then the kernel will
2831 silently corrupt the start of the filesystem. CONFIG_LBD is
2832 enabled for all CFS-supported kernels, but the possibility of
2833 this happening with a modified kernel config exists.
2835 Severity : enhancement
2837 Description: add client O_DIRECT support for 2.6 kernels
2838 Details : It is now possible to do O_DIRECT reads and writes to files
2839 in the Lustre client mountpoint on 2.6 kernel clients.
2841 Severity : enhancement
2843 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2844 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2845 parallel, not waiting for response from every OST before sending
2846 a rpc to the next OST.
2851 Description: setattr vs write race when updating file timestamps
2852 Details : Client processes that update a file timestamp into the past
2853 right after writing to the file (e.g. tar) it is possible that
2854 the updated file modification time can be reset to the current
2855 time due to a race between processing the setattr and write RPC.
2857 Severity : enhancement
2859 Description: Bring 'lfs find' closer in line with regular Linux find.
2860 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2861 -print0 options and obtains all the needed info through the lustre
2864 Severity : enhancement
2866 Description: support up to 1024 configured devices on one node
2867 Details : change obd_dev array from statically allocated to dynamically
2868 allocated structs as they are first used to reduce memory usage
2873 Description: Flush dirty partially truncated pages during truncate
2874 Details : Immediatelly flush partially truncated pages in filter_setattr,
2875 this way we completely avoid having any pages in page cache on OST
2876 and can retire ugly workarounds during writes to flush such pages.
2881 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2882 Details : protect inode from truncation within vfs_unlink() context
2883 just take a reference before calling vfs_unlink() and release it
2884 when parent's i_sem is free.
2889 Description: last_id value checked outside lock on OST caused LASSERT failure
2890 Details : If there were multiple MDS->OST object precreate requests in
2891 flight, it was possible that the OST's last object id was checked
2892 outside a lock and incorrectly tripped an assertion. Move checks
2893 inside locks, and discard old precreate requests.
2896 Frequency : always, if extents are used on OSTs
2898 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2899 Details : OSTs using the extents format would not zero the high 16 bits of
2900 the index physical block number. This is not a problem for any
2901 OST filesystems smaller than 16TB, and no kernels support ext3
2902 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2903 new/modified files) and can be fixed for existing filesystems
2904 with e2fsprogs-1.39-cfs1.
2909 Description: import connection selection may be incorrect if timer wraps
2910 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2911 import connections to be ignored if the 32-bit jiffies counter
2912 wraps. Use a 64-bit jiffies counter.
2915 Frequency : very large clusters immediately after boot
2917 Description: LNET request buffers exhausted under heavy short-term load
2918 Details : If a large number of client requests are generated on a service
2919 that has previously never seen so many requests it is possible
2920 that the request buffer growth cannot keep up with the spike in
2921 demand. Instead of dropping incoming requests, they are held in
2922 the LND until the RPC service can accept more requests.
2925 Frequency : Sometimes during replay
2927 Description: Assertion failure in ll_local_open after replay.
2928 Details : If replay happened on an open request reply before we were able
2929 to set replay handler, reply will become not swabbed tripping the
2930 assertion in ll_local_open. Now we set the handler right after
2931 recognising of open request
2934 Frequency : very rare
2936 Description: kernel reports "badness in vsnprintf"
2937 Details : Reading from the "recovery_status" /proc file in small chunks
2938 may cause a negative length in lprocfs_obd_rd_recovery_status()
2939 call to vsnprintf() (which is otherwise harmless). Exit early
2940 if there is no more space in the output buffer.
2942 Severity : enhancement
2944 Description: clear OBD RPC statistics by writing to them
2945 Details : It is now possible to clear the OBD RPC statistics by writing
2946 to the "stats" file.
2951 Description: Client mtime is not the same on different clients after utimes
2952 Details : In some cases, the client was using the utimes() syscall on
2953 a file cached on another node. The clients now validate the
2954 ctime from the MDS + OSTs to determine which one is right.
2959 Description: Inability to activate failout mode
2960 Details : lconf script incorrectly assumed that in pythong string's numeric
2961 value is used in comparisons.
2964 Frequency : always with multiple stripes per file
2966 Description: Inefficient object allocation for mutli-stripe files
2967 Details : When selecting which OSTs to stripe files over, for files with
2968 a stripe count that divides evenly into the number of OSTs,
2969 the MDS is always picking the same starting OST for each file.
2970 Return the OST selection heuristic to the original design.
2975 Description: mount failures may take full timeout to return an error
2976 Details : Under some heavy load conditions it is possible that a
2977 failed mount can wait for the full obd_timeout interval,
2978 possibly several minutes, before reporting an error.
2979 Instead return an error as soon as the status is known.
2981 Frequency : quota enabled and large files being deleted
2983 Description: releasing more than 4GB of quota at once hangs OST
2984 Details : If a user deletes more than 4GB of files on a single OST it
2985 will cause the OST to spin in an infinite loop. Release
2986 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2991 Description: statfs data retrieved from /proc may be stale or zero
2992 Details : When reading per-device statfs data from /proc, in the
2993 {kbytes,files}_{total,free,avail} files, it may appear
2994 as zero or be out of date.
2997 Frequency : systems with MD RAID1 external journal devices
2999 Description: lconf's call to blkid is confused by RAID1 journal devices
3000 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3001 of returning all block devices that match the journal UUID.
3004 Frequency : always, for aggregate stripe size over 4GB
3006 Description: assertion fails when trying to use 4GB stripe size
3007 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3008 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3011 Frequency : always on ppc64
3013 Description: the first write on an ext3 filesystem with mballoc got stuck
3014 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3015 perform endianness conversion.
3017 ------------------------------------------------------------------------------
3019 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3021 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3022 INTEROPERATE with older versions automatically. Please read the
3023 user documentation before upgrading any part of a live system.
3024 * WARNING: Lustre networking configuration changes are required with
3025 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3028 * Support for kernels:
3029 2.6.9-22.0.2.EL (RHEL 4)
3030 2.6.5-7.244 (SLES 9)
3031 2.6.12.6 vanilla (kernel.org)
3034 Severity : enhancement
3035 Bugzilla : 7981/8208
3036 Description: Introduced Lustre Networking (LNET)
3037 Details : LNET is new networking infrastructure for Lustre, it includes
3038 a reorganized network configuration mode (see the user
3039 documentation for full details) as well as support for routing
3040 between different network fabrics. Lustre Networking Devices
3041 (LNDs) for the supported network fabrics have also been
3042 created for this new infrastructure.
3044 Severity : enhancement
3045 Description: Introduced Access control lists
3046 Details : clients can set ACLs on files and directories in order to have
3047 more fine-grained permissions than the standard Unix UGO+RWX.
3048 The MDS must be started with the "-o acl" mount option.
3050 Severity : enhancement
3051 Description: Introduced filesystem quotas
3052 Details : Administrators may now establish per-user quotas on the
3055 Severity : enhancement
3057 Description: Configuration change for the XT3
3058 The PTLLND is now used to run Lustre over Portals on the XT3
3059 The configure option(s) --with-cray-portals are no longer used.
3060 Rather --with-portals=<path-to-portals-includes> is used to
3061 enable building on the XT3. In addition to enable XT3 specific
3062 features the option --enable-cray-xt3 must be used.
3067 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3068 Details : A race between allocating a new llcd and re-getting the llcd_lock
3069 allowed another thread to grab newly-allocated llcd.
3071 Severity : enhancement
3073 Description: 2.6 OST async journal commit and locking fix to improve performance
3074 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3075 2.6 kernels are now async as they already were in 2.4 kernels so
3076 that they can commit concurrently with the network bulk transfer.
3077 For block-allocated files the filter allocation semaphore is held
3078 to avoid filesystem fragmentation during allocation. BKL lock
3079 removed for 2.6 xattr operations where it is no longer needed.
3084 Description: lconf incorrectly determined whether two IP networks could talk
3085 Details : In some more complicated routing and multiple-network
3086 configurations, lconf will avoid trying to make a network
3087 connection to a disjoint part of the IP space. It was doing the
3088 math incorrectly for one set of cases.
3093 Description: Fix for potential infinite loop processing records in an llog.
3094 Details : If an llog record is corrupted/zeroed, it is possible to loop
3095 forever in llog_process(). Validate the llog record length
3096 and skip the remainder of the block on error.
3099 Frequency : occasional (liblustre only)
3101 Description: liblustre could not open files whose last component is a symlink
3102 Details : sysio_path_walk() would incorrectly pass the open intent to
3103 intermediate path components.
3106 Frequency : rare (liblustre only with non-standard tuning)
3107 Bugzilla : 7201 (7350)
3108 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3109 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3110 after it has been released, only do this for referenced locks
3112 Severity : enhancement
3114 Description: specifying an (invalid) directory default stripe_size of -1
3115 would reset the directory default striping
3116 Details : stripe_size -1 was used internally to signal directory stripe
3117 removal, now use "all default" to signal dir stripe removal
3118 as a directory striping of "all default" is not useful
3121 Frequency : common for large clusters running liblustre clients
3123 Description: doing an ls when liblustre clients are running is slow
3124 Details : sending a glimpse AST to a liblustre client waits for every AST
3125 to time out, as liblustre clients will not respond. Since they
3126 cannot cache data we refresh the OST lock LVB from disk instead.
3128 Severity : enhancement
3130 Description: doing an ls at the same time as file IO can be slow
3131 Details : enqueue and other "small" requests can be blocked behind many
3132 large IO requests. Create a new OST IO portal for non-IO
3133 requests so they can be processed faster.
3136 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3138 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3139 created file with mode 000
3140 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3141 allow an ftruncate() as a non-root user to a file with mode 000.
3142 Lustre now respects this flag to disable mode checks when
3143 truncating a file owned by the user
3146 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3149 Description: Revoking locks from clients that went dead or catatonic might take
3151 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3152 cancellation of such locks instant on servers without waiting for
3153 any reply from clients. Clients drops these locks when cancel
3154 notification from server is received without replying.
3157 Frequency : liblustre-only, when liblustre client dies or becomes busy
3159 Description: Doing ls on Linux clients can take a long time with active
3161 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3162 granting such locks to it in the first place if possible. Locks
3163 are taken by proxy on the OST during the read or write and
3164 dropped immediately afterward. Add connect flags handling, do
3165 not grant locks to liblustre clients for glimpse ASTs.
3167 Severity : enhancement
3169 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3170 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3171 sequential reads and behaves badly with applications doing random
3172 reads. Improve it by reading ahead at least read region, and
3173 avoiding excessive large RPC for small reads.
3175 Severity : enhancement
3177 Description: Creating more than 1000 files for a single job may cause a load
3178 imbalance on the OSTs if there are also a large number of OSTs.
3179 Details : qos_prep_create() uses an OST index reseed value that is an
3180 even multiple of the number of available OSTs so that if the
3181 reseed happens in the middle of the object allocation it will
3182 still utilize the OSTs as uniformly as possible.
3187 Description: OST or MDS may oops in ping_evictor_main()
3188 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3189 but doesn't restart at beginning of obd_exports_timed list
3192 Severity : enhancement
3194 Description: improve by-nid export eviction on the MDS and OST
3195 Details : allow multiple exports with the same NID to be evicted at one
3196 time without re-searching the exports list.
3199 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3201 Description: MDS may oops in groups_free()
3202 Details : in rare race conditions a newly allocated group_info struct is
3203 freed again, and this can be NULL. The 2.4 compatibility code
3204 for groups_free() checked for a NULL pointer, but 2.6 did not.
3207 Frequency : common for liblustre clients doing little filesystem IO
3208 Bugzilla : 9352, 7313
3209 Description: server may evict liblustre clients accessing contended locks
3210 Details : if a client is granted a lock or receives a completion AST
3211 with a blocking AST already set it would not reply to the AST
3212 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3215 Frequency : lfs setstripe, only systems with more than 160 OSTs
3217 Description: unable to set striping with a starting offset beyond OST 160
3218 Details : llapi_create_file() incorrectly limited the starting stripe
3219 index to the maximum single-file stripe count.
3222 Frequency : LDAP users only
3224 Description: lconf did not handle in-kernel recovery with LDAP properly
3225 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3227 Severity : enhancement
3229 Description: bind OST threads to NUMA nodes to improve performance
3230 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3231 node and do their allocations there to localize memory access
3233 Severity : enhancement
3235 Description: llmount can determine client NID directly from Myrinet (GM)
3236 Details : the client NID code from gmnalnid was moved directly into
3237 llmount, removing the need to use this or specifying the
3238 client NID explicitly when mounting GM clients with zeroconf
3241 Frequency : if client is started with down MDS
3243 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3244 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3245 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3246 ptlrpc_import_delayed_req() because we didn't check if the
3247 request was interrupted, and we also didn't break out of the
3248 event loop if there was no timeout
3253 Description: data loss during non-page-aligned writes to a single file from
3254 both multiple nodes and multiple threads on one node at same time
3255 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3256 inconsistency led to false short-reads, that were cached and later
3257 used by ->prepare_write() to fill in partially written page,
3258 leading to data loss.
3261 Frequency : always, if lconf --abort_recovery used
3263 Description: lconf --abort_recovery fails with 'Operation not supported'
3264 Details : lconf was attempting to abort recovery on the MDT device and not
3267 Severity : enhancement
3269 Description: remove cleanup logs
3270 Details : replace lconf-generated cleanup logs with lustre internal
3271 cleanup routines. Eliminates the need for client-cleanup and
3274 Severity : enhancement
3276 Description: add support for EAs (user and system) on lustre filesystems
3277 Details : it is now possible to store extended attributes in the Lustre
3278 client filesystem, and with the user_xattr mount option it
3279 is possible to allow users to store EAs on their files also
3281 Severity : enhancement
3283 Description: Add possibility (config option) to show minimal available OST free
3285 Details : When compiled with --enable-mindf configure option, statfs(2)
3286 (and so, df) will return least minimal free space available from
3287 all OSTs as amount of free space on FS, instead of summary of
3288 free spaces of all OSTs.
3290 Severity : enhancement
3292 Description: do not expand extent locks acquired on OST-side
3293 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3294 by server: they are not cached anyway.
3297 Frequency : when mmap is used/binaries executed from Lustre
3299 Description: Unmmap pages before throwing them away from read cache.
3300 Details : llap_shrink cache now attempts to unmap pages before discarding
3301 them (if unmapping failed - do not discard). SLES9 kernel has
3302 extra checks that trigger if this unmapping is not done first.
3307 Description: lconf didn't resolve symlinks before checking to see whether a
3308 given mountpoint was already in use
3311 Frequency : when migrating failover services
3312 Bugzilla : 6395, 9514
3313 Description: When migrating a subset of services from a node (e.g. failback
3314 from a failover service node) the remaining services would
3315 time out and evict clients.
3316 Details : lconf --force (implied by --failover) sets the global obd_timeout
3317 to 5 seconds in order to quickly disconnect, but this caused
3318 other RPCs to time out too quickly. Do not change the global
3319 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3321 Severity : enhancement
3322 Frequency : if MDS is started with down OST
3323 Bugzilla : 9439,5706
3324 Description: Allow startup/shutdown of an MDS without depending on the
3325 availability of the OSTs.
3326 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3327 Add appropriate locking and lov-osc refcounts for safe
3328 cleaning. Add osc abort_inflight calls in case the
3329 synchronize never started.
3332 Frequency : occasional (Cray XT3 only)
3334 Description: root not authorized to access files in CRAY_PORTALS environment
3335 Details : The client process capabilities were not honoured on the MDS in
3336 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3337 been accessed by an authorized user then root was able to access
3338 the file on the local client also. The root user capabilities
3339 are now allowed on the MDS, as this environment has secure UID.
3342 Frequency : occasional
3344 Description: ldiskfs "too long searching" message happens too often
3345 Details : A debugging message (otherwise harmless) prints too often on
3346 the OST console. This has been reduced to only happen when
3347 there are fragmentation problems on the filesystem.
3352 Description: Division by zero in statfs when all OSCs are inactive
3353 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3354 lov_statfs divides by value returned by lov_get_stripecnt().
3358 Bugzilla : 9489, 3273
3359 Description: First write from each client to each OST was only 4kB in size,
3360 to initialize client writeback cache, which caused sub-optimal
3361 RPCs and poor layout on disk for the first writen file.
3362 Details : Clients now request an initial cache grant at (re)connect time
3363 and so that they can start streaming writes to the cache right
3364 away and always do full-sized RPCs if there is enough data.
3365 If the OST is rebooted the client also re-establishes its grant
3366 so that client cached writes will be honoured under the grant.
3371 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3372 Details : Now I/O RPCs go to different portal number and (presumably) fast
3373 lock requests (and glimses) and other RPCs get their own service
3374 threads pool that should be able to service those RPCs
3377 Severity : enhancement
3379 Description: Ability to exchange lustre version between client and servers and
3380 issue warnings at client side if client is too old. Also for
3381 liblustre clients there is ability to refuse connection of too old
3383 Details : New 'version' field is added to connect data structure that is
3384 filled with version info. That info is later checked by server and
3388 Frequency : rare, liblustre only.
3389 Bugzilla : 9296, 9581
3390 Description: Two simultaneous writes from liblustre at offset within same page
3391 might proceed at the same time overwriting eachother with stale
3393 Details : I/O lock withing llu_file_prwv was released too early, before data
3394 actually was hitting the wire. Extended lock-holding time until
3395 server acknowledges receiving data.
3398 Frequency : extremely rare. Never observed in practice.
3400 Description: avoid generating lustre_handle cookie of 0.
3401 Details : class_handle_hash() generates handle cookies by incrementing
3402 global counter, and can hit 0 occasionaly (this is unlikely, but
3403 not impossible, because initial value of cookie counter is
3404 selected randonly). Value of 0 is used as a sentinel meaning
3405 "unassigned handle" --- avoid it. Also coalesce two critical
3406 sections in this function into one.
3408 Severity : enhancement
3410 Description: allow liblustre clients to delegate truncate locking to OST
3411 Details : To avoid overhead of locking, liblustre client instructs OST to
3412 take extent lock in ost_punch() on client's behalf. New connection
3413 flag is added to handle backward compatibility.
3415 Severity : enhancement
3416 Bugzilla : 4928, 7341, 9758
3417 Description: allow number of OST service threads to be specified
3418 Details : a module parameter allows the number of OST service threads
3419 to be specified via "options ost ost_num_threads={N}" in the
3420 OSS's /etc/modules.conf or /etc/modprobe.conf.
3424 Bugzilla : 6146, 9635, 9895
3425 Description: servers crash with bad pointer in target_handle_connect()
3426 Details : In rare cases when a client is reconnecting it was possible that
3427 the connection request was the last reference for that export.
3428 We would temporarily drop the export reference and get a new
3429 one, but this may have been the last reference and the export
3430 was just destroyed. Get new reference before dropping old one.
3432 Severity : enhancement
3433 Frequency : if client is started with failover MDS
3435 Description: Allow multiple MDS hostnames in the mount command
3436 Details : Try to read the configuration from all specified MDS
3437 hostnames during a client mount in case the "primary"
3440 Severity : enhancement
3442 Description: Stop sending data to evicted clients as soon as possible.
3443 Details : Check if the client we are about to send or are sending data to
3444 was evicted already. (Check is done every second of waiting,
3445 for which l_wait_event interface was extended to allow checking
3446 of exit condition at specified intervals).
3449 Frequency : rare, normally only when NFS exporting is done from client
3451 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3453 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3456 Frequency : always, if srand() is called before liblustre initialization
3458 Description: Liblustre uses system PRNG disturbing its usage by user application
3459 Details : Introduce internal to lustre fast and high-quality PRNG for
3460 lustre usage and make liblustre and some other places in generic
3461 lustre code to use it.
3463 Severity : enhancement
3464 Bugzilla : 9477, 9557, 9870
3465 Description: Verify that the MDS configuration logs are updated when xml is
3466 Details : Check if the .xml configuration logs are newer than the config
3467 logs stored on the MDS and report an error if this is the case.
3468 Request --write-conf, or allow starting with --old_conf.
git://git.whamcloud.com / fs / lustre-release.git / blob