3 2.6.5-7.287.3 (SLES 9),
4 2.6.9-55.0.12.EL (RHEL 4),
5 2.6.16.53-0.8 (SLES 10),
6 2.6.18-8.1.14.EL5 (RHEL 5).
7 2.6.18.8 vanilla (kernel.org)
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.9-42.0.10.EL (RHEL 4),
11 2.6.16 - 2.6.21 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.40.2-cfs1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Update to SLES9 kernel-2.6.5-7.287.3.
21 Description: Update to RHEL4 latest kernel.
25 Description: Remove 2.4 kernel support from lustre 1.8.0
26 Details : Remove 2.5.0 check from quota_check.c, quota_ctl.c,
27 quota_interface.c in quota subsystem.
28 Remove 2.4 kernel check from obd_mount.c, linux/linux-sysctl.c in
30 Remove lock_24kernel from fsfilt_ext3.c fsfilt_reiserfs.c
31 lvfs_linux.c upcall_cache.c in lvfs subsystem.
32 Remove 2.5.0 check from dcache.c dir.c file.c llite_internal.h
33 llite_lib.c llite_mmap.c lproc_llite.c namei.c rw.c symlink.c
35 Remove 2.5.0 check from handler.c lproc_mds.c mds_fs.c
36 mds_internal.h mds_lib.c mds_open.c mds_reint.c mds_xattr.c
38 Remove 2.4 kernel check from lproc_osc.c, osc_request.c in osc
40 Remove 2.4 kernel check from filter.c, lproc_obdfilter.c in
42 Remove 2.4 kernel code from lustre/obdecho/echo_client.c.
43 Remove 2.4 kernel check from lustre/mgs/lproc_mgs.c,
45 Remove 2.4 kernel code from lustre/mgc/mgc_request.c.
46 Remove 2.4 kernel check from lustre/mdt/mdt_lproc.c.
47 Remove 2.4 kernel code from lustre/mdc/mdc_request.c.
48 Remove 2.4 kernel check from lustre/lov/lproc_lov.c.
49 Remove 2.4 kernel check from lustre/lmv/lproc_lmv.c.
50 Remove lock_24kernel from lustre/lvfs/fsfilt_ext3.c.
51 Remove 2.4 kernel code from include/liblustre.h,
52 include/lustre_disk.h, include/lustre_quota.h,
53 include/linux/lprocfs_status.h, include/linux/lvfs.h,
54 include/linux/lustre_lite.h, include/linux/lustre_net.h,
55 include/linux/lustre_user.h, include/linux/lvfs_linux.h,
56 include/linux/obd_class.h, include/linux/lustre_compat25.h.
58 Severity : enhancement
60 Description: Build SLES10 patchless client fails
61 Details : The configure was broken by run ./configure with
62 --with-linux-obj=.... argument for patchless client. When the
63 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
64 can't search header adequately. Use absolute path such as
65 -I($LINUX)/include instead.
69 Description: interrupt oig_wait produce painc on resend.
70 Details : brw_redo_request can be used for resend requests from ptlrpcd and
71 private set, and this produce situation when rq_ptlrpcd_data not
72 copyed to new allocated request and triggered LBUG on assert
73 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
74 ptlrpcd set if request is changed and can be safety changed to use
79 Description: LASSERT_{REQ,REP}SWAB macros are buggy
80 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
81 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
82 these macros and replace them with their difinitions instead.
86 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
87 Details : Update which_patch & target file for SLES10 latest kernel.
90 Description: Remove client patches from SLES 10 kernel.
91 Details : This causes SLES 10 clients to behave as patchless clients
92 even on a Lustre-patched (server) kernel.
94 Severity : enhancement
96 Description: self-adjustable client's lru lists
97 Details : use adaptive algorithm for managing client cached locks lru
98 lists according to current server load, other client's work
99 pattern, memory activities, etc. Both, server and client
100 side namespaces provide number of proc tunables for controlling
103 Severity : enhancement
105 Description: light-weight GSS support
106 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
107 and reduce performance overhead.
109 Severity : enhancement
111 Description: Linux keyring support
112 Details : Support using service of Linux keyring for Lustre GSS internal
113 context refresh/cache mechanism.
117 Description: Fix errors in lfs documentation
118 Details : Fixes man pages
122 Description: don't use GFP_* in generic Lustre code.
123 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
127 Description: obdclass is limited by single OBD_ALLOC(idarray)
128 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
132 Description: warnings with build patchless client with vanila 2.6.19 and up
133 Details : change old ctl_table style and replace ctl_table/ctl_table_header
134 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
138 Frequency : Only for SLES
139 Description: sanity_quota fail test_1
140 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
141 file, which makes TSTID[2] inunique.
145 Frequency : Only for SLES9
146 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
147 Details : Update target/ChangeLog/which_patch .
151 Frequency : Only for test_10 in sanity-quota.sh
152 Description: a bug in quota test code
153 Details : checking a test flag in a wrong place
157 Frequency : Only for test_13 in sanity-quota.sh
158 Description: a bug in quota test script
159 Details : shouldn't check overfull allocated quota
163 Description: jbd statistics
164 Details : Port older jbd statistics patch for sles10
166 Severity : enhancement
168 Description: Add printing inode info into message about error in writepage.
173 Description: reply_lock_interpret crash due to race with it and lock cancel.
174 Details : Do not replay locks that are being cancelled. Do not reference
175 locks by their address during replay, just by their handle.
179 Frequency : When flocks are used.
180 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
181 Details : Flock locks might destroy just granted lock if it could be merged
182 with another existing flock, this is done in completion handler,
183 so teach ldlm_cli_enquque_fini that this is a valid case for
189 Description: Oops in read and write path when failing to allocate lock.
190 Details : Check if lock allocation failed and return error back.
194 Description: lstripe command fails for valid OST index
195 Details : The stripe offset is compared to lov->desc.ld_tgt_count
196 instead of lov->desc.ld_active_tgt_count.
200 Description: sanity.sh failed test 103
201 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
206 Description: df doesn't work properly if diskfs blocksize != 4K
207 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
211 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
212 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
216 Description: obd_health_check_timeout too short
217 Details : set obd_health_check_timeout as 1.5x of obd_timeout
221 Description: llapi_file_create() does not allow some changes
222 Details : add llapi_file_open() that allows specifying the mode and
223 open flags, and also returns an open file handle.
227 Description: merge and cleanup kernel patches.
228 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
232 Description: Add journal checksum support.(Kernel part)
233 Details : The journal checksum feature adds two new flags i.e
234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
235 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
236 indicates that the commit block contains the checksum for
237 the blocks described by the descriptor blocks. Now commit
238 record can be sent to disk without waiting for descriptor
239 blocks to be written to disk. This behavior is controlled
240 using JBD2_FEATURE_ASYNC_COMMIT flag.
244 Description: OSS needs mutliple precreate threads
245 Details : Add ability to start more than one create thread per OSS.
249 Description: Sanity.sh test_65k failed: setstripe should have succeeded
250 Details : execute lfs setstripe on client
254 Description: mds_obd_create error creating tmp object
255 Details : When the user sets quota on root, llog will be affected and can't
256 create files and write files.
259 Frequency : Always on ia64 patchless client, and possibly others.
261 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
262 Details : This allows the patchless client to be loaded on architectures
267 Description: RedHat Update kernel for RHEL5
268 Details : Add the kernel config file for RHEL5.
272 Description: RedHat Update kernel for RHEL5
273 Details : Modify the kernel config file more closer RHEL5.
277 Description: Build failure against Centos5 (RHEL5)
278 Details : Define PAGE_SIZE when it isn't present.
282 Description: test_7 on sanity-quota.sh can't be used on separate machines
283 Details : add facet to handle it
287 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
288 Details : with NFS, the anon dentry's parent was set to itself in
289 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
290 corresponding dentry other than use rec->ur_name.
292 Severity : enhancement
294 Description: lfs setstripe enhancement
295 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
299 Description: testing performance impact of enabling checksumming
300 Details : enable checksum by default, allow --disable-checksum
301 configure option and "-o nochecksum" mount option
305 Description: lustre support for RHEL5
306 Details : Add support for RHEL5.
310 Description: Client eviction due to failover config
311 Details : after a connection loss, the lustre client should attempt to
312 reconnect to the last active server first before trying the
313 other potential connections.
317 Description: when mds and osts use different quota unit(32bit and 64bit),
318 quota will be released repeatly.
319 Details : void sending multiple quota reqs to mds, which will keep the
320 status between the reqs.
324 Description: osts not allocated evenly to files
325 Details : change the condition to increase offset_idx
328 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
330 Description: Data corruption for OSTs that are formatted larger than 2TB
332 Details : When generating the bio request for lustre file writes the
333 sector number would overflow a temporary variable before being
334 used for the IO. The data reads correctly from Lustre (which
335 will overflow in a similar manner) but other file data or
336 filesystem metadata may be corrupted in some cases.
340 Description: Tune the kernel for good SCSI performance.
341 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
342 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
347 Description: rewrite ext2-derived code in obdclass/uuid.c
348 Details : rewrite inherited code (uuid parsing code from ext2 utils)
349 from scratch preserving functionality.
353 Description: "lfs find -obd UUID" prints directories
354 Details : "lfs find -obd UUID" will return all directory names instead
355 of just file names. It is incorrect because the directories
356 do not reside on the OSTs.
360 Description: use i_size_read and i_size_write in 2.6 port
361 Details : replace inode->i_size access with i_size_read/write()
364 Frequency : when removing large files
366 Description: scheduling issue during removal of large Lustre files
367 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
368 It causes scheduling issues when removing large files (17TB in the
372 Frequency : only with liblustre clients on XT3
374 Description: evictions taking too long
375 Details : allow llrd to evict clients directly on OSTs
378 Frequency : only on ppc
380 Description: /proc/fs/lustre/devices broken on ppc
381 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
382 We should make sure the type of 'index' is loff_t and then cast
383 explicitly as needed below. Do not assign an explicitly cast
387 Frequency : only for rhel5
389 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
390 Details : Modify the target file & which_kernel.
392 Severity : enhancement
394 Description: omit set fsid for export NFS
395 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
396 as FS_REQUIRES_DEV for avoid problems with generate fsid.
400 Frequency : Always, for kernels after 2.6.16
401 Description: Fix warning idr_remove called for id=.. which is not
403 Details : Last kernels save old s_dev before kill super and not allow
404 to restore from callback - restore it before call kill_anon_super.
408 Description: buffer overruns could theoretically occur
409 Details : llapi_semantic_traverse() modifies the "path" argument by
410 appending values to the end of the origin string, and a
411 overrun may occur. Adding buffer overrun check in liblustreapi.
415 Description: Fix error on 'ls .' at the top of the Lustre mount.
416 Details : Don't revalidate dentry if it is a root dentry.
420 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
421 Details : Modify vm-tunables-rhel4.patch.
425 Description: Kernel config for 2.6.18-vanilla.
426 Details : Modify targets/2.6-vanilla.target.in.
427 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
428 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
429 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
430 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
434 Description: improve handling recoverable errors
435 Details : if request processig with error which can be recoverable on server
436 request should be resend, otherwise page released from cache and
441 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
442 Details : Modify target file & which_patch.
443 A flaw was found in the IA32 system call emulation provided
444 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
445 value could be stored in the %RAX register, which could trigger an
446 out-of-bounds system call table access. An untrusted local user
447 could exploit this flaw to run code in the kernel
448 (ie a root privilege escalation). (CVE-2007-4573).
452 Description: O_DIRECT bypasses client statistics.
453 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
454 read_ahead_stats were updated. copied the stats section from
455 osc_send_oap_rpc() into async_internal().
459 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
463 Description: change order of libsysio includes
464 Details : '#include sysio.h' should always come before '#include xtio.h'
468 Description: handle "serious error: objid * already exists" more gracefully
469 Details : If LAST_ID value on disk is smaller than the objects existing in
470 the O/0/d* directories, it indicates disk corruption and causes an
471 LBUG(). If the object is 0-length, then we should use the existing
472 object. This will help to avoid a full fsck in most cases.
474 Severity : enhancement
476 Description: adapt the lustre_config script to support the upgrade case
477 Details : Add "-u" option for lustre_config script to support upgrading 1.4
478 server targets to 1.6 in parallel.
482 Description: To avoid grant space > avaible space when the disk is almost
483 full. Without this patch you might see the error "grant XXXX >
484 available" or some LBUG about grant, when the disk is almost
486 Details : In filter_check_grant, for non_grant cache write, we should
487 check the left space by if (*left > ungranted + bytes), instead
488 of (*left > ungranted), because only we are sure the left space
489 is enough for another "bytes", then the ungrant space should be
490 increase. In client, we should update cl_avail_grant only there
491 is OBD_MD_FLGRANT in the reply.
495 Description: Update RHEL 4 kernel to fix local root privilege escalation.
496 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
497 described in CVE-2007-4573. This problem could allow untrusted
498 local users to gain root access.
502 Frequency : when using O_DIRECT and quotas
504 Description: Incorrect file ownership on O_DIRECT output files
505 Details : block usage reported by 'lfs quota' does not take into account
506 files that have been written with O_DIRECT.
508 --------------------------------------------------------------------------------
510 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
512 * Support for kernels:
513 2.6.5-7.283 (SLES 9),
514 2.6.9-55.EL (RHEL 4),
515 2.6.16.46-0.14 (SLES 10),
516 2.6.18.8 vanilla (kernel.org)
517 * Client support for unpatched kernels:
518 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
519 2.6.16 - 2.6.22 vanilla (kernel.org)
520 * Due to recently discovered recovery problems, we do not recommend
521 using patchless RHEL 4 clients with this or any earlier release.
522 * Recommended e2fsprogs version: 1.39.cfs8
523 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
524 * Starting with this release, the ldiskfs backing filesystem required
525 by Lustre is now in its own package, lustre-ldiskfs. This package
526 should be installed. It is versioned separately from Lustre and
527 may be released separately in future.
531 Description: block reactivating mgc import until all deactivates complete
532 Details : Fix race when failing back MDT/MGS to itself (testing)
535 Frequency : at statup only
537 Description: mds_lov_synchronize race leads to various problems
538 Details : simultaneous MDT->OST connections at startup can cause the
539 sync to abort, leaving the OSC in a bad state.
541 Severity : enhancement
543 Description: add optional extra BUILD_VERSION info
544 Details : add a new environment variable (namely LUSTRE_VERS) which allows
545 to override the lustre version.
548 Frequency : 2.6.18 servers only
550 Description: ll_kern_mount() doesn't release the module reference
551 Details : The ldiskfs module reference count never drops down to 0
552 because ll_kern_mount() doesn't release the module reference.
557 Description: server LBUG when using old ost_num_threads parameter
558 Details : Accept the old ost_num_threads parameter but warn that it
559 is deprecated, and fix an off-by-one error that caused an LBUG.
564 Description: Transient SCSI error results in persistent IO issue
565 Details : iobuf->dr_error is not reinitialized to 0 between two
569 Frequency : sometimes when underlying device returns I/O errors
571 Description: OSTs not going read-only during write failures
572 Details : OSTs are not remounted read-only when the journal commit threads
573 get I/O errors because fsfilt_ext3 calls journal_start/stop()
574 instead of the ext3 wrappers.
578 Description: poor connect scaling with increasing client count
579 Details : Don't run filter_grant_sanity_check for more than 100 exports
580 to improve scaling for large numbers of clients.
583 Frequency : SLES10 only
585 Description: sanity-quota.sh quotacheck failed: rc = -22
586 Details : Quotas cannot be enabled on SLES10.
589 Frequency : liblustre clients only
591 Description: getdirentries does not give error when run on compute nodes
592 Details : getdirentries does not fail when the size specified as an argument
593 is too small to contain at least one entry
595 Severity : enhancement
597 Description: Add LNET router traceability for debug purposes
598 Details : If a checksum failure occurs with a router as part of the
599 IO path, the NID of the last router that forwarded the bulk data
600 is printed so it can be identified.
605 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
606 Details : Due to a race condition, liblustre clients were occasionally
609 Severity : enhancement
611 Description: lfs setstripe use optional parameters instead of postional
614 Severity : enhancement
616 Description: Nanosecond timestamp support for ldiskfs
617 Details : The on-disk ldiskfs filesystem has added support for nanosecond
618 resolution timestamps. There is not yet support for this at
619 the Lustre filesystem level.
622 Frequency : during server recovery
624 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
625 Details : request with rq_no_resend flag not awake l_wait_event if they get a
629 Frequency : nfs export on patchless client
631 Description: connectathon hang when test nfs export over patchless client
632 Details : Disconnected dentry cannot be found with lookup, so we do not need
633 to unhash it or make it invalid
636 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
637 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
638 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
643 Description: open req refcounting wrong on reconnect
644 Details : If reconnect happened between getting open reply from server and
645 call to mdc_set_replay_data in ll_file_open, we will schedule
646 replay for unreferenced request that we are about to free.
647 Subsequent close will crash in variety of ways.
648 Check that request is still eligible for replay in
649 mdc_set_replay_data().
654 Description: disable writes to filesystem when reading health_check file
655 Details : the default for reading the health_check proc file has changed
656 to NOT do a journal transaction and write to disk, because this
657 can cause reads of the /proc file to hang and block HA state
658 checking on a healthy but otherwise heavily loaded system. It
659 is possible to return to the previous behaviour during configure
660 with --enable-health-write.
662 Severity : enhancement
664 Description: 64-bit inode version
665 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
666 to the inode. This will be required for version-based recovery.
671 Description: MDS fails to start if a duplicate client export is detected
672 Details : in some rare cases it was possible for a client to connect to
673 an MDS multiple times. Upon recovery the MDS would detect this
674 and fail during startup. Handle this more gracefully.
676 Severity : enhancement
678 Description: Add -o localflock option to simulate old noflock
680 Details : This will achieve local-only flock/fcntl locks
686 Description: log_commit_thread vs filter_destroy race leads to crash
687 Details : Take import reference before releasing llog record semaphore
692 Description: Wrong request locking in request set processing
693 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
694 imp_delayed_list, in this place should be used imp_lock.
697 Frequency : when reconnection
699 Description: Grant Leak when osc reconnect to OST
700 Details : When osc reconnect ost, OST(filter) should check whether it
701 should grant more space to client by comparing fed_grant and
702 cl_avail_grant, and return the granted space to client instead
703 of "new granted" space, because client will call osc_init_grant
704 to update the client grant space info.
707 Frequency : when client reconnect to OST
709 Description: Grant Leak when osc do resend and replay bulk write
710 Details : When osc reconnect to OST, OST(filter)should clear grant info of
711 bulk write request, because the grant info while be sync between
712 OSC and OST when reconnect, and we should ignore the grant info
713 these of resend/replay write req.
718 Description: Grant space more than avaiable left space sometimes.
719 Details : When then OST is about to be full, if two bulk writing from
720 different clients came to OST. Accord the avaliable space of the
721 OST, the first req should be permitted, and the second one
722 should be denied by ENOSPC. But if the seconde arrived before
723 the first one is commited. The OST might wrongly permit second
724 writing, which will cause grant space > avaiable space.
727 Frequency : when client is evicted
729 Description: Grant might be wrongly erased when osc is evicted by OST
730 Details : when the import is evicted by server, it will fork another
731 thread ptlrpc_invalidate_import_thread to invalidate the
732 import, where the grant will be set to 0. While the original
733 thread will update the grant it got when connecting. So if
734 the former happened latter, the grant will be wrongly errased
735 because of this race.
740 Description: Checking Stale with correct fid
741 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
742 to check whether it is stale, because sometimes, we want the
743 enqueue happened anyway, and op_data.fid2 will not be initialized.
745 Severity : enhancement
747 Description: update patchless client
748 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
751 Frequency : only with 2.4 kernel
753 Description: random memory corruption
754 Details : size of struct ll_inode_info is to big for union inode.u and this
755 can be cause of random memory corruption.
760 Description: Memory leak in recovery
761 Details : Lov_mds_md was not free in an error handler in mds_create_object.
762 It should also check obd_fail before fsfilt_start, otherwise if
763 fsfilt_start return -EROFS,(failover mds during mds recovery).
764 then the req will return with repmsg->transno = 0 and rc = EROFS.
765 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
766 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
767 be freed no matter whether fsfilt_commit success or not.
770 Frequency : only with huge count clients
772 Description: Prevents from taking the superblock lock in llap_from_page for
774 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
775 for llap_from_page call in ll_removepage prevents from taking the
776 superblock lock for a soon died page.
781 Description: Not check open intent error before release open handle
782 Details : in some rare cases, the open intent error is not checked before
783 release open handle, which may cause
784 ASSERTION(open_req->rq_transno != 0), because it tries to release
785 the failed open handle.
790 Description: Set cat log bitmap only after create log success.
791 Details : in some rare cases, the cat log bitmap is set too early. and it
792 should be set only after create log success.
796 Description: Accessing a block bevice can re-enable I/O when Lustre is
797 tearing down a device.
798 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
803 Description: service threads may hog cpus when there are a lot of requests
805 Details : Insert cond_resched to give other threads a chance to use some of
811 Description: the cat log was not initialized in recovery
812 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
813 unlink log on mds will not be initialized until mds post
814 recovery. And also in mds post recovery, the unlink log will
815 initialization will be done asynchronausly, so there will be race
816 between add unlink log and unlink log initialization.
820 Description: brw_stats were being printed incorrectly
821 Details : brw_stats were being printed as log2 but all of them were not
822 recorded as log2. Also remove some code duplication arising from
823 filter_tally_{read,write}.
827 Frequency : rare, only in recovery.
828 Description: ASSERTION(req->rq_type != LI_POISON) failed
829 Details : imp_lock should be held while iterating over imp_sending_list for
830 prevent destroy request after get timeout in ptlrpc_queue_wait.
834 Description: replay-single.sh test 52 fails
835 Details : A lock's skiplist need to be cleanup when it being unlinked
836 from its resource list.
840 Description: Short directio read returns full requested size rather than
842 Details : Direct I/O operations should return actual amount of bytes
843 transferred rather than requested size.
845 Severity : enhancement
847 Description: metadata RPC reduction (e.g. for rm performance)
848 Details : decrease the amount of synchronous RPC between clients and servers
849 by canceling conflicing lock before the operation on the client side
850 and packing thier handles into the main operation RPC to server.
852 Severity : enhancement
854 Description: Async OSC create to avoid the blocking unnecessarily.
855 Details : If a OST has no remain object, system will block on the creating
856 when need to create a new object on this OST. Now, ways use
857 pre-created objects when available, instead of blocking on an
858 empty osc while others are not empty. If we must block, we block
859 for the shortest possible period of time.
861 --------------------------------------------------------------------------------
863 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
868 Frequency : on some architectures
870 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
871 Details : Uninitialized flags sometimes cause configuration commands to
875 Frequency : patchless clients only
877 Description: missing __iget() symbol export
878 Details : The __iget() symbol export is missing. To avoid the need for
879 this on patchless clients the deathrow inode reaper is turned
880 off, and we depend on the VM to clean up old inodes. This
881 dependency was during via the fix for bug 12181.
883 --------------------------------------------------------------------------------
885 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
887 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
888 INTEROPERATE with older versions automatically. In many cases a
889 special upgrade step is needed. Please read the
890 user documentation before upgrading any part of a live system.
891 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
892 version will not interoperate with 1.6 betas before beta5 (1.5.95).
893 * WARNING: Lustre configuration and startup changes are required with
894 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
900 Severity : enhancement
902 Description: MountConf
903 Details : Lustre configuration is now managed via mkfs and mount
904 commands instead of lmc and lconf. New obd types (MGS, MGC)
905 are added for dynamic configuration management. See
906 https://mail.clusterfs.com/wikis/lustre/MountConf for
909 Severity : enhancement
911 Description: dynamic OST addition
912 Details : OSTs can now be added to a live filesystem
914 Severity : enhancement
916 Description: startup order invariance
917 Details : MDTs and OSTs can be started in any order. Clients only
918 require the MDT to complete startup.
920 Severity : enhancement
922 Description: parallel, asynchronous orphan cleanup
923 Details : orphan cleanup is now performed in separate threads for each
924 OST, allowing parallel non-blocking operation.
926 Severity : enhancement
928 Description: optimized stripe assignment
929 Details : stripe assignments are now made based on ost space available,
930 ost previous usage, and OSS previous usage, in order to try
931 to optimize storage space and networking resources.
933 Severity : enhancement
935 Description: Permanently set tunables
936 Details : All writable /proc/fs/lustre tunables can now be permanently
937 set on a per-server basis, at mkfs time or on a live
940 Severity : enhancement
942 Description: Lustre message v2
943 Details : Add lustre message format v2.
945 Severity : enhancement
947 Description: client OST exclusion list
948 Details : Clients can be started with a list of OSTs that should be
949 declared "inactive" for known non-responsive OSTs.
953 Description: SPEC SFS validation failure on NFS v2 over lustre.
954 Details : Changes the blocksize for regular files to be 2x RPC size,
955 and not depend on stripe size.
957 Severity : enhancement
959 Description: Multiple MD RPCs in flight.
960 Details : Further unserialise some read-only MDS RPCs - learn about intents.
961 To avoid overly-overloading MDS, introduce a limit on number of
962 MDS RPCs in flight for a single client and add /proc controls
963 to adjust this limit.
965 Severity : enhancement
967 Description: client read/write statistics
968 Details : Add client read/write call usage stats for performance
969 analysis of user processes.
970 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
971 file access. extents_stats shows chunk size distribution.
972 extents_stats_per_process show chunk size distribution per
975 Severity : enhancement
977 Description: mds statistics
978 Details : Add detailed mds operations statistics in
979 /proc/fs/lustre/mds/*/stats.
983 Description: Failure of copying files with lustre special EAs.
984 Details : Client side always return success for setxattr call for lustre
985 special xattr (currently only "trusted.lov").
990 Description: Refcount LNET uuids
991 Details : The global LNET uuid list grew linearly with every startup;
992 refcount repeated list entries instead of always adding to
995 Severity : enhancement
997 Description: Dynamic service threads
998 Details : Within a small range, start extra service threads
999 automatically when the request queue builds up.
1002 Frequency : mixed-endian client/server environments
1004 Description: mixed-endian crashes
1005 Details : The new msg_v2 system had some failures in mixed-endian
1008 Severity : enhancement
1010 Description: Easy OST removal
1011 Details : OSTs can be permanently deactivated with e.g. 'lctl
1012 conf_param lustre-OST0001.osc.active=0'
1014 Severity : enhancement
1016 Description: MGS proc entries
1017 Details : Added basic proc entries for the MGS showing what filesystems
1020 Severity : enhancement
1022 Description: provide MGS failover
1023 Details : Added config lock reacquisition after MGS server failover.
1025 Severity : enhancement
1027 Description: add Linux 2.4 support
1028 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1032 Description: a large application tries to do I/O to the same resource and dies
1033 in the middle of it.
1034 Details : Check the req->rq_arrival time after the call to
1035 ost_brw_lock_get(), but before we do anything about
1036 processing it & sending the BULK transfer request. This
1037 should help move old stale pending locks off the queue as
1038 quickly as obd_timeout.
1041 Frequency : when an incorrect nid is specified during startup
1043 Description: ptlrpc connect to non-existant node causes kernel crash
1044 Details : LNET can't be re-entered from an event callback, which
1045 happened when we expire a message after the export has been
1046 cleaned up. Instead, hand the zombie cleanup off to another
1049 Severity : enhancement
1051 Description: plain/inodebits lock performance improvement
1052 Details : Grouping plain/inodebits in granted list by their request modes
1053 and bits policy, thus improving the performance of search through
1057 Frequency : only if OST filesystem is corrupted
1059 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1060 Details : for a short time RPCs with bulk IO are in the replay list,
1061 but replay of bulk IOs is unimplemented. If the OST filesystem
1062 is corrupted due to disk cache incoherency and then replay is
1063 started it is possible to trip an assertion. Avoid putting
1064 committed RPCs into the replay list at all to avoid this issue.
1067 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1069 Description: System hang on startup
1070 Details : This bug allowed the liblustre (e.g. catamount) client to
1071 return to the app before handling all startup RPCs. This
1072 could leave the node unresponsive to lustre network traffic
1073 and manifested as a server ptllnd timeout.
1075 Severity : enhancement
1077 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1078 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1079 boolean to control whether to print peer debug info when a
1080 client's RPC times out.
1083 Frequency : only for kernels with patches from Lustre below 1.4.3
1085 Description: Remove old rdonly API
1086 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1089 Frequency : only for devices with external journals
1091 Description: Set external device read-only also
1092 Details : During a commanded failover stop, we set the disk device
1093 read-only while the server shuts down. We now also set any
1094 external journal device read-only at the same time.
1097 Frequency : when upgrading from 1.4 while trying to change parameters
1099 Description: The wrong (new) MDC name was used when setting parameters for
1100 upgraded MDT's. Also allows changing of OSC (and MDC)
1101 parameters if --writeconf is specified at tunefs upgrade time.
1104 Frequency : when setting specific ost indicies
1106 Description: QOS code breaks on skipped indicies
1107 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1108 created with --index need not be sequential.
1110 Severity : enhancement
1112 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1113 Details : The uninit_groups feature works in conjunction with the kernel
1114 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1115 up the pass1 processing of e2fsck. This is a read-only feature
1116 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1117 mount filesystems that have had this feature enabled.
1119 Severity : enhancement
1121 Description: Improve multi-block allocation algorithm to avoid fragmentation
1122 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1123 allocation locality and avoid filesystem fragmentation.
1125 ------------------------------------------------------------------------------
1127 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1129 * Support for kernels:
1130 2.4.21-47.0.1.EL (RHEL 3)
1131 2.6.5-7.283 (SLES 9)
1132 2.6.9-42.0.10.EL (RHEL 4)
1133 2.6.12.6 vanilla (kernel.org)
1134 2.6.16.27-0.9 (SLES 10)
1135 * Recommended e2fsprogs version: 1.39.cfs5
1137 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1141 Frequency : occasional, depends on client load and configuration
1142 Bugzilla : 12181, 12203
1143 Description: data loss for recently-modified files
1145 Details : In some cases it is possible that recently written or created
1146 files may not be written to disk in a timely manner (this should
1147 normally be within 30s unless client IO load is very high).
1148 The problem appears as zero-length files or files that are a
1149 multiple of 1MB in size after a client crash or client eviction
1150 that are missing data at the end of the file.
1152 This problem is more likely to be hit on clients where files are
1153 repeatedly created and unlinked in the same directory, clients
1154 have a large amount of RAM, have many CPUs, the filesystem has
1155 many OSTs, the clients are rebooted frequently, and/or the files
1156 are not accessed by other nodes after being written.
1158 The presence of the problem can be detected by looking at
1159 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1160 smaller than the second (nr_unused) then dirty files will not
1161 be flushed automatically to disk. "sync; sleep 10" should be
1162 run several times on the node before unmounting it to update
1163 Lustre (this is also safe to run on nodes without this problem).
1165 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1166 that can cause this same problem, so customers using that kernel
1167 also need to update the kernel in addition to Lustre. In order
1168 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1170 It is normal that files written just before a client crash (less
1171 than 30s) may not yet have been flushed to disk, even for local
1175 Frequency : frequent on thin XT3 nodes
1177 Description: UUID collision on thin XT3 Linux nodes
1178 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1179 randomly, since we relied on an insufficiently-seeded PRNG.
1184 Description: OSS hangs after "All ost request buffers busy"
1185 Details : A deadlock between quota and journal operations caused OSS
1186 hangs after printing "All ost request buffers busy."
1189 Frequency : always on liblustre builds
1191 Description: Cleanup compiler warnings on liblustre
1194 Frequency : always on liblustre builds on XT3
1196 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1200 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1201 > 32000 subdirectories
1202 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1203 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1204 subdirectory count crosses 32000. This will aid e2fsck to
1205 correctly handle more than 32000 subdirectories.
1208 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1210 Description: System hang on startup
1211 Details : This bug allowed the liblustre (e.g. catamount) client to
1212 return to the app before handling all startup RPCs. This
1213 could leave the node unresponsive to lustre network traffic
1214 and manifested as a server ptllnd timeout.
1216 Severity : enhancement
1218 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1219 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1220 boolean to control whether to print peer debug info when a
1221 client's RPC times out.
1226 Description: make O_SYNC working on 2.6 kernels
1227 Details : 2.6 kernels use different method for mark pages for write,
1228 so need add a code to lustre for O_SYNC work.
1233 Description: Failure to close file and release space on NFS
1234 Details : Put inode details into lock acquired in ll_intent_file_open.
1235 Use mdc_intent_lock in ll_intent_open to properly
1236 detect all kind of errors unhandled by mdc_enqueue.
1241 Description: proc file read during shutdown sometimes raced obd removal,
1243 Details : Add lock to prevent obd access after proc file removal.
1246 Frequency : Only for files larger than 4GB on 32-bit clients.
1248 Description: improperly doing page alignment of locks
1249 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1250 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1255 Description: RPCs being resent when they shouldn't be
1256 Details : Some RPCs that should not be resent are being resent. This
1257 can cause inconsistencies in the RPC state machine. Do not
1258 resend such requests.
1261 Frequency : rare, only with NFS export
1263 Description: Crash on NFS re-export node
1264 Details : under very unusual load conditions an assertion is hit in
1265 ll_intent_file_open()
1268 Frequency : only if OST filesystem is corrupted
1270 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1271 Details : for a short time RPCs with bulk IO are in the replay list,
1272 but replay of bulk IOs is unimplemented. If the OST filesystem
1273 is corrupted due to disk cache incoherency and then replay is
1274 started it is possible to trip an assertion. Avoid putting
1275 committed RPCs into the replay list at all to avoid this issue.
1280 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1281 Details : Large single O_DIRECT read and write calls can fail to allocate
1282 a sufficiently large buffer to process the request. In case of
1283 allocation failure the allocation is retried with a smaller
1284 buffer and broken into smaller requests.
1286 Severity : enhancement
1288 Description: Add -o localflock option to simulate old noflock behaviour.
1289 Details : This will achieve local-only flock/fcntl locks coherentness.
1294 Description: versioning check is incomplete
1295 Details : Checking the version difference of client vs. server, report
1296 error if the gap is too big.
1301 Description: add support PG_writeback bit
1302 Details : add support for PG_writeback bit for Lustre, for more carefull
1303 work with page cache in 2.6 kernel. This also fix some deadlocks
1304 and remove hack for work O_SYNC with 2.6 kernel.
1306 Severity : enhancement
1308 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1309 Details : The uninit_groups feature works in conjunction with the kernel
1310 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1311 up the pass1 processing of e2fsck. This is a read-only feature
1312 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1313 mount filesystems that have had this feature enabled.
1315 Severity : enhancement
1317 Description: Improve multi-block allocation algorithm to avoid fragmentation
1318 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1319 allocation locality and avoid filesystem fragmentation.
1321 ------------------------------------------------------------------------------
1323 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1325 * Support for kernels:
1326 2.6.9-42.0.3.EL (RHEL 4)
1327 2.6.5-7.276 (SLES 9)
1328 2.4.21-47.0.1.EL (RHEL 3)
1329 2.6.12.6 vanilla (kernel.org)
1330 2.6.16.21-0.8 (SLES10)
1331 * Recommended e2fsprogs version: 1.39.cfs2-0
1333 * The backwards-compatible /proc/sys/portals symlink has been removed
1334 in this release. Before upgrading, please ensure that you change
1335 any configuration scripts or /etc/sysctl.conf files that access
1336 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1337 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1338 in advance of the upgrade on any system running Lustre 1.4.6 or
1339 newer, since /proc/sys/lnet was added in that version.
1340 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1344 Frequency : only when quota is used
1346 Description: avoid scanning export list for quota master
1347 Details : Change the algorithms to avoid scanning export list in order
1348 to improve the efficiency.
1351 Frequency : MDS failover only, very rarely
1353 Description: "went back in time" messages on mds failover
1354 Details : The greatest transno may be lost when the current operation
1355 finishes with an error (transno==0) and the client's last_rcvd
1356 record is over-written. Save the greatest transno in the
1357 mds_last_transno for this case.
1360 Frequency : always for specific kernels and striping counts
1362 Description: client may get "Matching packet too big" without ACL support
1363 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1364 when trying to access files in certain configurations. The
1365 clients should in fact be denied when mounting because they do
1366 not understand ACLs.
1369 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1371 Description: many clients connecting with IO in progress causes connect timeouts
1372 Details : Avoid synchronous journal commits to avoid delays caused by many
1373 clients connecting/disconnecting when bulk IO is in progress.
1374 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1375 OST_IO_PORTAL to avoid delays behind potentially many pending
1379 Frequency : occasionally with multiple writers to a single file
1381 Description: shared writes to file may result in wrong size reported by stat()
1382 Details : Allow growing of kms when extent lock is cancelled
1385 Frequency : always with random mmap IO to multi-striped file
1387 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1388 Details : Only if the hole is at the end of OST object so that kms is too
1389 small. Fix is to increase kms accordingly in ll_nopage.
1392 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1394 Description: writes to a missing object would leak memory on the OST
1395 Details : If there is an inconsistency between the MDS and OST filesystems,
1396 such that the MDS references an object that doesn't exist, writes
1397 to that object will leak memory due to incorrect cleanup in the
1398 error handling path, eventually running out of memory on the OST.
1403 Description: Creating too long symlink causes lustre errors
1404 Details : Check symlink and name lengths before sending requests to MDS.
1407 Frequency : only if flock is enabled (not on by default)
1409 Description: posix locks not released on fd closure on 2.6.9+
1410 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1411 kernels, this caused such locks not to be released on fd close and
1412 then assertions on fs unmount about still used locks.
1415 Frequency : MDS failover only, very rarely
1417 Description: clients may get ASSERTION(granted_lock != NULL)
1418 Details : When request was taking a long time, and a client was resending
1419 a getattr by name lock request. The were multiple lock requests
1420 with the same client lock handle and
1421 mds_getattr_name->fixup_handle_for_resent_request found one of the
1422 lock handles but later failed with ASSERTION(granted_lock != NULL).
1427 Description: handle->h_buffer_credits > 0, assertion failure
1428 Details : h_buffer_credits is zero after truncate, causing assertion
1429 failure. This patch extends the transaction or creates a new
1433 Frequency : NFS re-export or patchless client
1434 Bugzilla : 11179, 10796
1435 Description: Crash on NFS re-export node (__d_move)
1436 Details : We do not want to hash the dentry if we don't have a lock.
1437 But if this dentry is later used in d_move, we'd hit uninitialised
1438 list head d_hash, so we just do this to init d_hash field but
1439 leave dentry unhashed.
1442 Frequency : NFS re-export or patchless client
1444 Description: NFS exports has problem with symbolic link
1445 Details : lustre client didn't properly install dentry when re-exported
1446 to NFS or running patchless client.
1449 Frequency : NFS re-export or patchless client
1451 Description: Various nfs/patchless fixes.
1452 Details : fixes reuse disconected alias for lookup process - this fixes
1453 warning "find_exported_dentry: npd != pd",
1454 fix permission error with open files at nfs.
1455 fix apply umask when do revalidate.
1458 Frequency : occasional
1460 Description: Crash on NFS re-export node
1461 Details : calling clear_page() on the wrong pointer triggered oops in
1462 generic_mapping_read().
1465 Frequency : rarely, using O_DIRECT IO
1467 Description: unaligned directio crashes client with LASSERT
1468 Details : check for unaligned buffers before trying any requests.
1471 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1473 Description: stale data returned from RAID cache
1474 Details : If only a small amount of IO is done to the RAID device before
1475 reading it again it is possible to get stale data from the RAID
1476 cache instead of reading it from disk.
1479 Frequency : always for sles10 kernel
1481 Description: sles10 support
1482 Details : ll_follow_link: compile fixes and using of nd_set_link
1483 under newer kernels.
1486 Frequency : depends on arch, kernel and compiler version, always on sles10
1489 Description: recursive or deep enough symlinks cause stack overflow
1490 Details : getting rid of large stack-allocated variable in
1494 Frequency : depends on hardware
1496 Description: lustre write performance loss in the SLES10 kernel
1497 Details : the performance loss is caused by using of write barriers in the
1498 ext3 code. The SLES10 kernel turns barrier support on by
1499 default. The fix is to undo that change for ldiskfs.
1501 ------------------------------------------------------------------------------
1503 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1505 * Support for kernels:
1506 2.6.9-42.0.3EL (RHEL 4)
1507 2.6.5-7.276 (SLES 9)
1508 2.4.21-47.0.1.EL (RHEL 3)
1509 2.6.12.6 vanilla (kernel.org)
1513 Frequency : quota enabled and large files being deleted
1515 Description: releasing more than 4GB of quota at once hangs OST
1516 Details : If a user deletes more than 4GB of files on a single OST it
1517 will cause the OST to spin in an infinite loop. Release
1518 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1523 Description: statfs data retrieved from /proc may be stale or zero
1524 Details : When reading per-device statfs data from /proc, in the
1525 {kbytes,files}_{total,free,avail} files, it may appear
1526 as zero or be out of date.
1529 Frequency : systems with MD RAID1 external journal devices
1531 Description: lconf's call to blkid is confused by RAID1 journal devices
1532 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1533 of returning all block devices that match the journal UUID.
1536 Frequency : always, for aggregate stripe size over 4GB
1538 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1539 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1540 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1543 Frequency : always if "lfs find" used on a local file/directory
1545 Description: "lfs find" segfaults if used on a local file/directory
1546 Details : The case where a directory component was not specified wasn't
1547 handled correctly. Handle this properly.
1550 Frequency : always on ppc64
1552 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1553 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1554 perform endianness conversion.
1557 Frequency : rarely (truncate to non-zero file size after write under load)
1558 Bugzilla : 10730, 10687
1559 Description: Files padded with zeros to next 4K multiple
1560 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1561 it is possible that files that are truncated to a non-zero size
1562 immediately after being written are filled with zero bytes beyond
1563 the truncated size. No file data is lost.
1565 Severity : enhancement
1567 Description: Allow recovery/failover for liblustre clients.
1568 Details : liblustre clients were unaware of failover configurations until
1571 Severity : enhancement
1573 Description: user file locks should fail when not mounting with flock option
1574 Details : Set up an error-returning stub in ll_file_operations.lock field
1575 to prevent incorrect behaviour when client is mounted without
1576 flock option. Also, set up properly f_op->flock field for
1580 Frequency : always on ia64
1582 Description: "lfs df" loops on printing out MDS statfs information
1583 Details : The obd_ioctl_data was not initialized and in some systems
1584 this caused a failure during the ioctl that did not return
1585 an error. Initialize the struct and return an error on failure.
1588 Frequency : SLES 9 only
1590 Description: Error of copying files with lustre special EAs as root
1591 Details : Client side always return success for setxattr call for lustre
1592 special xattr (currently only "trusted.lov").
1595 Frequency : rarely on clusters with both ia64+i386 clients
1597 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1598 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1599 clients are doing concurrent writes to the same file it is
1600 possible that the ia64 clients may LASSERT because the OST
1601 extent locks are not PAGE_SIZE aligned. Ensure that grown
1602 locks are always aligned on the request boundary.
1605 Frequency : specific use, occasional
1607 Description: Overwriting in use executable truncates on-disk binary image
1608 Details : If one node attempts to overwrite an executable in use by
1609 another node, we now correctly return ETXTBSY instead of
1610 truncating the file.
1612 Severity : enhancement
1614 Description: Async OSC create to avoid the blocking unnecessarily.
1615 Details : If a OST has no remain object, system will block on the creating
1616 when need to create a new object on this OST. Now, ways use
1617 pre-created objects when available, instead of blocking on an
1618 empty osc while others are not empty. If we must block, we block
1619 for the shortest possible period of time.
1624 Description: chmod on Lustre root is propagated to other clients
1625 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1626 invalid by the follow_mount time.
1629 Frequency : liblustre clients only
1631 Description: Race in 'instant cancel' lock handling could lead to such locks
1632 never to be granted in case of SMP MDS
1633 Details : Do not destroy not yet granted but cbpending locks in
1637 Frequency : replay/resend of open
1639 Description: non null lock assetion failure in mds_intent_policy
1640 Details : Trying to replay/resend lockless open requests resulted in
1641 mds_open() returning 0 with no lock. Now it sets a flag if
1642 a lock is going to be returned.
1644 Severity : enhancement
1646 Description: Checksum enhancements
1647 Details : New checksum enhancements allow for resending RPCs that failed
1650 Severity : enhancement
1652 Description: Tunables on number of dirty pages in cacche
1653 Details : Allow to set limit on number of dirty pages cached.
1658 Description: client crash on unmount - lock still has references
1659 Details : In some error handling cases it was possible to leak a lock
1660 reference on a client while accessing a file. This was not
1661 harmful to the client during operation, but would cause the
1662 client to crash when the filesystem is unmounted.
1665 Frequency : specific case, rare
1667 Description: ETXTBSY on mds though file not in use by client
1668 Details : ETXTBSY is no longer incorrectly returned when attempting to
1669 chmod or chown a directory that the user previously tried to
1670 execute or a currently-executing binary.
1673 Frequency : extremely rare except on liblustre-based clients
1675 Description: Lustre space not freed when files are deleted
1676 Details : Clean up open-unlinked files after client eviction. Previously
1677 the unlink was skipped and the files remained as orphans.
1682 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1683 Details : In some cases it was possible to send a blocking callback to
1684 a client doing a glimpse, even though that client didn't get
1685 a lock granted. When the glimpse lock is cancelled on the OST
1686 the freed lock is left on the waiting list and corrupted the list.
1689 Frequency : all core dumps
1691 Description: Broke core dumps to lustre
1692 Details : Negative dentry may be unhashed if parent does not have UPDATE
1693 lock, but some callers, e.g. do_coredump, expect dentry to be
1694 hashed after successful create, hash it in ll_create_it.
1696 ------------------------------------------------------------------------------
1698 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1700 * Support for kernels:
1701 2.6.9-42.0.2.EL (RHEL 4)
1702 2.6.5-7.276 (SLES 9)
1703 2.4.21-40.EL (RHEL 3)
1704 2.6.12.6 vanilla (kernel.org)
1708 Frequency : always on RHEL 3
1710 Description: Number of open files grows over time
1711 Details : The number of open files grows over time, whether or not
1712 Lustre is started. This was due to a filp leak introduced
1713 by one of our kernel patches.
1715 ------------------------------------------------------------------------------
1717 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
1719 * Support for kernels:
1720 2.6.9-42.EL (RHEL 4)
1721 2.6.5-7.276 (SLES 9)
1722 2.4.21-40.EL (RHEL 3)
1723 2.6.12.6 vanilla (kernel.org)
1728 Bugzilla : 5719, 9635, 9792, 9684
1729 Description: OST (or MDS) trips assertions in (re)connection under heavy load
1730 Details : If a server is under heavy load and cannot reply to new
1731 connection requests before the client resends the (re)connect,
1732 the connection handling code can behave badly if two service
1733 threads are concurrently handing separate (re)connections from
1734 the same client. Add better locking to the connection handling
1735 code, and ensure that only a single connection will be processed
1736 for a given client UUID, even if the lock is dropped.
1738 Severity : enhancement
1740 Description: add TCP zero-copy support to kernel
1741 Details : Add support to the kernel TCP stack to allow zero-copy bulk
1742 sends if the hardware supports scatter-gather and checksumming.
1743 This allows socklnd to do client-write and server-read more
1744 efficiently and reduce CPU utilization from skbuf copying.
1747 Frequency : only if NFS exporting from client
1749 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
1750 Details : If setting attributes on a file created under NFS that had
1751 never been opened it would be possible to oops the client
1752 if the file had no objects.
1755 Frequency : always for liblustre
1757 Description: liblustre client does MDS+OSTs setattr RPC for each write
1758 Details : When doing a write from a liblustre client, the client
1759 incorrectly issued an RPC to the MDS and each OST the file was
1760 striped over in order to update the timestamps. When writing
1761 with small chunks and many clients this could overwhelm the MDS
1762 with RPCs. In all cases it would slow down the write because
1763 these RPCs are unnecessary.
1765 Severity : enhancement
1767 Description: allow number of MDS service threads to be changed at module load
1768 Details : It is now possible to change the number of MDS service threads
1769 running. Adding "options mds mds_num_threads={N}" to the MDS's
1770 /etc/modprobe.conf will set the number of threads for the next
1771 time Lustre is restarted (assuming the "mds" module is also
1772 reloaded at that time). The default number of threads will
1773 stay the same, 32 for most systems.
1778 Description: OST crash if filesystem is unformatted or corrupt
1779 Details : If an OST is started on a device that has never been formatted
1780 or if the filesystem is corrupt and cannot even mount then the
1781 error handling cleanup routines would dereference a NULL pointer.
1786 Description: NULL pointer deref in llap_from_page.
1787 Details : get_cache_page_nowait can return a page with NULL (or otherwise
1788 incorrect) mapping if the page was truncated/reclaimed while it was
1789 searched for. Check for this condition and skip such pages when
1790 doing readahead. Introduce extra check to llap_from_page() to
1791 verify page->mapping->host is non-NULL (so page is not anonymous).
1794 Frequency : Sometimes when using sys_sendfile
1796 Description: "page not covered by a lock" warnings from ll_readpage
1797 Details : sendfile called ll_readpage without right page locks present.
1798 Now we introduced ll_file_sendfile that does necessary locking
1799 around call to generic_file_sendfile() much like we do in
1803 Frequency : with certain MDS communication failures at client mount time
1805 Description: NULL pointer deref after failed client mount
1806 Details : a client connection request may delayed by the network layer
1807 and not be sent until after the PTLRPC layer has timed out the
1808 request. If the client fails the mount immediately it will try
1809 to clean up before the network times out the request. Add a
1810 reference from the request import to the obd device and delay
1811 the cleanup until the network drops the request.
1814 Frequency : occasionally during client (re)connect
1816 Description: assertion failure during client (re)connect
1817 Details : processing a client connection request may be delayed by the
1818 client or server longer than the client connect timeout. This
1819 causes the client to resend the connection request. If the
1820 original connection request is replied in this interval, the
1821 client may trip an assertion failure in ptlrpc_connect_interpret()
1822 which thought it would be the only running connect process.
1825 Frequency : only with obd_echo servers and clients that are rebooted
1827 Description: kernel BUG accessing uninitialized data structure
1828 Details : When running an obd_echo server it did not start the ping_evictor
1829 thread, and when a client was evicted an uninitialized data
1830 structure was accessed. Start the ping_evictor in the RPC
1831 service startup instead of the OBD startup.
1833 Severity : enhancement
1834 Bugzilla : 10193 (patchless)
1835 Description: Remove dependency on various unexported kernel interfaces.
1836 Details : No longer need reparent_to_init, exit_mm, exit_files,
1837 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
1840 Frequency : rare (only users of deprecated and unsupported LDAP config)
1842 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
1843 Details : LDAP apparently contains 'lustreName' attributes instead of
1844 'name'. A simple remapping of the name is sufficient.
1847 Frequency : rare (only with non-default dump_on_timeout debug enabled)
1849 Description: waiting_locks_callback trips kernel BUG if client is evicted
1850 Details : Running with the dump_on_timeout debug flag turned on makes
1851 it possible that the waiting_locks_callback() can try to dump
1852 the Lustre kernel debug logs from an interrupt handler. Defer
1853 this log dumping to the expired_lock_main() thread.
1855 Severity : enhancement
1857 Description: Support NFS exporting on 2.6 kernels.
1858 Details : Implement non-rawops metadata methods for NFS server to use without
1859 changing NFS server code.
1862 Frequency : very rare (synthetic metadata workload only)
1864 Description: two racing renames might cause an MDS thread to deadlock
1865 Details : Running the "racer" program may cause one MDS thread to rename
1866 a file from being the source of a rename to being the target of
1867 a rename at exactly the same time that another thread is doing
1868 so, and the second thread has already enqueued these locks after
1869 doing a lookup of the target and is trying to relock them in
1870 order. Ensure that we don't try to re-lock the same resource.
1873 Frequency : only very large systems with liblustre clients
1875 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
1876 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
1877 all OSTs in parallel. This allows the eviction of stale liblustre
1878 clients to proceed much faster than if they were done in series,
1879 and also offers similar improvements for other set_info RPCs.
1884 Description: excessive CPU usage during initial read phase on client
1885 Details : During the initial read phase on a client, it would agressively
1886 retry readahead on the file, consuming too much CPU and impacting
1887 performance (since 1.4.5.8). Improve the readahead algorithm
1888 to avoid this, and also improve some other common cases (read
1889 of small files in particular, where "small" is files smaller than
1890 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
1895 Description: MDS crash when receiving packet with unknown intent.
1896 Details : Do not LBUG in unknown intent case, just return -EFAULT
1898 Severity : enhancement
1899 Bugzilla : 9293, 9385
1900 Description: MDS RPCs are serialised on client. This is unnecessary for some.
1901 Details : Do not serialize getattr (non-intent version) and statfs.
1904 Frequency : occasional, when OST network is overloaded/intermittent
1906 Description: client evicted by OST after bulk IO timeout
1907 Details : If a client sends a bulk IO request (read or write) the OST
1908 may evict the client if it is unresposive to its data GET/PUT
1909 request. This is incorrect if the network is overloaded (takes
1910 too long to transfer the RPC data) or dropped the OST GET/PUT
1911 request. There is no need to evict the client at all, since
1912 the pinger and/or lock callbacks will handle this, and the
1913 client can restart the bulk request.
1916 Frequency : Always when mmapping file with no objects
1918 Description: client crashes when mmapping file with no objects
1919 Details : Check that we actually have objects in a file before doing any
1920 operations on objects in ll_vm_open, ll_vm_close and
1926 Description: Request leak when working with deleted CWD
1927 Details : Introduce advanced request refcount tracking for requests
1928 referenced from lustre intent.
1930 Severity : Enhancement
1932 Description: Cache open file handles on client.
1933 Details : MDS now will return special lock along with openhandle, if
1934 requested and client is allowed to hold openhandle, even if unused,
1935 until such a lock is revoked. Helps NFS a lot, since NFS is opening
1936 closing files for every read/write openration.
1938 Severity : Enhancement
1940 Description: Cache open negative dentries on client when possible.
1941 Details : Guard negative dentries with UPDATE lock on parent dir, drop
1942 negative dentries on lock revocation.
1947 Description: Remounting a client read-only wasn't possible with a zconf mount
1948 Details : It wasn't possible to remount a client read-only with llmount.
1950 Severity : enhancement
1951 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
1952 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
1953 MPICH is included to add Lustre-specific ADIO interfaces.
1954 This is based closely on the UFS ADIO layer and only differs
1955 in file creation, in order to allow the OST striping to be set.
1956 This is user-contributed code and not supported by CFS.
1961 Description: extended inode attributes (immutable, append-only) work improperly
1962 when 2.4 and 2.6 kernels are used on client/server or vice versa
1963 Details : Introduce kernel-independent values for these flags.
1965 Severity : enhancement
1968 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
1969 Details : Many of the /proc/ tunables can only be tuned at a megabyte
1970 granularity. Now, Fractional MB granularity is be supported,
1971 this is very useful for low memory system.
1973 Severity : enhancement
1975 Description: Getattr by fid
1976 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
1977 locks, avoids extra getattr rpc requests to MDS, allows '/' to
1978 have locks and avoids getattr rpc requests for it on every stat.
1981 Frequency : Always, for filesystems larger than 2TB
1983 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1984 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
1985 filesystems larger than 2TB to be mounted with mballoc enabled.
1988 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
1990 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
1991 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
1992 filesystems larger than 2TB is mounted then the kernel will
1993 silently corrupt the start of the filesystem. CONFIG_LBD is
1994 enabled for all CFS-supported kernels, but the possibility of
1995 this happening with a modified kernel config exists.
1997 Severity : enhancement
1999 Description: add client O_DIRECT support for 2.6 kernels
2000 Details : It is now possible to do O_DIRECT reads and writes to files
2001 in the Lustre client mountpoint on 2.6 kernel clients.
2003 Severity : enhancement
2005 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2006 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2007 parallel, not waiting for response from every OST before sending
2008 a rpc to the next OST.
2013 Description: setattr vs write race when updating file timestamps
2014 Details : Client processes that update a file timestamp into the past
2015 right after writing to the file (e.g. tar) it is possible that
2016 the updated file modification time can be reset to the current
2017 time due to a race between processing the setattr and write RPC.
2019 Severity : enhancement
2021 Description: Bring 'lfs find' closer in line with regular Linux find.
2022 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2023 -print0 options and obtains all the needed info through the lustre
2026 Severity : enhancement
2028 Description: support up to 1024 configured devices on one node
2029 Details : change obd_dev array from statically allocated to dynamically
2030 allocated structs as they are first used to reduce memory usage
2035 Description: Flush dirty partially truncated pages during truncate
2036 Details : Immediatelly flush partially truncated pages in filter_setattr,
2037 this way we completely avoid having any pages in page cache on OST
2038 and can retire ugly workarounds during writes to flush such pages.
2043 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2044 Details : protect inode from truncation within vfs_unlink() context
2045 just take a reference before calling vfs_unlink() and release it
2046 when parent's i_sem is free.
2051 Description: last_id value checked outside lock on OST caused LASSERT failure
2052 Details : If there were multiple MDS->OST object precreate requests in
2053 flight, it was possible that the OST's last object id was checked
2054 outside a lock and incorrectly tripped an assertion. Move checks
2055 inside locks, and discard old precreate requests.
2058 Frequency : always, if extents are used on OSTs
2060 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2061 Details : OSTs using the extents format would not zero the high 16 bits of
2062 the index physical block number. This is not a problem for any
2063 OST filesystems smaller than 16TB, and no kernels support ext3
2064 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2065 new/modified files) and can be fixed for existing filesystems
2066 with e2fsprogs-1.39-cfs1.
2071 Description: import connection selection may be incorrect if timer wraps
2072 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2073 import connections to be ignored if the 32-bit jiffies counter
2074 wraps. Use a 64-bit jiffies counter.
2077 Frequency : very large clusters immediately after boot
2079 Description: LNET request buffers exhausted under heavy short-term load
2080 Details : If a large number of client requests are generated on a service
2081 that has previously never seen so many requests it is possible
2082 that the request buffer growth cannot keep up with the spike in
2083 demand. Instead of dropping incoming requests, they are held in
2084 the LND until the RPC service can accept more requests.
2087 Frequency : Sometimes during replay
2089 Description: Assertion failure in ll_local_open after replay.
2090 Details : If replay happened on an open request reply before we were able
2091 to set replay handler, reply will become not swabbed tripping the
2092 assertion in ll_local_open. Now we set the handler right after
2093 recognising of open request
2096 Frequency : very rare
2098 Description: kernel reports "badness in vsnprintf"
2099 Details : Reading from the "recovery_status" /proc file in small chunks
2100 may cause a negative length in lprocfs_obd_rd_recovery_status()
2101 call to vsnprintf() (which is otherwise harmless). Exit early
2102 if there is no more space in the output buffer.
2104 Severity : enhancement
2106 Description: clear OBD RPC statistics by writing to them
2107 Details : It is now possible to clear the OBD RPC statistics by writing
2108 to the "stats" file.
2113 Description: Client mtime is not the same on different clients after utimes
2114 Details : In some cases, the client was using the utimes() syscall on
2115 a file cached on another node. The clients now validate the
2116 ctime from the MDS + OSTs to determine which one is right.
2121 Description: Inability to activate failout mode
2122 Details : lconf script incorrectly assumed that in pythong string's numeric
2123 value is used in comparisons.
2126 Frequency : always with multiple stripes per file
2128 Description: Inefficient object allocation for mutli-stripe files
2129 Details : When selecting which OSTs to stripe files over, for files with
2130 a stripe count that divides evenly into the number of OSTs,
2131 the MDS is always picking the same starting OST for each file.
2132 Return the OST selection heuristic to the original design.
2137 Description: mount failures may take full timeout to return an error
2138 Details : Under some heavy load conditions it is possible that a
2139 failed mount can wait for the full obd_timeout interval,
2140 possibly several minutes, before reporting an error.
2141 Instead return an error as soon as the status is known.
2143 Frequency : quota enabled and large files being deleted
2145 Description: releasing more than 4GB of quota at once hangs OST
2146 Details : If a user deletes more than 4GB of files on a single OST it
2147 will cause the OST to spin in an infinite loop. Release
2148 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2153 Description: statfs data retrieved from /proc may be stale or zero
2154 Details : When reading per-device statfs data from /proc, in the
2155 {kbytes,files}_{total,free,avail} files, it may appear
2156 as zero or be out of date.
2159 Frequency : systems with MD RAID1 external journal devices
2161 Description: lconf's call to blkid is confused by RAID1 journal devices
2162 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2163 of returning all block devices that match the journal UUID.
2166 Frequency : always, for aggregate stripe size over 4GB
2168 Description: assertion fails when trying to use 4GB stripe size
2169 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2170 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2173 Frequency : always on ppc64
2175 Description: the first write on an ext3 filesystem with mballoc got stuck
2176 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2177 perform endianness conversion.
2179 ------------------------------------------------------------------------------
2181 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2183 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2184 INTEROPERATE with older versions automatically. Please read the
2185 user documentation before upgrading any part of a live system.
2186 * WARNING: Lustre networking configuration changes are required with
2187 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2190 * Support for kernels:
2191 2.6.9-22.0.2.EL (RHEL 4)
2192 2.6.5-7.244 (SLES 9)
2193 2.6.12.6 vanilla (kernel.org)
2196 Severity : enhancement
2197 Bugzilla : 7981/8208
2198 Description: Introduced Lustre Networking (LNET)
2199 Details : LNET is new networking infrastructure for Lustre, it includes
2200 a reorganized network configuration mode (see the user
2201 documentation for full details) as well as support for routing
2202 between different network fabrics. Lustre Networking Devices
2203 (LNDs) for the supported network fabrics have also been
2204 created for this new infrastructure.
2206 Severity : enhancement
2207 Description: Introduced Access control lists
2208 Details : clients can set ACLs on files and directories in order to have
2209 more fine-grained permissions than the standard Unix UGO+RWX.
2210 The MDS must be started with the "-o acl" mount option.
2212 Severity : enhancement
2213 Description: Introduced filesystem quotas
2214 Details : Administrators may now establish per-user quotas on the
2217 Severity : enhancement
2219 Description: Configuration change for the XT3
2220 The PTLLND is now used to run Lustre over Portals on the XT3
2221 The configure option(s) --with-cray-portals are no longer used.
2222 Rather --with-portals=<path-to-portals-includes> is used to
2223 enable building on the XT3. In addition to enable XT3 specific
2224 features the option --enable-cray-xt3 must be used.
2229 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2230 Details : A race between allocating a new llcd and re-getting the llcd_lock
2231 allowed another thread to grab newly-allocated llcd.
2233 Severity : enhancement
2235 Description: 2.6 OST async journal commit and locking fix to improve performance
2236 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2237 2.6 kernels are now async as they already were in 2.4 kernels so
2238 that they can commit concurrently with the network bulk transfer.
2239 For block-allocated files the filter allocation semaphore is held
2240 to avoid filesystem fragmentation during allocation. BKL lock
2241 removed for 2.6 xattr operations where it is no longer needed.
2246 Description: lconf incorrectly determined whether two IP networks could talk
2247 Details : In some more complicated routing and multiple-network
2248 configurations, lconf will avoid trying to make a network
2249 connection to a disjoint part of the IP space. It was doing the
2250 math incorrectly for one set of cases.
2255 Description: Fix for potential infinite loop processing records in an llog.
2256 Details : If an llog record is corrupted/zeroed, it is possible to loop
2257 forever in llog_process(). Validate the llog record length
2258 and skip the remainder of the block on error.
2261 Frequency : occasional (liblustre only)
2263 Description: liblustre could not open files whose last component is a symlink
2264 Details : sysio_path_walk() would incorrectly pass the open intent to
2265 intermediate path components.
2268 Frequency : rare (liblustre only with non-standard tuning)
2269 Bugzilla : 7201 (7350)
2270 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2271 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2272 after it has been released, only do this for referenced locks
2274 Severity : enhancement
2276 Description: specifying an (invalid) directory default stripe_size of -1
2277 would reset the directory default striping
2278 Details : stripe_size -1 was used internally to signal directory stripe
2279 removal, now use "all default" to signal dir stripe removal
2280 as a directory striping of "all default" is not useful
2283 Frequency : common for large clusters running liblustre clients
2285 Description: doing an ls when liblustre clients are running is slow
2286 Details : sending a glimpse AST to a liblustre client waits for every AST
2287 to time out, as liblustre clients will not respond. Since they
2288 cannot cache data we refresh the OST lock LVB from disk instead.
2290 Severity : enhancement
2292 Description: doing an ls at the same time as file IO can be slow
2293 Details : enqueue and other "small" requests can be blocked behind many
2294 large IO requests. Create a new OST IO portal for non-IO
2295 requests so they can be processed faster.
2298 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2300 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2301 created file with mode 000
2302 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2303 allow an ftruncate() as a non-root user to a file with mode 000.
2304 Lustre now respects this flag to disable mode checks when
2305 truncating a file owned by the user
2308 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2311 Description: Revoking locks from clients that went dead or catatonic might take
2313 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2314 cancellation of such locks instant on servers without waiting for
2315 any reply from clients. Clients drops these locks when cancel
2316 notification from server is received without replying.
2319 Frequency : liblustre-only, when liblustre client dies or becomes busy
2321 Description: Doing ls on Linux clients can take a long time with active
2323 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2324 granting such locks to it in the first place if possible. Locks
2325 are taken by proxy on the OST during the read or write and
2326 dropped immediately afterward. Add connect flags handling, do
2327 not grant locks to liblustre clients for glimpse ASTs.
2329 Severity : enhancement
2331 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2332 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2333 sequential reads and behaves badly with applications doing random
2334 reads. Improve it by reading ahead at least read region, and
2335 avoiding excessive large RPC for small reads.
2337 Severity : enhancement
2339 Description: Creating more than 1000 files for a single job may cause a load
2340 imbalance on the OSTs if there are also a large number of OSTs.
2341 Details : qos_prep_create() uses an OST index reseed value that is an
2342 even multiple of the number of available OSTs so that if the
2343 reseed happens in the middle of the object allocation it will
2344 still utilize the OSTs as uniformly as possible.
2349 Description: OST or MDS may oops in ping_evictor_main()
2350 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2351 but doesn't restart at beginning of obd_exports_timed list
2354 Severity : enhancement
2356 Description: improve by-nid export eviction on the MDS and OST
2357 Details : allow multiple exports with the same NID to be evicted at one
2358 time without re-searching the exports list.
2361 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2363 Description: MDS may oops in groups_free()
2364 Details : in rare race conditions a newly allocated group_info struct is
2365 freed again, and this can be NULL. The 2.4 compatibility code
2366 for groups_free() checked for a NULL pointer, but 2.6 did not.
2369 Frequency : common for liblustre clients doing little filesystem IO
2370 Bugzilla : 9352, 7313
2371 Description: server may evict liblustre clients accessing contended locks
2372 Details : if a client is granted a lock or receives a completion AST
2373 with a blocking AST already set it would not reply to the AST
2374 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2377 Frequency : lfs setstripe, only systems with more than 160 OSTs
2379 Description: unable to set striping with a starting offset beyond OST 160
2380 Details : llapi_create_file() incorrectly limited the starting stripe
2381 index to the maximum single-file stripe count.
2384 Frequency : LDAP users only
2386 Description: lconf did not handle in-kernel recovery with LDAP properly
2387 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2389 Severity : enhancement
2391 Description: bind OST threads to NUMA nodes to improve performance
2392 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2393 node and do their allocations there to localize memory access
2395 Severity : enhancement
2397 Description: llmount can determine client NID directly from Myrinet (GM)
2398 Details : the client NID code from gmnalnid was moved directly into
2399 llmount, removing the need to use this or specifying the
2400 client NID explicitly when mounting GM clients with zeroconf
2403 Frequency : if client is started with down MDS
2405 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2406 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2407 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2408 ptlrpc_import_delayed_req() because we didn't check if the
2409 request was interrupted, and we also didn't break out of the
2410 event loop if there was no timeout
2415 Description: data loss during non-page-aligned writes to a single file from
2416 both multiple nodes and multiple threads on one node at same time
2417 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2418 inconsistency led to false short-reads, that were cached and later
2419 used by ->prepare_write() to fill in partially written page,
2420 leading to data loss.
2423 Frequency : always, if lconf --abort_recovery used
2425 Description: lconf --abort_recovery fails with 'Operation not supported'
2426 Details : lconf was attempting to abort recovery on the MDT device and not
2429 Severity : enhancement
2431 Description: remove cleanup logs
2432 Details : replace lconf-generated cleanup logs with lustre internal
2433 cleanup routines. Eliminates the need for client-cleanup and
2436 Severity : enhancement
2438 Description: add support for EAs (user and system) on lustre filesystems
2439 Details : it is now possible to store extended attributes in the Lustre
2440 client filesystem, and with the user_xattr mount option it
2441 is possible to allow users to store EAs on their files also
2443 Severity : enhancement
2445 Description: Add possibility (config option) to show minimal available OST free
2447 Details : When compiled with --enable-mindf configure option, statfs(2)
2448 (and so, df) will return least minimal free space available from
2449 all OSTs as amount of free space on FS, instead of summary of
2450 free spaces of all OSTs.
2452 Severity : enhancement
2454 Description: do not expand extent locks acquired on OST-side
2455 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2456 by server: they are not cached anyway.
2459 Frequency : when mmap is used/binaries executed from Lustre
2461 Description: Unmmap pages before throwing them away from read cache.
2462 Details : llap_shrink cache now attempts to unmap pages before discarding
2463 them (if unmapping failed - do not discard). SLES9 kernel has
2464 extra checks that trigger if this unmapping is not done first.
2469 Description: lconf didn't resolve symlinks before checking to see whether a
2470 given mountpoint was already in use
2473 Frequency : when migrating failover services
2474 Bugzilla : 6395, 9514
2475 Description: When migrating a subset of services from a node (e.g. failback
2476 from a failover service node) the remaining services would
2477 time out and evict clients.
2478 Details : lconf --force (implied by --failover) sets the global obd_timeout
2479 to 5 seconds in order to quickly disconnect, but this caused
2480 other RPCs to time out too quickly. Do not change the global
2481 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2483 Severity : enhancement
2484 Frequency : if MDS is started with down OST
2485 Bugzilla : 9439,5706
2486 Description: Allow startup/shutdown of an MDS without depending on the
2487 availability of the OSTs.
2488 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2489 Add appropriate locking and lov-osc refcounts for safe
2490 cleaning. Add osc abort_inflight calls in case the
2491 synchronize never started.
2494 Frequency : occasional (Cray XT3 only)
2496 Description: root not authorized to access files in CRAY_PORTALS environment
2497 Details : The client process capabilities were not honoured on the MDS in
2498 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2499 been accessed by an authorized user then root was able to access
2500 the file on the local client also. The root user capabilities
2501 are now allowed on the MDS, as this environment has secure UID.
2504 Frequency : occasional
2506 Description: ldiskfs "too long searching" message happens too often
2507 Details : A debugging message (otherwise harmless) prints too often on
2508 the OST console. This has been reduced to only happen when
2509 there are fragmentation problems on the filesystem.
2514 Description: Division by zero in statfs when all OSCs are inactive
2515 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2516 lov_statfs divides by value returned by lov_get_stripecnt().
2520 Bugzilla : 9489, 3273
2521 Description: First write from each client to each OST was only 4kB in size,
2522 to initialize client writeback cache, which caused sub-optimal
2523 RPCs and poor layout on disk for the first writen file.
2524 Details : Clients now request an initial cache grant at (re)connect time
2525 and so that they can start streaming writes to the cache right
2526 away and always do full-sized RPCs if there is enough data.
2527 If the OST is rebooted the client also re-establishes its grant
2528 so that client cached writes will be honoured under the grant.
2533 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2534 Details : Now I/O RPCs go to different portal number and (presumably) fast
2535 lock requests (and glimses) and other RPCs get their own service
2536 threads pool that should be able to service those RPCs
2539 Severity : enhancement
2541 Description: Ability to exchange lustre version between client and servers and
2542 issue warnings at client side if client is too old. Also for
2543 liblustre clients there is ability to refuse connection of too old
2545 Details : New 'version' field is added to connect data structure that is
2546 filled with version info. That info is later checked by server and
2550 Frequency : rare, liblustre only.
2551 Bugzilla : 9296, 9581
2552 Description: Two simultaneous writes from liblustre at offset within same page
2553 might proceed at the same time overwriting eachother with stale
2555 Details : I/O lock withing llu_file_prwv was released too early, before data
2556 actually was hitting the wire. Extended lock-holding time until
2557 server acknowledges receiving data.
2560 Frequency : extremely rare. Never observed in practice.
2562 Description: avoid generating lustre_handle cookie of 0.
2563 Details : class_handle_hash() generates handle cookies by incrementing
2564 global counter, and can hit 0 occasionaly (this is unlikely, but
2565 not impossible, because initial value of cookie counter is
2566 selected randonly). Value of 0 is used as a sentinel meaning
2567 "unassigned handle" --- avoid it. Also coalesce two critical
2568 sections in this function into one.
2570 Severity : enhancement
2572 Description: allow liblustre clients to delegate truncate locking to OST
2573 Details : To avoid overhead of locking, liblustre client instructs OST to
2574 take extent lock in ost_punch() on client's behalf. New connection
2575 flag is added to handle backward compatibility.
2577 Severity : enhancement
2578 Bugzilla : 4928, 7341, 9758
2579 Description: allow number of OST service threads to be specified
2580 Details : a module parameter allows the number of OST service threads
2581 to be specified via "options ost ost_num_threads={N}" in the
2582 OSS's /etc/modules.conf or /etc/modprobe.conf.
2586 Bugzilla : 6146, 9635, 9895
2587 Description: servers crash with bad pointer in target_handle_connect()
2588 Details : In rare cases when a client is reconnecting it was possible that
2589 the connection request was the last reference for that export.
2590 We would temporarily drop the export reference and get a new
2591 one, but this may have been the last reference and the export
2592 was just destroyed. Get new reference before dropping old one.
2594 Severity : enhancement
2595 Frequency : if client is started with failover MDS
2597 Description: Allow multiple MDS hostnames in the mount command
2598 Details : Try to read the configuration from all specified MDS
2599 hostnames during a client mount in case the "primary"
2602 Severity : enhancement
2604 Description: Stop sending data to evicted clients as soon as possible.
2605 Details : Check if the client we are about to send or are sending data to
2606 was evicted already. (Check is done every second of waiting,
2607 for which l_wait_event interface was extended to allow checking
2608 of exit condition at specified intervals).
2611 Frequency : rare, normally only when NFS exporting is done from client
2613 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2615 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2618 Frequency : always, if srand() is called before liblustre initialization
2620 Description: Liblustre uses system PRNG disturbing its usage by user application
2621 Details : Introduce internal to lustre fast and high-quality PRNG for
2622 lustre usage and make liblustre and some other places in generic
2623 lustre code to use it.
2625 Severity : enhancement
2626 Bugzilla : 9477, 9557, 9870
2627 Description: Verify that the MDS configuration logs are updated when xml is
2628 Details : Check if the .xml configuration logs are newer than the config
2629 logs stored on the MDS and report an error if this is the case.
2630 Request --write-conf, or allow starting with --old_conf.
2632 Severity : enhancement
2634 Description: Handle symlinks in the path when checking if Lustre is mounted.
2635 Details : Resolve intermediate symlinks when checking if a client has
2636 mounted a filesystem to avoid duplicate client mounts.
2641 Description: lconf can hit an error exception but still return success.
2642 Details : The lconf command catches the Command error exception at the top
2643 level script context and will exit with the associated exit
2644 status, but doesn't ensure that this exit status is non-zero.
2649 Description: failure of ptlrpc thread startup can cause oops
2650 Details : Starting a ptlrpc service thread can fail if there are a large
2651 number of threads or the server memory is very fragmented.
2652 Handle this without oopsing.
2655 Frequency : always, only if liblustre and non-default acceptor port was used
2657 Description: liblustre cannot connect to servers with non-default acceptor port
2658 Details : tcpnal_set_default_params() was not called and was therefore
2659 ignoring the environment varaible TCPNAL_PORT, as well as other
2660 TCPNAL_ environment variables
2665 Description: two objects could be created on the same OST for a single file
2666 Details : If an OST is down, in some cases it was possible to create two
2667 objects on a single OST for a single file. No problems other
2668 than potential performance impact and spurious error messages.
2672 Bugzilla : 5681, 9562
2673 Description: Client may oops in ll_unhash_aliases
2674 Details : Client dcache may become inconsistent in race condition.
2675 In some cases "getcwd" can fail if the current directory is
2681 Description: Inode refcounting problems in NFS export code
2682 Details : link_raw functions used to call d_instantiate without obtaining
2683 extra inode reference first.
2687 Bugzilla : 9942, 9903
2688 Description: Referencing freed requests leading to crash, memleaks with NFS.
2689 Details : We used to require that call to ll_revalidate_it was always
2690 followed by ll_lookup_it. Also with revalidate_special() it is
2691 possible to call ll_revalidate_it() twice for the same dentry
2692 even if first occurence returned success. This fix changes semantic
2693 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2694 reference on a request referred from the intent.
2695 ll_intent_release() then releases such a request.
2698 Frequency : rare, normally benchmark loads only
2700 Description: unlinked inodes were kept in memory on the client
2701 Details : If a client is repeatedly creating and unlinking files it
2702 can accumulate a lot of stale inodes in the inode slab cache.
2703 If there is no other client load running this can cause the
2704 client node to run out of memory. Instead flush old inodes
2705 from client cache that have the same inode number as a new inode.
2708 Frequency : SLES9 2.6.5 kernel and long filenames only
2709 Bugzilla : 9969, 10379
2710 Description: utime reports stale NFS file handle
2711 Details : SLES9 uses out-of-dentry names in some cases, which confused
2712 the lustre dentry revalidation. Change it to always use the
2716 Frequency : rare, unless heavy write-truncate concurrency is continuous
2717 Bugzilla : 4180, 6984, 7171, 9963, 9331
2718 Description: OST becomes very slow and/or deadlocked during object unlink
2719 Details : filter_destroy() was holding onto the parent directory lock
2720 while truncating+unlinking objects. For very large objects this
2721 may block other threads for a long time and slow overall OST
2722 responsiveness. It may also be possible to get a lock ordering
2723 deadlock in this case, or run out of journal credits because of
2724 the combined truncate+unlink. Solution is to do object truncate
2725 first in one transaction without parent lock, and then do the
2726 final unlink in a new transaction with the parent lock. This
2727 reduces the lock hold time dramatically.
2730 Frequency : rare, 2.4 kernels only
2732 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
2733 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
2734 before unmounting their filesystem in order to avoid deadlock.
2735 The kernel_locked() function in 2.4 kernels only checks whether
2736 the kernel lock is held, not whether it is this process that is
2737 holding it as 2.6 kernels do.
2742 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
2743 Details : The client ptlrpc code may be trying to reconnect to a down
2744 server before a previous connection attempt has timed out.
2745 Increase the reconnect interval to be longer than the connection
2746 timeout interval to avoid sending duplicate connections to
2750 Frequency : echo_client brw_test command
2752 Description: fix echo_client to work with OST preallocated code
2753 Details : OST preallocation code (5137) didn't take echo_client IO path
2754 into account: echo_client calls filter methods outside of any
2755 OST thread and, hence, there is no per-thread preallocated
2756 pages and buffers to use. Solution: hijack pga pages for IO. As
2757 a byproduct, this avoids unnecessary data copying.
2761 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
2762 Description: Client can oops in mdc_commit_close() after open replay
2763 Details : It was possible for the MDS to return an open request with no
2764 transaction number in mds_finish_transno() if the client was
2765 evicted, but without actually returning an error. Clients
2766 would later try to replay that open and may trip an assertion
2767 Simplify the client close codepath, and always return an error
2768 from the MDS in case the open is not successful.
2771 Frequency : rare, 2.6 OSTs only
2773 Description: OST may deadlock under high load on fragmented files
2774 Details : If there was a heavy load and highly-fragmented OST filesystems
2775 it was possible to have all the OST threads deadlock waiting on
2776 allocation of biovecs, because the biovecs were not released
2777 until the entire RPC IO was completed. Instead, release biovecs
2778 as soon as they are complete to ensure forward IO progress.
2780 Severity : enhancement
2782 Description: Support for specifying external journal device at mount
2783 Details : If an OST or MDS device is formatted with an external journal
2784 device, this device major/minor is stored in the ext3 superblock
2785 and may not be valid for failover. Allow detecting and
2786 specifying the external journal at mount time.
2791 Description: Mounting an MDS with pending unlinked files may cause oops
2792 Details : target_finish_recovery() calls mds_postrecov() which returned
2793 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
2794 considers this an error and immediately begins cleaning up the
2795 lov, just after starting the mds_lov process
2797 Severity : enhancement
2799 Description: Implement 'lfs df' to report actual free space on per-OST basis
2800 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
2801 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
2802 usage of objects; '-h' to report in human readable format.
2804 ------------------------------------------------------------------------------
2806 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
2813 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
2814 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
2815 at startup during orphan unlinking. Instead, skip buddy update
2816 before setup, as it will be regenerated after recovery is complete.
2821 Description: If an OST is inactive, its locks might reference stale inodes.
2822 Details : lov_change_cbdata() must iterate over all namespaces, even if
2823 they are inactive to clear inode references from the lock.
2825 Severity : enhancement
2826 Frequency : occasional, if non-standard max_dirty_mb used
2828 Description: Client will block write RPCs if not enough grant
2829 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
2830 then the client will block writes while waiting for another RPC
2831 to complete instead of consuming its dirty limit. With change
2832 we get improved performance when max_dirty_mb is small.
2834 Severity : enhancement
2835 Bugzilla : 3389, 6253
2836 Description: Add support for supplementary groups on the MDS.
2837 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
2838 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
2839 lookups for user supplementary groups into a cache.
2843 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
2844 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
2845 for liblustre O_EXCL is enforced by MDS.
2849 Description: readdir never returns NULL in liblustre.
2850 Details : Corrected llu_iop_getdirentries logic, to return offset of next
2851 dentry in struct dirent.
2855 Frequency : liblustre only, depends on application IO pattern
2856 Description: liblustre clients evicted if not contacting servers
2857 Details : Don't put liblustre clients into the ping_evictor list, so
2858 they will not be evicted by the pinger ever.
2860 Severity : enhancement
2862 Description: Add ability to evict clients by NID from MDS.
2863 Details : By echoing "nid:$NID" string into
2864 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
2865 $NID would be instantly evicted from this MDS and from all active
2866 OSTs connected to it.
2870 Description: Do not query file size twice, somewhat slowing stat(2) calls.
2871 Details : lookup_it_finish() used to query file size from OSTs that was not
2876 Description: service threads change working directory to that of init
2877 Details : Starting lustre service threads may pin the working directory
2878 of the parent thread, making that filesystem busy. Threads
2879 now change to the working directory of init to avoid this.
2883 Frequency : during shutdown only
2884 Description: shutdown with a failed MDS or OST can cause unmount to hang
2885 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
2890 Frequency : 2.6 only
2891 Description: chmod/chown may include an extra supplementary group
2892 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
2893 supplementary group and if none is specified this is used.
2896 Bugzilla : 5479 (6816)
2898 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
2899 Details : If lookup is in progress on a file that is unlinked we might try
2900 to revalidate the inode and fail in revalidate after lookup is
2901 complete and ll_file_open() enqueues the open again but
2902 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
2905 Frequency : always, if lconf --abort_recovery used
2907 Description: lconf --abort_recovery fails with 'Operation not supported'
2908 Details : lconf was attempting to abort recovery on the MDT device and not
2911 ------------------------------------------------------------------------------
2913 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
2918 Frequency : rare (only unsupported configurations with a node running as an
2920 Bugzilla : 6514, 5137
2921 Description: Mounting a Lustre file system on a node running as an OST could
2923 Details : OSTs now preallocates memory needed to write out data at
2924 startup, instead of when needed, to avoid having to
2925 allocate memory in possibly low memory situations.
2926 Specifically, if the file system is mounted on on OST,
2927 memory pressure could force it to try to write out data,
2928 which it needed to allocate memory to do. Due to the low
2929 memory, it would be unable to do so and the node would
2930 become unresponsive.
2932 Severity : enhancement
2934 Description: Addition of lconf --service command line option
2935 Details : lconf now accepts a '--service <arg>' option, which is
2936 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
2938 Severity : enhancement
2940 Description: Failover mode is now the default for OSTs.
2941 Details : By default, OSTs will now run in failover mode. To return to
2942 the old behaviour, add '--failout' to the lmc line for OSTs.
2944 Severity : enhancement
2946 Description: Health checks are now provided for MDS and OSTs
2947 Details : Additional detailed health check information on MSD and OSTs
2948 is now provided through the procfs health_check value.
2951 Frequency : occasional, depends on IO load
2953 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
2954 after numerous create/delete cycles
2955 Details : The ext3 inode allocation policy would not allocate new inodes
2956 very well on the OSTs because there are no new directories
2957 being created. Instead we look for groups with free space if
2958 the parent directories are nearly full.
2963 Description: Network or server problems during mount may cause partially
2964 mounted clients instead of returning an error.
2965 Details : The config llog parsing code may overwrite the error return
2966 code during mount error handling, returning success instead
2972 Description: MDS can fail to allocate large reply buffers
2973 Details : After long uptimes the MDS can fail to allocate large reply
2974 buffers (e.g. zconf client mount config records) due to memory
2975 fragmentation or consumption by the buffer cache. Preallocate
2976 some large reply buffers so that these replies can be sent even
2977 under memory pressure.
2981 Frequency : rare (liblustre)
2982 Description: fsx running with liblustre complained that using truncate() to
2983 extend the file doesn't work. This patch corrects that issue.
2984 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
2985 ATTR_SIZE and lsm use in llu_setattr_raw.
2989 Frequency : rare, only 2.6 kernels
2990 Description: Unusual file access patterns on the MDS may result in inode
2991 data being lost in very rare circumstances.
2992 Details : Bad interaction between the ea-in-inode patch and the "no-read"
2993 code in the 2.6 kernel caused the inode and/or EA data not to
2994 be read from disk, causing single-file corruption.
2998 Frequency : rare, only 2.6 filesystems using extents
2999 Description: Heavy concurrent write and delete load may cause data corruption.
3000 Details : It was possible under high-load situations to have an extent
3001 metadata block in the block device cache from a just-unlinked
3002 file overwrite a newly-allocated data block. We now unmap any
3003 metadata buffers that alias just-allocated data blocks.
3007 Frequency : filesystems with default stripe_count larger than 77
3008 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3009 Details : lconf specifies an inode size of 4096 bytes when the default
3010 stripe_count is larger than 77. This conflicts with the default
3011 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3012 this case to avoid pinning too much memory for large EAs.
3014 ------------------------------------------------------------------------------
3016 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3021 Frequency : rare (extremely heavy IO load with hundreds of clients)
3023 Description: Client is evicted, gets IO error writing to file
3024 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3025 caused clients to be evicted for AST timeouts. The fixes in
3026 bug 5192 mean we no longer need to have such short AST timeouts
3027 so ldlm_timeout has been increased.
3030 Frequency : occasional during --force or --failover shutdown under load
3031 Bugzilla : 5949, 4834
3032 Description: Server oops/LBUG if stopped with --force or --failover under load
3033 Details : a collection of import/export refcount and cleanup ordering
3034 issues fixed for safer force cleanup
3037 Frequency : only filesystems larger than 120 OSTs
3038 Bugzilla : 5990, 6223
3039 Description: lfs getstripe would oops on a very large filesystem
3040 Details : lov_getconfig used kfree on vmalloc'd memory
3043 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3044 Bugzilla : 6242, 6243
3045 Description: reading from files that had been truncated to a non-zero size
3046 but never opened returned no data
3047 Details : ll_file_read() reads zeros from no-object files to EOF
3052 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3053 Details : There's little harm in aborting MDS/OSS recovery and letting it
3054 try again, so I removed the LASSERT and return an error instead.
3056 Severity : enhancement
3058 Description: New debugging infrastructure for tracking down data corruption
3059 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3060 (b) cover more of the client-side code path, and (c) try to narrow
3061 down where problems occurred
3065 Bugzilla : 3819, 4364, 4397, 6313
3066 Description: Racing close and eviction MDS could cause assertion in mds_close
3067 Details : It was possible to get multiple mfd references during close and
3068 client eviction, leading to one thread referencing a freed mfd.
3070 Severity: : enhancement
3071 Bugzilla : 3262, 6359
3072 Description: Attempts to reconnect to servers are now more aggressive.
3073 Details : This builds on the enhanced upcall-less recovery that was added
3074 in 1.4.2. When trying to reconnect to servers, clients will
3075 now try each server in the failover group every 10 seconds. By
3076 default, clients would previously try one server every 25 seconds.
3081 Description: After recovery, certain operations trigger a failed
3082 assertion on a client.
3083 Details : Failing over an mds, using lconf -d --failover, while a
3084 client was doing a readdir() call would cause the client to
3085 LBUG after recovery completed and the readdir() was resent.
3087 Severity : enhancement
3089 Description: Default groups are now added by lconf
3090 Details : You can now run lconf --group <servicename> without having to
3091 manually add groups with lmc.
3094 Frequency : occasional
3096 Description: Nodes with an elan id of 0 trigger a failed assertion
3099 Frequency : always when accessing e.g. tty/console device nodes
3101 Description: tty and some other devices nodes cannot be used on lustre
3102 Details : file's private_data field is used by device data and lustre
3103 values in there got lost. New field was added to struct file to
3104 store fs-specific private data.
3107 Frequency : when exporting Lustre via NFS
3109 Description: NFSD failed occasionally when looking up a path component
3110 Details : NFSD is looking up ".." which was broken in ext3 directories
3111 that had grown large enough to become hashed.
3114 Frequency : Clusters with multiple interfaces not on the same subnet
3116 Description: Nodes will repeatedly try to reconnect to an interface which it
3117 cannot reach and report an error to the log.
3118 Details : Extra peer list entries will be created by lconf with some peers
3119 unreachable. lconf now validates the peer before adding it.
3122 Frequency : Only if a default stripe is set on the filesystem root.
3124 Description: Setting a default stripe on the filesystem root prevented the
3125 filesystem from being remounted.
3126 Details : The client was sending extra request flags in the root getattr
3127 request and did not allocate a reply buffer for the dir EA.
3130 Frequency : occasional, higher if lots of files are accessed by one client
3131 Bugzilla : 6159, 6097
3132 Description: Client trips assertion regarding lsm mismatch/magic
3133 Details : While revalidating inodes the VFS looks up inodes with ifind()
3134 and in rare cases can find an inode that is being freed.
3135 The ll_test_inode() code will free the lsm during ifind()
3136 when it finds an existing inode and then the VFS later attaches
3137 this free lsm to a new inode.
3141 Bugzilla : 6422, 7030
3142 Description: MDS deadlock between mkdir and client eviction
3143 Details : Creating a new file via mkdir or mknod (starting a transaction
3144 and getting the ns lock) can deadlock with client eviction
3145 (gets ns lock and trying to finish a synchronous transaction).
3148 Frequency : occasional
3149 Description: While starting a server, the fsfilt_ext3 module could not be
3151 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3152 kernels. Previously, lconf would still use the ext3 name
3153 when trying to load modules. Now, it will correctly use
3154 ext3 on 2.4 and ldiskfs on 2.6.
3156 Severity : enhancement
3157 Description: The default stripe count has been changed to 1
3158 Details : The interpretation of the default stripe count (0, to lfs
3159 or lmc) has been changed to mean striping across a single
3160 OST, rather than all available. For general usage we have
3161 found a stripe count of 1 or 2 works best.
3163 Severity : enhancement
3164 Description: Add support for compiling against Cray portals.
3165 Details : Conditional compiling for some areas that are different
3169 Frequency : occasional
3170 Bugzilla : 6409, 6834
3171 Description: Creating files with an explicit stripe count may lead to
3172 a failed assertion on the MDS
3173 Details : If some OSTs are full or unavailable, creating files may
3174 trigger a failed assertion on the MDS. Now, Lustre will
3175 try to use other servers or return an error to the
3179 Frequency : occasional
3181 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3182 caused lock timeout during readahead (since 1.4.2).
3183 Details : Processes doing readahead might match a lock that hasn't been
3184 granted yet if there are overlapping and conflicting lock
3185 requests. The readahead process waits on ungranted lock
3186 (original lock is CBPENDING), while OST waits for that process
3187 to cancel CBPENDING read lock and eventually evicts client.
3189 Severity : enhancement
3191 Description: Initial enabling of flock support for clients
3192 Details : Implements fcntl advisory locking and file status functions.
3193 This feature is provided as an optional mount flag (default
3194 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3195 locking are implemented yet, and those that are are not guaranteed
3196 to be completely correct in production environments.
3197 mount -t lustre -o [flock|noflock] ...
3200 Frequency : occasional
3202 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3203 assertion in the ext3 JBD (journaling) layer.
3204 Details : The b_committed_data struct is protected by the big kernel lock
3205 in 2.4 kernels, serializing journal_commit_transaction() and
3206 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3207 access to this struct. In 2.6 kernels there is finer grained
3208 locking to improve SMP performance of the JBD layer.
3212 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3217 Description: merge and cleanup kernel patches.
3220 -----------------------------------------------------------------------------
3222 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3224 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3225 versions of Lustre. Please update all servers and clients to
3226 version 1.4.2 or later at the same time. You must also run
3227 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3228 to update the configuration logs.
3230 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3231 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3232 - swab llog records of type '0' so we get proper header size/idx (5861)
3233 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3234 - fix rename of one directory over another leaking an inode (5953)
3235 - avoid SetPageDirty on 2.6 (5981)
3236 - don't re-add just-being-destroyed locks to the waiting list (5653)
3237 - when creating new directories, inherit the parent's custom
3238 striping settings if present parent (3048)
3239 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3240 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3241 - don't hold client locks on temporary worklist from l_lru (5666)
3242 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3243 - automatically evict dead clients (5921)
3244 - Update file size properly in create+truncate+fstat case (6196)
3245 - Do not unhash mountpoint dentries, do not allow removal of
3247 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3248 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3249 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3250 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3251 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3252 - Correctly handle reads of files with no objects (6243)
3253 - lctl recover will also mark a device active if deactivate used (5933)
3255 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3256 - allow --write-conf on an MDS with different nettype than client (5619)
3257 - don't write config llogs to MDS for mounts not from that MDS (5617)
3258 - lconf should create multiple TCP connections from a client (5201)
3259 - init scripts are now turned off by default; run chkconfig --on
3260 lustre and chkconfig --on lustrefs to use them
3261 - upcalls are no longer needed for clients to recover to failover
3263 - add --abort-recovery option to lconf to abort recovery on device
3265 - add support for an arbitrary number of OSTs (3026)
3266 - Quota support protocol changes.
3267 - forward compatibility changes to wire structs (6007)
3268 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3269 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3270 - improved "lustre status" script
3271 - initialize blocksize for non-regular files (6062)
3272 - added --disable-server and --disable-client configure options (5782)
3273 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3274 - Vanilla 2.4.29 support
3275 - increase maximum number of obd devices to 520 (6242)
3276 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3277 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3278 SLES 9 2.6.5 kernels
3279 - limit stripes per file to 160 (the maximum EA size) (6093)
3281 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3284 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3285 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3286 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3287 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3288 - small ext3 extents cleanups and fixes (5733)
3289 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3290 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3291 - avoid needless client->OST connect, fix handle mismatch (5317)
3292 - fix DLM error path that led to out-of-sync client, long delays (5779)
3293 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3294 - fix several locking issues related to i_size (5492,5624,5654,5672)
3295 - don't move pending lock onto export if it is already evicted (5683)
3296 - fix kernel oops when creating .foo in unlinked directory (5548)
3297 - fix deadlock in obdfilter statistics vs. object create (5811)
3298 - use time_{before,after} to avoid timer jiffies wrap (5882)
3299 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3300 - Do not leak request if server was not able to process it (5154)
3301 - If mds_open unable to find parent dir, make that negative lookup(5154)
3302 - don't create new directories with extent-mapping (5909, 5936)
3304 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3305 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3306 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3307 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3308 - support for RHEL4 kernel on i686 (5773)
3309 - provide error messages when incompatible logs are encountered (5898)
3311 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3312 * version 1.4.0.10 (1.4.1 release candidate 1)
3314 - don't keep a lock reference when lock is not granted (4238)
3315 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3316 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3317 - drop import inflight refcount on signal_completed_replay error (5255)
3318 - unlock page after async write error during send (3677)
3319 - handle missing objects in filter_preprw_read properly (5265)
3320 - no transno return for symlink open, don't save no-trasno open (3440)
3321 - don't try to complete elan receive that already failed (4012)
3322 - free RPC server reply state on error (5406)
3323 - clean up thread from ptlrpc_start_thread() on error (5160)
3324 - readahead could read extra page into cache that wasn't ejected (5388)
3325 - prevent races in class_attach/setup/cleanup/detach (5260)
3326 - don't dereference de->d_inode after l_dput of de (5458)
3327 - use "int" for stripe value returned from lock_to_stripe (5544)
3328 - mballoc allocation and error-checking fixes in 2.6 (5504)
3329 - block device patches to fix I/O request sizes in 2.6 (5482)
3330 - look up hostnames for IB nals (5602)
3331 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3332 - don't start multiple acceptors for the same port (5277)
3333 - fix incorrect LASSERT in mds_getattr_name (5635)
3334 - export a proc file for general "ping" checking (5628)
3335 - fix "lfs check" to not block when the MDS is down (5628)
3337 - service request history (4965)
3338 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3339 - create an "evict_client" /proc entry on OSTs, like the MDS has
3340 - fix mount usage message, return errors per mount(8) (5168)
3341 - change grep [] to grep "[]" in tests so they work in more UMLs
3342 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3343 - remove ancient LOV_MAGIC_V0 EA support (5047)
3344 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3345 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3346 - allow readahead allocations to fail when low on memory (5383)
3347 - mmap locking landed again, after considerable improvement (2828)
3348 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3350 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3353 - send OST transaction number in read/write reply to free req (4966)
3354 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3355 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3356 - the watchdog thread now runs as interruptible (5246)
3357 - flock/lockf fixes (but it's still disabled, pending 5135)
3358 - don't use EXT3 constants in llite code (5094)
3359 - memory shortage at startup could cause assertion (5176)
3361 - reorganization of lov code
3362 - single portals codebase
3364 - add extents/mballoc support (5025)
3365 - direct I/O reads in the obdfilter (4048)
3366 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3368 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3371 - send OST transaction number in read/write reply to free req (4966)
3372 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3373 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3374 - the watchdog thread now runs as interruptible (5246)
3375 - handle missing objects in filter_preprw_read properly (5265)
3376 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3377 - drop import inflight refcount on signal_completed_replay error (5255)
3378 - unlock page after async write error during send (3677)
3379 - return original error code on reconstructed replies (3761)
3380 - no transno return for symlink open, don't save no-trasno open (3440)
3382 - add pid to ldlm debugging output (4922)
3383 - bump the watchdog timeouts -- we can't handle 30sec yet
3384 - extra debugging for orphan dentry/inode bug (5259)
3386 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3389 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3390 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3391 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3392 - fix NULL dereference in /proc/sys/portals/routes (4827)
3393 - allow failed mdc_close() operations to be interrupted (4561)
3394 - stop precreate on OST before MDS would time out on it (4778)
3395 - don't send partial-page writes before EOF from client (4410)
3396 - discard client grant for sub-page writes on large-page clients (4520)
3397 - don't free dentries not owned by NFS code, check generation (4806)
3398 - fix lsm leak if mds_create_objects() fails (4801)
3399 - limit debug_daemon file size, always print CERROR messages (4789)
3400 - use transno after validating reply (3892)
3401 - process timed out requests if import state changes (3754)
3402 - update mtime on OST during writes, return in glimpse (4829)
3403 - add mkfsoptions to LDAP (4679)
3404 - use ->max_readahead method instead of zapping global ra (5039)
3405 - don't interrupt __l_wait_event() during strace
3407 - add software watchdogs to catch hung threads quickly (4941)
3408 - make lustrefs init script start after nfs is mounted
3409 - fix CWARN/ERROR duplication (4930)
3410 - return async write errors to application if possible (2248)
3411 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3412 - print NAL number in %x format (4645)
3413 - update barely-supported suse-2.4.21-171 series (4842)
3414 - support for sles 9 %post scripts
3415 - support for building 2.6 kernel-source packages
3416 - support for sles km_* packages
3418 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3421 - ignore -ENOENT errors in osc_destroy (3639)
3422 - notify osc create thread that OSC is being cleaned up (4600)
3423 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3424 - reconstruct ost_handle() like mds_handle() (4657)
3425 - create a new thread to do import eviction to avoid deadlock (3969)
3426 - let lconf resolve symlinked-to devices (4629)
3427 - don't unlink "objects" from directory with default EA (4554)
3428 - hold socknal file ref over connect in case target is down (4394)
3429 - allow more than 32000 subdirectories in a single directory (3244)
3430 - fix blocks count for O_DIRECT writes (3751)
3431 - OST returns ENOSPC from object create when no space left (4539)
3432 - don't send truncate RPC if file size isn't changing (4410)
3433 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3434 - bind to privileged port in socknal and tcpnal (3689)
3436 - rate limit CERROR/CWARN console message to avoid overload (4519)
3437 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3438 - basic mmap support (3918)
3439 - kernel patch series update from b1_4 (4711)
3441 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3444 - avoid crash during MDS cleanup with OST shut down (2775)
3445 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3446 - don't use bad inodes on the MDS (3744)
3447 - dynamic object preallocation to improve recovery speed (4236)
3448 - don't hold spinlock over lock dumping or change debug flags (4401)
3449 - don't zero obd_dev when it is force cleaned (3651)
3450 - print grants to console if they go negative (4431)
3451 - "lctl deactivate" will stop automatic recovery attempts (3406)
3452 - look for existing locks in ldlm_handle_enqueue() (3764)
3453 - don't resolve lock handle twice in recovery avoiding race (4401)
3454 - revalidate should check working dir is a directory (4134)
3456 - don't always mark "slow" obdfilter messages as errors (4418)
3458 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3461 - don't close LustreDB during write_conf until it is done (3860)
3462 - fix typo in lconf for_each_profile (3821)
3463 - allow dumping logs from multiple threads at one time (3820)
3464 - don't allow multiple threads in OSC recovery (3812)
3465 - fix debug_size parameters (3864)
3466 - fix mds_postrecov to initialize import for llog ctxt (3121)
3467 - replace config semaphore with spinlock (3306)
3468 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3469 - don't allow enqueue to complete on a destroyed export (3822)
3470 - down write_lock before checking llog header bitmap (3825)
3471 - recover from lock replay timeout (3764)
3472 - up llog sem before sending rpc (3652)
3473 - reduce ns lock hold times when setting kms (3267)
3474 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3475 - fix NULL deref and obd_dev leak on setup error (3312)
3476 - replace some LBUG about llog ops with error handling (3841)
3477 - don't match INVALID dentries from d_lookup and spin (3784)
3478 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3479 - fix invalid assertion in ptlrpc_set_wait (3880)
3481 - add libwrap support for the TCP acceptor (3996)
3482 - add /proc/sys/portals/routes for non-root route listing (3994)
3483 - allow setting MDS UUID in .xml (2580)
3484 - print the stack of a process that LBUGs (4228)
3486 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3489 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3490 - make sure to unset replay flag from failed open requests (3440)
3491 - if default stripe count is 0, use OST count for inode size (3636)
3492 - update parent mtime/ctime on client for create/unlink (2611)
3493 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3494 - free recovery state on server during a forced cleanup (3571)
3495 - unregister_reply for resent reqs (3063)
3496 - loop back devices mounting and status check on 2.6 (3563)
3497 - fix resource-creation race that can provoke i_size == 0 (3513)
3498 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3499 - more debugging for page-accounting assertion (3746)
3500 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3501 - avoid deadlock after precreation failure (3758)
3502 - fix race and lock order deadlock in orphan handling (3450, 3750)
3503 - add validity checks when grabbing inodes from l_ast_data (3599)
3505 - add /proc/.../recovery_status to obdfilter (3428)
3506 - lightweight CDEBUG infrastructure, debug daemon (3668)
3507 - change default OSC RPC parameters to be better on small clusters
3508 - turn off OST read cache for files smaller than 32MB
3509 - install man pages and include them in rpms (3100)
3510 - add new init script for (un)mounting lustre filesystems (2593)
3511 - run chkconfig in %post for init scripts (3701)
3512 - drop scimac NAL (unmaintained)
3514 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3517 - clean kiobufs before and after use (3485)
3518 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3519 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3520 - change init script to fail more clearly if not run as root (1528)
3521 - allow clients to reconnect during replay (1742)
3522 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3523 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3524 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3525 - fix cond_resched (3554)
3526 - only evict unfinished clients after recovery (3515)
3527 - allow bulk resend, prevent data loss (3570)
3528 - dynamic ptlrpc request buffer allocation (2102)
3529 - don't allow unlinking open directory if it isn't empty (2904)
3530 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3531 - remove extraneous obd dereference causing LASSERT failure (3334)
3532 - don't use get_cycles() when creating temp. files on the mds (3156)
3533 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3534 - handle EEXIST for set-stripe, set proper directory name (3336)
3536 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3537 - fix ksocknal_fmb_callback() error messages (2918)
3539 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3542 - don't copy lvb into (possibly NULL) reply on error (2983)
3543 - don't deref dentry after dput, don't free lvb on error (2922)
3544 - use the kms to determine writeback rpc length (2947)
3545 - increment oti_logcookies when osc is inactive (2948)
3546 - update client's i_blocks count via lvb messages (2543)
3547 - handle intent open/close of special files properly (1557)
3548 - mount MDS with errors=remount-ro, like obdfilter (2009)
3549 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3550 - don't use cancelling-locks' kms values (2947)
3551 - use highest lock extent for kms, not last one (2925)
3552 - don't dereference ERR_PTR() dentry in error handling path (3107)
3553 - fix thread race in portals_debug_dumplog() (3122)
3554 - create lprocfs device entries at setup instead of at attach (1519)
3555 - common AST error handler, don't evict client on completion race (3145)
3556 - zero nameidata in detach_mnt in 2.6 (3118)
3557 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3558 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3559 - initialize RPC timeout timer earlier for 2.6 (3219)
3560 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3561 - print nal/nid for unknown nid (3258)
3562 - additional checks for oscc recovery before doing precreate (3284)
3563 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3564 - don't crash in mdc_close for bad permissions on open (3285)
3565 - zero i_rdev for non-device files (3147)
3566 - clear page->private before handing to FS, better assertion (3119)
3567 - tune the read pipeline (3236)
3568 - fix incorrect decref of invalidated dentry (2350)
3569 - provide read-ahead stats and refine rpc in flight stats (3328)
3570 - don't hold journal transaction open across create RPC (3313)
3571 - update atime on MDS at close time (3265)
3572 - close LDAP connection when recovering to avoid server load (3315)
3573 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3574 - don't leak open file on MDS after open resend (3325)
3575 - serialize filter_precreate and filter_destroy_precreated (3329)
3576 - loop device shouldn't call sync_dev() for nul device (3092)
3577 - clear page cache after eviction (2766)
3578 - resynchronize MDS->OST in background (2824)
3579 - refuse to mount the same filesystem twice on same mountpoint (3394)
3580 - allow llmount to create routes for mounting behind routers (3320)
3581 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3582 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3583 - fix rare problem with rename on htree directories (3417)
3585 - allow default OST striping configuration per directory (1414)
3586 - fix compilation for qswnal for 2.6 kernels (3125)
3587 - increase maximum number of MDS request buffers for large systems
3588 - change liblustreapi to be useful for external progs like lfsck (3098)
3589 - increase local configuration timeout for slow disks (3353)
3590 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3592 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3595 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3596 - don't overwrite extent policy data in reply if lock was blocked (2901)
3597 - drop filter export grants atomically with removal from device (2663)
3598 - del obd_self_export from work_list in class_disconnect_exports (2908)
3599 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3600 - swab reply message in mdc_close, other PPC fixes (2464)
3601 - fix destroying of named logs (2325)
3602 - overwrite old logs when running lconf --write_conf (2264)
3603 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3604 - fix race in target_handle_connect (2898)
3605 - mds_reint_create() should take same inode create lock (2926)
3606 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3607 - don't close files for self_export to avoid uninitialized obd (2936)
3608 - allow MDS with the same name as client node (2939)
3609 - hold dentry reference for closed log files for unlink (2325)
3610 - reserve space for all logs during transactions (2059)
3611 - don't evict page beyond end of stripe extent (2925)
3612 - don't oops on a deleted current working directory (2399)
3613 - handle hard links to targets without a parent properly (2517)
3614 - don't dereference NULL lock when racing during eviction (2867)
3615 - don't grow lock extents when lots of conflicting locks (2919)
3617 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3620 - account for cache space usage on clients to avoid data loss (974)
3621 - lfsck support in lustre kernel code (2349)
3622 - reduce journal credits needed for BRW writes (2370)
3623 - orphan handling to avoid losing space on client/server crashes
3624 - ptlrpcd can be blocked, stopping ALL progress (2477)
3625 - use lock value blocks to assist in proper KMS, faster stat (1021)
3626 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3627 - recovery for initial connections (2355)
3628 - fixes for mds_cleanup_orphans (1934)
3629 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3630 - block all file creations until orphan recovery completes (1901)
3631 - client remove rq_connection from request struct (2423)
3632 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3633 - recovery timer race (2670)
3634 - mdc_close recovey bug (2532)
3635 - ptlrpc cleanup bug (2710)
3636 - mds timeout on local locks (2588)
3637 - namespace lock held during RPCs (2431)
3638 - handle interrupted sync write properly (2503)
3639 - don't try to handle a message that hasn't been replied to (2699)
3640 - client assert failure during cleanup after abort recovery (2701)
3641 - leak mdc device after failed mount (2712)
3642 - ptlrpc_check_set allows timedout requests to complete (2714)
3643 - wait for inflight reqs when ptlrpcd finishes (2710)
3644 - make sure unregistered services are removed from the srv_list
3645 - reset bulk XID's when resending them (caught by 1138 test)
3646 - unregister_bulk after timeout
3647 - fix lconf error (2694)
3648 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3649 - readahead locks pages, leaves pending causing memory pressure (2673)
3650 - increase OST request buffers to 4096 on large machines (2729)
3651 - fix up permission of existing directories in simple_mkdir (2661)
3652 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3653 - don't assign transno to errored transactions (2742)
3654 - don't delete objects on OST if given a bogus objid from MDS (2751)
3655 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3656 - if rq_no_resend, then timeout request after recovery (2432)
3657 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3658 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3659 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3660 - MDS recovery shouldn't delete objects that it has given out (2730)
3661 - if enqueue arrives after completion, don't clobber LVB (2819)
3662 - don't unlock pages twice when trigger_group_io returns error (2814)
3663 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3664 - don't write pages to disk if there was an error (1450)
3665 - don't ping imports that have recovery disabled (2676)
3666 - take buffered bytes into account when balancing socknal conn (2817)
3667 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3668 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3669 - remove little-endian swabbing of llog records (1987)
3670 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3671 - retry reposting request buffers if they fail (1191)
3672 - grow extent at grant time to avoid granting a revoked lock (2809)
3673 - lock revoke doesn't evict page if covered by a second lock (2765)
3674 - disable VM readahead to avoid reading outside lock extents (2805)
3676 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3677 - updated kernel patches for hp-2.4.20 kernel (2681)
3679 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3682 - fix truncated write corruption (2366)
3683 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3685 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3686 - ensure len(uuid) < 37 in lmc (1171)
3687 - fix ia64 OOPS in llog_test (2255)
3688 - zero end of page at obdfilter for partial page writes (2648)
3689 - don't leave stale dentries around after renames (bug 2428)
3690 - fix timeouts when evicting a client with a single lock held (2642)
3691 - set deadline for the initial HELLO message to drain (2634)
3692 - print out dotted-quad IP addresses in the socknal (2302)
3694 - additional debugging for MDS client eviction problem (2443)
3695 - fix mkfsoptions support for osts (2603, 2604)
3697 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3700 - add series for the vanilla 2.6.0 kernel
3701 - add series for the vanilla 2.4.24 kernel
3702 - add series for a cray x86/64 UL kernel drop
3703 - fix xattr patches for the vanilla 2.4.19 series
3705 - generate true UUIDs in lmc (1171)
3706 - have portals stack dumping break in UML (2466)
3707 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
3708 - allocate with _NOFS in ldlm to avoid deadlock (1933)
3709 - wake callback waiting threads on client eviction (2460)
3710 - Add --ptldebug and --subsystem to lmc (1719)
3711 - update assertion to allow safe interrupt allocation
3712 - set rq_no_resend for cancel requests (2432)
3713 - recalculate ptlrpcd timeout after resend (2494)
3714 - call vfs_rmdir when removing pending directories (2368)
3715 - fix renaming a file to itself (2429)
3716 - lmc creates a default one-stripe lov (2454)
3717 - expand procfs space to handle large clusters (2326)
3718 - increase UML stack to avoid overflow
3719 - update lconf's list of debug and subsystem masks
3720 - fix lfs find --obd (2510)
3721 - /proc tunable for disabling filter read caching (2591)
3722 - stop rpm packages from altering slapd.conf (2301)
3723 - disable nagle in the socknal under 0conf (2578)
3724 - choose mds inode size based on stripe count (2572)
3725 - fix kernel-source rpm problems (2516)
3727 - add --disable-doc to avoid pdf generation (2421)
3728 - update documentation, tests, type-os, comments
3729 - avoid format warnings on ia64
3730 - remove the TOE NAL
3731 - tiny code cleanups by removing unused fields
3733 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
3736 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
3737 - remove the most blatant lies from BUILDING file (2371)
3738 - change default debug level to reasonable production setting
3739 - reduce client side cache size to reduce cache flush time
3740 - reduce max RPCs in flight to avoid unnecessary file fragmentation
3741 - make TCP zerocopy and pinger support enabled by default (2476)
3742 - sync writes completed after process exits caused crashes (2319)
3743 - maintain correct mount count on the MDS (2356)
3744 - backout 1557, because 2316 wasn't really fixed
3745 - better file I/O statistics gathering in /proc
3746 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
3747 - another kernel patch to fix zero-copy TCP function export
3748 - don't take duplicate lock when processing re-sent getattr (2420)
3749 - lctl uses obd_self_export instead of creating new conn (2353)
3750 - MDS/OST recovery case which requires object creation asserted (2425)
3751 - move lfs from /usr/sbin to /usr/bin in packages
3752 - fix race between mds_client_add and mds_client_free (2417)
3753 - use kmalloc instead of slabs in portals (2430)
3754 - don't create duplicate records when a failover MDS is present (2442)
3755 - remove unnecessary mount age check (2332)
3756 - don't remove directory inodes from locks prematurely (2451)
3757 - don't break if MDS service name is the same as hostname (2103)
3758 - fix races in client write RPC generation when cache full (2482)
3760 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
3763 - remove now-unused request->rq_obd (278)
3764 - if an allocation fails, print out how much memory we've used (1933)
3765 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
3766 - add the "configurable stack size" patch to most series files (1256)
3767 - ability to write large log records, for 100+ OST configs (2306)
3768 - fix NULL deref when filter_prep fails (2314)
3769 - fix operator precedence error in filter_sync
3770 - dynamic allocation of socknal TX descriptors (2315)
3771 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
3772 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
3773 - ensure that CWARN messages in Portals always get to the syslog
3774 - __init/__exit are not for prototype decls (ldlm_init/exit)
3775 - x86-64 compile warning fixes
3776 - fix gateway LMC keyword conflict (2318)
3777 - fix MDS lock inversions in getattr/reint paths (1844)
3778 - fix a rare lock re-ordering bug, which caused deadlock (2322)
3779 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
3780 - DLM race condition prevented some lock evictions (2328)
3781 - ENOMEM detection and retry on socknal sends (2230)
3782 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
3783 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
3784 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
3785 - better error messages when a client is rejected during recovery (1505)
3786 - avoid cancelling locks which were never granted, after failure (2330)
3787 - fix i_sem/journal inversion in mds_client_add (2333)
3788 - fix truncate/getattr lock cycle deadlock (2334)
3789 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
3790 - fix two rare exit paths which could leak an l_lock() ref (2321)
3791 - fencepost error in MDS/OST orphan recovery (2226)
3792 - make log record alignment 8 bytes (1988)
3793 - lstripe now fails when requested offset > ost_count (2237)
3794 - ensure that all kernel series have a complete list.h (1607)
3795 - fix crashes in special-file operations (2316)
3796 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
3798 - allow configurable automake binary, for testing new versions
3799 - small update to the lfs documentation
3801 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
3803 * fix negative export reference count in fsfilt_sync (2312)
3805 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
3806 * release candidate 0.9.1
3808 - orphans are moved into the PENDING directory for possible recovery
3809 - replayed opens now open by fid for orphan/rename safety (1042)
3810 - last close of an orphan inode generates a transno (683)
3811 - chdir() and mount() now pin the directory entry (1020)
3812 - avoid CERROR in normal ll_setattr_raw() error case (1500)
3813 - discard very old requests without processing them (1502)
3814 - remove some common, well-understood CERRORs (1505)
3815 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
3816 - clear "grant" flags in OST replies until OST grant code lands (1644)
3817 - fix read performance by not clobbering i_blksize on client (1598)
3818 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
3819 - make LRU size a /proc tunable, clears locks when reduced (707)
3820 - fix some lprocfs rot that prevented ptlbd from loading (1732)
3821 - server locks take references on exports now (1558)
3822 - build fixes for 2.4.20-rh trees (1663)
3823 - return an error from lov_create if all OSCs are inactive (1751)
3824 - fix import levels when a reconnect happens without a timeout (1597)
3825 - exit early from mds_open if we get a lookup error (1749)
3826 - partial page read at EOF wouldn't wait for disk before sending (1642)
3827 - avoid NULL deref in obdfilter when reading page past EOF (1592)
3828 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
3829 - fix LBUG in ll_it_open_error with rc = -2 (1861)
3830 - write/truncate lock inversion (1639)
3831 - Don't auto-load obdclass, portals modules during cleanup (1495)
3832 - fix timestamps from jumping to "now" (1763)
3833 - extra journal assertions (1648)
3834 - add an extra multiunlink test (1771)
3835 - fix read_record/write_record API (1776)
3836 - fix leak of offset_extent, possible incorrect i_size later (1772)
3837 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
3838 - Debugging for the kqswnal_get_idle_tx problems (1820)
3839 - Allow recovery to be attempted multiple times (1536)
3840 - Write out MDS last_rcvd file after it is first created (1600)
3841 - Fix tx_descriptor leak in failed transmit situations (1827)
3842 - ext3 journaling fixes for assertion failure after IO error (1871)
3843 - class_export_put() on freed export after completion AST error (1896)
3844 - Fix revalidate looping in VFS (1322)
3845 - Don't access a freed export during MDS_REINT timeout (1521)
3846 - Add open-unlink recovery support on the MDS (1673,1764)
3847 - Return an error if no MDS data was read from last_rcvd (1946)
3848 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
3849 - Don't setup a disk device that doesn't match exported UUID (317)
3850 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
3851 - avoid committing NULL handle in force close
3852 - local.sh is now a one-stripe LOV configuration
3853 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
3854 - fix ext3 htree duplicate directory entry corruption (1516)
3855 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
3856 - update to new LOV EA format (2097)
3857 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
3859 2003-06-15 Phil Schwan <phil@clusterfs.com>
3862 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
3863 - per-import/export recovery handling (958, 931, 959)
3864 - multiple last-rcvd slots, for serving multiple FSes (949)
3865 - connections are again shared between multiple imp/exports (963, 964)
3866 - "umount -f" would hang if any requests needed to be sent (393, 978)
3867 - avoid pinning large req buffer by copying for queued messages (989)
3868 - add "uuid" to "lctl device" command to help upcalls (991)
3869 - "open" RPCs with transnos would confuse recovery counters (1037)
3870 - do proper endian conversion of all wire messages (288, 340, 891)
3871 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
3872 - call waiting locks callback from LDLM recovery thread (1127, 1151)
3873 - fix ptlrpc_connection leak in target_handle_connect (1174)
3874 - fix import refcounting bug in OST and MDS cleanup (1134)
3875 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
3876 - fix very unlikely obd_types race condition (501)
3877 - remove osc_open hack for echo_client (1187)
3878 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
3879 - a failure in read_inode2 leads to deadlock (1139)
3880 - cancel ack-locks as soon as transaction is committed (1072)
3881 - fix major leaks and crashes in the bulk I/O path (937, 1057)
3882 - make sure to commitrw after any preprw to avoid deadlock (1162)
3883 - failing to execute a file in a lustre FS would lock inode (1203)
3884 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
3885 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
3886 - obd_uuid2tgt would walk off the end of the list (1255)
3887 - on IA64 the osc would give portals incorrect bulk size (1258)
3888 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
3889 - fix lock inversion caused by new llite matching code (1282)
3890 - limit the number of dirty pages on a client to 10MB (1286)
3891 - timed out locks were not being corrected cancelled (1289)
3892 - fix O_DIRECT above 4GB on IA-32 (1292)
3893 * major user-visible changes
3894 - fail out/fail over policy now controlled by the upcall (993)
3896 - add OBD_PING to check server availability and failure (954)
3897 - lustre messages are now sent in sending host order (288, 340, 891)
3898 - add eadatalen to MDS getattr reply (340)
3899 - OST read replies may contain second buffer, with per-page status (593)
3901 2003-03-11 Phil Schwan <phil@clusterfs.com>
3904 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
3905 - failed open()s could cause deadlock; fixed (867, 869)
3906 - stop cancelling OST locks when files are closed (481)
3907 - overlapping XID spaces caused network corruption (851, 853)
3908 - fix unsafe fsfilt counter arithmetic; change to atomic_t
3909 - setattr_raw added, to do single-RPC, server-side setattrs
3910 - lmc/lconf syntax change for OST UUIDs
3911 - fix crashy race condition between ptlrpc_free_req and osc_close
3912 - don't use request in mdc_enqueue if we hit a timeout (889)
3913 - don't set the inode i_size for regular files from the MDS (896)
3914 - handle out of order completion AST (842)
3915 - don't LBUG if a lock request times out after receiving AST (913)
3916 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
3917 - if a bad lock AST arrives, send an error instead of dropping entirely
3918 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
3919 - fix leak in bulk IO when only partially completed (899, 900, 926)
3920 - fix O_DIRECT for ia64 (55)
3921 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
3922 - C-z after timeout could hang a process forever; fixed (977)
3924 - client-side I/O cache (678, 924, 929, 941, 970)
3926 - READPAGE and SETATTRs which don't take server-side locks get
3929 2003-02-11 Phil Schwan <phil@clusterfs.com>
3932 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
3933 - Fix fsfilt_extN_readpage() to read a full page of directory
3934 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
3935 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
3936 - Fix imbalanced LOV object allocation and out-of-bound access (469)
3937 - Most intent operations were removed, in favour of a new RPC mode
3938 that does a single RPC to the server and bypasses most of the VFS
3939 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
3940 - Aggressively cancel local locks on DLM servers
3941 - mds_reint_unlink sends EA to the client if it's the last nlink.
3942 client uses that EA to unlink OST objects.
3943 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
3944 - recursive symlinks were fixed (439)
3945 - fixed NULL deref in DEBUG_REQ
3946 - filter_update_lastobjid no longer calls sync, which annoyed extN
3947 - fixed multi-client small-writes to a single file problem (445)
3948 - fixed mtime updates during file writes (607)
3949 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
3950 - fixed bug in obd_brw_read/write() (under guise of testing 367)
3951 - fixed Linux OST size reporting problem (444, 656)
3952 - OST now updates object mtime with writes or setattr (607, 619)
3953 - client verifies file size before zeroing page past EOF (445)
3954 - OST now writes last allocated objid to disk with allocation (108)
3955 - LOV on echo now works (409)
3957 - mds_reint_unlink sends a new buffer, with the EA included. this
3958 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
3959 set if a regular file was being unlinked, and it was the last link
3960 - use PtlGet from the target for bulk writes (315)
3961 - OST now updates object mtime with writes or setattr (607, 619)
3962 - LDLM now has a grant-time callback to revalidate locked items, if
3964 - Many MDS operations were reorganized to combat race conditions
3966 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
3967 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
3969 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
3972 - Fully reactivate OST imports after reconnection (512, others)
3973 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
3974 - More graceful error handling for truncating on dead OST (515)
3975 - Don't error out unless we're actually accessing dead stripes (474)
3976 - Fix garbage sizes when stripes are missing (410)
3977 - LRU counters were broken, causing constant lock purge (433, 432)
3978 - garbage on read from stripes with failed OSTs (441)
3979 - mark OSCs as active before reconnecting during recovery (438)
3980 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
3981 - lfind did not preserve OST order in output (443)
3982 - symlinks cause hung clients, incorrect data (439)
3983 - stop dereferencing request after dropping refcount (457)
3984 - don't LASSERT(spin_is_locked) on non-SMP (455)
3985 - fixes for many rename() bugs
3986 - fstat didn't correctly synchronize attributes (399)
3987 - server must handle lock cancellation during blocking AST prep (487)
3988 - bulk descriptors were free()d too soon (511)
3989 - fix paths in lconf, which would load incorrect modules (451, 507)
3990 - fix confusing lconf 'host not found' error message (386)
3991 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
3992 - fix race condition in mdc_blocking_ast() for inode access (526)
3993 - fix lov_unpackmd() unpacking wrong number of stripes (537)
3994 - fix lov_set_osc_active() marking wrong OSC inactive (440)
3995 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
3996 - fix multiple writes of stripe MD to MDS (358, maybe 519)
3997 - fix lstripe in several ways (kernel side) (527)
3998 - fix request leak in ldlm_cli_enqueue (262)
3999 - incorrect OSC was marked inactive after OST failure
4000 - call mds_fs_cleanup before unmounting filesystem (524)
4001 - fix races between taking ns_lock and ldlm_lock_change_resource
4002 - fix races updating LOV export open file list
4003 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4004 - fix recovery NULL deref in ldlm_cli_cancel_unused
4005 - fix some DLM races by using new hash table for lock handles (419)
4006 - permit the client to specify desired inodes, at replay
4007 - duplicate requests when we queue them for replay reintegration
4008 - fix last_rcvd offset calculation
4009 - sync after each recovered transaction, so we always make progress
4010 - never, not always, ERESTART requests without transnos
4011 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4013 - skip replay if the MDS says that the client is already connected
4014 - don't check for a recovery-enabled export to match lctl's UUID
4015 - don't INC_USE_COUNT for phantom exports
4016 - don't crash when cleaning up phantom exports (567)
4017 - don't double-finish or set replay data for errored mdc_open requests
4018 - abort requests when they time out, so we don't get old replies
4019 - send/receive replies for AST messages again
4020 - if the client says that it doesn't have the lock, cancel it on the
4022 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4023 mark it as destroyed, it will all work out when decref is called
4024 - fix module use counts (22, 581)
4026 - ASTs now expect a reply (server cancels lock on error reply)
4028 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4031 - fix many simultaneous client startup (392)
4032 - fix dentry->d_it clobbering
4033 - credentials weren't being shipped for readdir/getattr operations
4034 - remove invalid assertions triggered during some concurrent MD
4036 - proper Lustre versions added (336, 389)
4037 - fix memory leak for create error case (398)
4038 - fix LOV locking bug that would get cli/srv out of sync
4039 - fix echo client over LOV (409)
4040 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4041 - fix double-O_EXCL intent crash (424)
4042 - avoid sending multiple lock CANCELs (352)
4044 - MDS can do multi-client recovery (modulo bugs in new code)
4046 - many updates, edits, cleanups
4048 2002-11-18 Phil Schwan <phil@clusterfs.com>
4051 - fix null d_it dereference (346)
4052 - fix full OST/dbench hang (333)
4053 - fix permission problem with file removal (286)
4054 - fix removal of OSCs from LOV when they fail
4055 - fix NULL deref during bulk timeout (214)
4056 - fix problems related to multiple filesystems on one MDS (241)
4057 - fixed serious subtle metadata locking bugs
4058 - free locks on clients when inodes are removed due to memory
4060 - fix inode pointer in lock data (285)
4061 - partial support for multiple MDS on a single host (241)
4062 - data locks weren't cancelled at clear_inode time (290, 311)
4063 - intent locks could lead to unbounded lock growth (205)
4064 - added a maximum lock count, an LRU list, and a flusher
4065 - fix multiple rename (365)
4066 - properly abstracted the echo client
4067 - OSC locked 1 byte too many; fixed
4068 - rewrote brw callback code:
4069 - fixed recovery bugs related to LOVs (306)
4070 - fixed too-many-pages-in-one-write crash (191)
4071 - fixed (again) crash in sync_io_timeout (214)
4072 - probably fixed callback-related race (385)
4074 - Add capability to MDS protocol
4075 - LDLM cancellations and callbacks on different portals
4077 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4080 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4081 - defer open object destruction to close time (601981, 138)
4082 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4083 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4084 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4085 - don't call obd_setattr in ll_file_release for destroyed objects
4086 * protocol change to lustre_msg: move |version| and add |flags|
4087 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4088 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4089 * added replay of create, unlink, link and rename operations during
4090 MDS failover; recovery should be much more robust now
4091 * remove failed OSCs from LOVs (only lov_create uses this so far)
4092 * the lustre-HOWTO was brought (more) up to date (582544)
4094 2002-10-23 Phil Schwan <phil@clusterfs.com>
4097 - in-use dentries weren't being reused properly (617851)
4098 - prevent multiple LDLM setup (599178)
4099 - fix LOV size calculations for truncate (617853)
4100 - fix client handling of MDS intent errors (POSIX)
4101 - fix permission bug in lovstripe.c test (624321)
4102 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4103 - truncate past end of file could corrupt data
4104 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4105 - a race in recovery could return ETIMEDOUT to apps (623947)
4106 - building outside the source directory was fixed
4107 * the lustre-HOWTO was brought (more) up to date (582544)
4108 * major progress was made on recovery functionality
4110 2002-10-10 Phil Schwan <phil@clusterfs.com>
4113 - recovery deadlock fix
4114 - rm -rf causes LBUG fix (617817)
4115 - file open by multiple tasks fix (618962)
4116 - directory permissions bugs (602707 and 620007)
4117 - journal_stop fixed with locking (611313)
4118 - O_APPEND failures resolved (618273, perhaps 614459)
4119 - lconf PATH fix (619770)
4120 - IA64 build fix (621450)
4121 - RPC buffer sizes scale with amount of memory
4123 2002-10-01 Phil Schwan <phil@clusterfs.com>
4126 - locks would be cancelled without throwing away data pages,
4127 resulting in inconsistent data (605627)
4128 - inode attributes were not always being refreshed (605627, 612449)
4129 - lconf now continues to cleanup after lctl reports an error
4130 - MDS now enforces user permissions (602707)
4131 - lprocfs cleanup fixed, but not yet enabled (614157)
4132 - fixed infinite server hang, should a client not respond to an AST
4133 - avoid going into recovery if user calls readlink() with a buffer
4134 that's too small (613941)
4135 - AST RPCs no longer require replies (614867) -- this may be changed
4136 - don't crash server if client sends an IOV that's too big (611336)
4137 - fixed lock conversion deadlock (611892)
4138 - fixed the following of symlinks (614622)
4139 * recovery: the server can remove locks from a client that dies, other
4140 clients can make progress
4141 * more extN patch fixes
4142 * compile-time configurable ptlrpc buffer allocations
4144 - collaborative read cache document
4145 - Lustre Lite Performance CDR document-in-progress
4147 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4150 - fix typo in patch-2.4.18
4152 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4155 - clear ptlrpc request each time in handle_incoming_request()
4156 - unlink of files now destroys the object on the OST
4158 2002-09-19 Peter Braam <braam@clusterfs.com>
4160 * add hard link support
4161 * change obdfile creation method
4162 * kernel patch changed
4164 2002-09-19 Peter Braam <braam@clusterfs.com>
4167 - stack overflow bug in extN fixed
4169 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4171 * documentation updates
4172 - add man pages for config tools
4173 - update tests/README to describe testing with new config tools
4174 - finish metadata API descriptions
4175 * bug fixes and cleanups
4176 - statfs workaround for 16TB limit
4177 - LOV stripe allocation improved, can stripe on subset of OSTs
4178 - LOV file size/IO offset was wrong for files > 4GB in size
4179 - object EA data was being dropped, caused files to be unreadable
4180 - memory overflow with non-LOV OST caused memory corruption
4181 - fixed regression tests to work with new config tools, obdfilter
4182 - fixed bug when directory size became larger than 1 block
4183 - fixed bug (for single client case) when PWD was deleted
4184 - invalidate local directory pages when doing intent-based ops
4185 - avoid LDLM oops when lock callback contained bad data
4187 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4189 * documentation updates
4190 * bug fixes and cleanups
4191 - configuration tools
4194 - 64-bit compile warnings
4195 - 64-bit internal statfs data
4197 * test_brw on persistent OST devices
4199 * lprocfs (disabled)
4201 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4203 * documentation updates
4204 * bug fixes and cleanups
4205 * configuration tools
4207 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4210 * many small fixes to 0_5_4
4211 * io/network handling
4212 * thinkos in MDS operations
4214 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4217 * crucial basic fixes to 0.5.3
4218 * IOR, Iozone work over Elan
4221 2002-08-07 Phil Schwan <phil@clusterfs.com>
4222 * version 0_5_3, our first alpha
4223 * we use the new Portals iovs
4224 * documentation updates
4225 * bug fixes and cleanups
4226 * small changes in the DLM wire protocol
4228 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4229 * version 0_5_1 with some initial stability,
4230 * locking on MD and file I/O.
4231 * documentation updates
4232 * several bug fixes since 0.5.0
4233 * small changes in wire protocol
4235 2002-07-18 Phil Schwan <phil@clusterfs.com>
4237 * delivered as Lustre Light Alpha
4238 * fixed a crash after handling invalid MDS requests
4239 * fixed directory pages for architectures with non-4k pages sizes
4241 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4242 * release version v0_4_4
4243 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4245 * Fixes a number of interruption problems with OST operations.
4246 * Update documentation for portals header changes
4247 * Move all wire protocol structs/defines to lustre_idl.h
4248 * Fixes symlink length bug.
4249 * Add tcpdump to repository.
4251 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4252 * release version v0_4_3
4253 * Fixes statfs for inodes on extN.
4254 * Fixes bug in runtests which would delete /etc/hosts.
4255 * Use 64-bit object IDs wherever possible (not into VFS though)
4256 Remove ost_get_info, which is unused by lustre, and out of date.
4258 2002-07-03 Peter Braam <braam@clusterfs.com>
4259 * release version v0_4_2 Fixes a lookup error (type not passed)
4260 * move forward to head of Portals
4261 * move forward to latest Lustre kernel
4263 2002-06-25 Peter Braam <braam@clusterfs.com>
4264 * release version v0_4_1. Hopefully stable on single node use.