1 tbd Sun Microsystems, Inc.
4 2.6.9-67.0.4.EL (RHEL 4),
5 2.6.16.54-0.2.5 (SLES 10),
6 2.6.18-53.1.13.el5 (RHEL 5).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.4-cfs1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
16 Frequency : on start mds
18 Description: Implement get_info(last_id) in obdfilter.
21 Frequency : occasional
23 Description: Correctly check stale fid, not start epoch if ost not support SOM
24 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
25 overwrited with new generated fid, but mds can anwer with one of these
26 two fids and both is not stale. setattr incorectly start epoch and
27 assume will be called done_writeting, but without SOM done_writing
31 Frequency : rare, depends on device drivers and load
33 Description: MDS or OSS nodes crash due to stack overflow
34 Details : Code changes in 1.8.0 increased the stack usage of some functions.
35 In some cases, in conjunction with device drivers that use a lot
36 of stack the MDS (or possibly OSS) service threads could overflow
37 the stack. One change which was identified to consume additional
38 stack has been reworked to avoid the extra stack usage.
41 Frequency : occasional
43 Description: Do not fail import if osc_interpret_create gets -EAGAIN
44 Details : If osc_interpret_create got -EAGAIN it immediately exits and
45 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
46 oscc_has_objects and see OSC has no objests and call
47 oscc_internal_create to resend create request.
49 Severity : enhancement
51 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
53 Severity : enhancement
55 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
60 Description: Fix replay if there is an un-replied request and open
61 Details : In some cases, older replay request will revert the
62 mcd->mcd_last_xid on MDS which is used to record the client's
65 Severity : enhancement
67 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
69 Severity : enhancement
71 Description: Add rhel5 support to HEAD.
73 Serverity : enhancement
75 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
80 Description: Don't allow skipping OSTs if index has been specified.
81 Details : Don't allow skipping OSTs if index has been specified, make locking
82 in internal create lots better.
86 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
87 Details : access to bitfield in structure is always rounded to long
88 and this produce problem with not atomic change any bit.
92 Description: Lustre make rpms failed.
93 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
94 builds Lustre rpms from tarball.
97 Frequency : rare on shutdown ost
99 Description: If llog cancel was not send before clean_exports phase, this can
100 produce deadlock in llog code.
101 Details : If llog thread has last reference to obd and call class_import_put
102 this produce deadlock because llog_cleanup_commit_master wait when
103 last llog_commit_thread exited, but this never success because was
104 called from llog_commit_thread.
108 Description: allow userland application know is lost one of stripes.
109 Details : fill lvb_blocks with error code on ost and return it to
110 application if error flag found.
114 Description: NULL lov_tgts causing MDS oops
115 Details : more safe checks for NULL lov_tgts for avoid oops.
117 Severity : enhancement
119 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
123 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
124 Details : If an OST is down the MDS will hang indefinitely in
125 obd_statfs_rqset() waiting for the statfs data. While for
126 MDS QOS usage of statfs, it should not stuck in waiting.
128 Severity : enhancement
130 Description: remote_acl support
131 Details : Support ACL-based permission check for remote user.
132 Support setfacl/getfacl for remote user with the utils
133 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
134 the system "{s,g}etfacl" utils.
136 Severity : enhancement
138 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
140 Severity : enhancement
142 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
146 Description: make mgs_setparam() handle fsname containing dash
147 Details : fsname containing a dash does not work with lctl conf_param
149 Severity : enhancement
151 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
153 Severity : enhancement
155 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
159 Frequency : rare, at shutdown
160 Description: access already free / zero obd_namespace.
161 Details : if client_disconnect_export was called without force flag set,
162 and exist connect request in flight, this can produce access to
163 NULL pointer (or already free pointer) when connect_interpret
164 store ocd flags in obd_namespace.
168 Frequency : only at startup
169 Description: not alloc memory with spinlock held.
170 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
171 if any spinlock held.
173 Severity : enhancement
175 Description: make lustre randomly fail allocating memory
176 Details : Make lustre randomly failed allocating memory for testing purpose.
178 Severity : enhancement
180 Description: lost problems with lov objid file
181 Details : Fixes some scability and access to not inited memory problems
182 in work with lov objdid file.
187 Description: lfs find does not continue on file error
188 Details : Continue other files processing when a file/dir is absent.
192 Description: Inconsistent usage of lustre_pack_reply()
193 Details : Standardize the usage of lustre_pack_reply() such that it
194 always generate a CERROR on failure.
197 Frequency : occasional
199 Description: MDS hang or stay in waiting lock
200 Details : If client receive lock with CBPENDING flag ldlm need send lock
201 cancel as separate rpc, to avoid situation when cancel request
202 can't processed due all i/o threads stay in wait lock.
206 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
210 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
214 Description: Update to SLES9 kernel-2.6.5-7.287.3.
218 Description: Update to RHEL4 latest kernel.
220 Severity : enhancement
222 Description: Build SLES10 patchless client fails
223 Details : The configure was broken by run ./configure with
224 --with-linux-obj=.... argument for patchless client. When the
225 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
226 can't search header adequately. Use absolute path such as
227 -I($LINUX)/include instead.
231 Description: interrupt oig_wait produce painc on resend.
232 Details : brw_redo_request can be used for resend requests from ptlrpcd and
233 private set, and this produce situation when rq_ptlrpcd_data not
234 copyed to new allocated request and triggered LBUG on assert
235 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
236 ptlrpcd set if request is changed and can be safety changed to use
241 Description: LASSERT_{REQ,REP}SWAB macros are buggy
242 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
243 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
244 these macros and replace them with their difinitions instead.
248 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
249 Details : Update which_patch & target file for SLES10 latest kernel.
252 Description: Remove client patches from SLES 10 kernel.
253 Details : This causes SLES 10 clients to behave as patchless clients
254 even on a Lustre-patched (server) kernel.
256 Severity : enhancement
258 Description: self-adjustable client's lru lists
259 Details : use adaptive algorithm for managing client cached locks lru
260 lists according to current server load, other client's work
261 pattern, memory activities, etc. Both, server and client
262 side namespaces provide number of proc tunables for controlling
265 Severity : enhancement
267 Description: light-weight GSS support
268 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
269 and reduce performance overhead.
271 Severity : enhancement
273 Description: Linux keyring support
274 Details : Support using service of Linux keyring for Lustre GSS internal
275 context refresh/cache mechanism.
279 Description: Fix errors in lfs documentation
280 Details : Fixes man pages
284 Description: don't use GFP_* in generic Lustre code.
285 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
289 Description: obdclass is limited by single OBD_ALLOC(idarray)
290 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
294 Description: warnings with build patchless client with vanila 2.6.19 and up
295 Details : change old ctl_table style and replace ctl_table/ctl_table_header
296 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
300 Frequency : Only for SLES
301 Description: sanity_quota fail test_1
302 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
303 file, which makes TSTID[2] inunique.
307 Frequency : Only for SLES9
308 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
309 Details : Update target/ChangeLog/which_patch .
313 Frequency : Only for test_10 in sanity-quota.sh
314 Description: a bug in quota test code
315 Details : checking a test flag in a wrong place
319 Frequency : Only for test_13 in sanity-quota.sh
320 Description: a bug in quota test script
321 Details : shouldn't check overfull allocated quota
325 Description: jbd statistics
326 Details : Port older jbd statistics patch for sles10
328 Severity : enhancement
330 Description: Add printing inode info into message about error in writepage.
335 Description: reply_lock_interpret crash due to race with it and lock cancel.
336 Details : Do not replay locks that are being cancelled. Do not reference
337 locks by their address during replay, just by their handle.
341 Frequency : When flocks are used.
342 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
343 Details : Flock locks might destroy just granted lock if it could be merged
344 with another existing flock, this is done in completion handler,
345 so teach ldlm_cli_enquque_fini that this is a valid case for
351 Description: Oops in read and write path when failing to allocate lock.
352 Details : Check if lock allocation failed and return error back.
356 Description: lstripe command fails for valid OST index
357 Details : The stripe offset is compared to lov->desc.ld_tgt_count
358 instead of lov->desc.ld_active_tgt_count.
362 Description: sanity.sh failed test 103
363 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
368 Description: df doesn't work properly if diskfs blocksize != 4K
369 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
373 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
374 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
378 Description: obd_health_check_timeout too short
379 Details : set obd_health_check_timeout as 1.5x of obd_timeout
383 Description: llapi_file_create() does not allow some changes
384 Details : add llapi_file_open() that allows specifying the mode and
385 open flags, and also returns an open file handle.
389 Description: merge and cleanup kernel patches.
390 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
394 Description: Add journal checksum support.(Kernel part)
395 Details : The journal checksum feature adds two new flags i.e
396 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
397 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
398 indicates that the commit block contains the checksum for
399 the blocks described by the descriptor blocks. Now commit
400 record can be sent to disk without waiting for descriptor
401 blocks to be written to disk. This behavior is controlled
402 using JBD2_FEATURE_ASYNC_COMMIT flag.
406 Description: OSS needs mutliple precreate threads
407 Details : Add ability to start more than one create thread per OSS.
411 Description: Sanity.sh test_65k failed: setstripe should have succeeded
412 Details : execute lfs setstripe on client
416 Description: mds_obd_create error creating tmp object
417 Details : When the user sets quota on root, llog will be affected and can't
418 create files and write files.
421 Frequency : Always on ia64 patchless client, and possibly others.
423 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
424 Details : This allows the patchless client to be loaded on architectures
429 Description: RedHat Update kernel for RHEL5
430 Details : Add the kernel config file for RHEL5.
434 Description: RedHat Update kernel for RHEL5
435 Details : Modify the kernel config file more closer RHEL5.
439 Description: Build failure against Centos5 (RHEL5)
440 Details : Define PAGE_SIZE when it isn't present.
444 Description: test_7 on sanity-quota.sh can't be used on separate machines
445 Details : add facet to handle it
449 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
450 Details : with NFS, the anon dentry's parent was set to itself in
451 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
452 corresponding dentry other than use rec->ur_name.
454 Severity : enhancement
456 Description: lfs setstripe enhancement
457 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
461 Description: enable data checksumming by default
462 Details : enable checksum by default, allow --disable-checksum
463 configure option and "-o nochecksum" mount option. Checksums
464 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
468 Description: lustre support for RHEL5
469 Details : Add support for RHEL5.
473 Description: Client eviction due to failover config
474 Details : after a connection loss, the lustre client should attempt to
475 reconnect to the last active server first before trying the
476 other potential connections.
480 Description: when mds and osts use different quota unit(32bit and 64bit),
481 quota will be released repeatly.
482 Details : void sending multiple quota reqs to mds, which will keep the
483 status between the reqs.
487 Description: osts not allocated evenly to files
488 Details : change the condition to increase offset_idx
491 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
493 Description: Data corruption for OSTs that are formatted larger than 2TB
495 Details : When generating the bio request for lustre file writes the
496 sector number would overflow a temporary variable before being
497 used for the IO. The data reads correctly from Lustre (which
498 will overflow in a similar manner) but other file data or
499 filesystem metadata may be corrupted in some cases.
503 Description: Tune the kernel for good SCSI performance.
504 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
505 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
510 Description: rewrite ext2-derived code in obdclass/uuid.c
511 Details : rewrite inherited code (uuid parsing code from ext2 utils)
512 from scratch preserving functionality.
516 Description: "lfs find -obd UUID" prints directories
517 Details : "lfs find -obd UUID" will return all directory names instead
518 of just file names. It is incorrect because the directories
519 do not reside on the OSTs.
523 Description: use i_size_read and i_size_write in 2.6 port
524 Details : replace inode->i_size access with i_size_read/write()
527 Frequency : when removing large files
529 Description: scheduling issue during removal of large Lustre files
530 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
531 It causes scheduling issues when removing large files (17TB in the
535 Frequency : only with liblustre clients on XT3
537 Description: evictions taking too long
538 Details : allow llrd to evict clients directly on OSTs
541 Frequency : only on ppc
543 Description: /proc/fs/lustre/devices broken on ppc
544 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
545 We should make sure the type of 'index' is loff_t and then cast
546 explicitly as needed below. Do not assign an explicitly cast
550 Frequency : only for rhel5
552 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
553 Details : Modify the target file & which_kernel.
555 Severity : enhancement
557 Description: omit set fsid for export NFS
558 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
559 as FS_REQUIRES_DEV for avoid problems with generate fsid.
563 Frequency : Always, for kernels after 2.6.16
564 Description: Fix warning idr_remove called for id=.. which is not allocated.
565 Details : Last kernels save old s_dev before kill super and not allow
566 to restore from callback - restore it before call kill_anon_super.
570 Description: buffer overruns could theoretically occur
571 Details : llapi_semantic_traverse() modifies the "path" argument by
572 appending values to the end of the origin string, and a
573 overrun may occur. Adding buffer overrun check in liblustreapi.
577 Description: Fix error on 'ls .' at the top of the Lustre mount.
578 Details : Don't revalidate dentry if it is a root dentry.
582 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
583 Details : Modify vm-tunables-rhel4.patch.
587 Description: Kernel config for 2.6.18-vanilla.
588 Details : Modify targets/2.6-vanilla.target.in.
589 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
590 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
591 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
592 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
596 Description: improve handling recoverable errors
597 Details : if request processig with error which can be recoverable on server
598 request should be resend, otherwise page released from cache and
603 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
604 Details : Modify target file & which_patch.
605 A flaw was found in the IA32 system call emulation provided
606 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
607 value could be stored in the %RAX register, which could trigger an
608 out-of-bounds system call table access. An untrusted local user
609 could exploit this flaw to run code in the kernel
610 (ie a root privilege escalation). (CVE-2007-4573).
614 Description: O_DIRECT bypasses client statistics.
615 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
616 read_ahead_stats were updated. copied the stats section from
617 osc_send_oap_rpc() into async_internal().
621 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
625 Description: change order of libsysio includes
626 Details : '#include sysio.h' should always come before '#include xtio.h'
630 Description: handle "serious error: objid * already exists" more gracefully
631 Details : If LAST_ID value on disk is smaller than the objects existing in
632 the O/0/d* directories, it indicates disk corruption and causes an
633 LBUG(). If the object is 0-length, then we should use the existing
634 object. This will help to avoid a full fsck in most cases.
636 Severity : enhancement
638 Description: adapt the lustre_config script to support the upgrade case
639 Details : Add "-u" option for lustre_config script to support upgrading 1.4
640 server targets to 1.6 in parallel.
644 Description: To avoid grant space > avaible space when the disk is almost
645 full. Without this patch you might see the error "grant XXXX >
646 available" or some LBUG about grant, when the disk is almost
648 Details : In filter_check_grant, for non_grant cache write, we should
649 check the left space by if (*left > ungranted + bytes), instead
650 of (*left > ungranted), because only we are sure the left space
651 is enough for another "bytes", then the ungrant space should be
652 increase. In client, we should update cl_avail_grant only there
653 is OBD_MD_FLGRANT in the reply.
657 Description: Update RHEL 4 kernel to fix local root privilege escalation.
658 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
659 described in CVE-2007-4573. This problem could allow untrusted
660 local users to gain root access.
663 Frequency : when using O_DIRECT and quotas
665 Description: Incorrect file ownership on O_DIRECT output files
666 Details : block usage reported by 'lfs quota' does not take into account
667 files that have been written with O_DIRECT.
672 Description: touch file failed when fs is not full
673 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
674 otherwise we can get ENOSP while fs is not full.
678 Description: parallel lock callbacks
679 Details : Instead of sending blocking and completion callbacks as separated
680 requests, adding them to a set and sending in parallel.
683 Frequency : only for Cray XT3
684 Bugzilla : 12829/13455
685 Description: Changing primary group doesn't change the group lustre assigns to
687 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
688 overridden with the primary group provided by the group upcall,
689 whereas the supplied fsgid can be trusted if it is in the list of
690 supplementary groups returned by the group upcall.
692 Severity : enhancement
694 Description: Allow masking D_WARNING, D_ERROR messages from console
695 Details : Console messages can now be disabled via lnet.printk.
699 Description: User code with malformed file open parameter crashes client node
700 Details : Before packing join_file req, all the related reference should be
701 checked carefully in case some malformed flags cause fake
702 join_file req on client.
706 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
707 Details : ldlm_completion_ast() assumes that a lock is granted when the req
708 mode is equal to the granted mode. However, it should also check
709 that LDLM_FL_CP_REQD is not set.
713 Description: Heavy nfs access might result in deadlocks
714 Details : After ELC code landed, it is now improper to enqueue any mds
715 locks under och_sem, because enqueue might want to decide to
716 cancel open locks for same inode we are holding och_sem for.
720 Description: Client eviction while running blogbench
721 Details : A lot of unlink operations with concurrent I/O can lead to a
722 deadlock causing evictions. To address the problem, the number of
723 oustanding OST_DESTROY requests is now throttled to
724 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
725 callbacks are processed in priority.
729 Description: enable ACLs on MDS by default
730 Details : ACLs must be enabled on MDS by default.
733 Frequency : PPC/PPC64 only
735 Description: conflicts between asm-ppc64/types.h and lustre_types.h
736 Details : fix duplicated definitions between asm-ppc64/types.h and
737 lustre_types.h on PPC.
740 Frequency : PPC/PPC64 only
742 Description: asm-ppc/segment.h does not exist
743 Details : fix compile issue on PPC.
747 Description: better handle error messages in extents code
750 Frequency : RHEL4 only
752 Description: mkfs is very slow on IA64/RHEL4
753 Details : A performance regression has been discovered in the MPT Fusion
754 driver between versions 3.02.73rh and 3.02.99.00rh. As a
755 consequence, we have downgraded the MPT Fusion driver in the RHEL4
756 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
758 Severity : enhancement
760 Description: SNMP support enhancement
761 Details : Adding total number of sampled request for an MDS node in snmp
764 Severity : enhancement
766 Description: Optimize ldlm waiting list processing for PR extent locks
767 Details : When processing waiting list for read extent lock and meeting read
768 lock that is same or wider to it that is not contended, skip
769 processing rest of the list and immediatelly return current
770 status of conflictness, since we are guaranteed there are no
771 conflicting locks in the rest of the list.
775 Description: Time out and refuse to reconnect
776 Details : When the failover node is the primary node, it is possible
777 to have two identical connections in imp_conn_list. We must
778 compare not conn's pointers but NIDs, otherwise we can defeat
779 connection throttling.
783 Description: port llog fixes from b1_6 into HEAD
784 Details : Port llog reference couting and some llog cleanups from b1_6
785 (bug 10800) into HEAD, for protect from panic and access to already
786 free llog structures.
790 Description: Detect stride IO mode in read-ahead
791 Details : When a client does stride read, read-ahead should detect that and
792 read-ahead pages according to the detected stride pattern.
796 Description: data checksumming impacts single node performance
797 Details : add support for several checksum algorithm. Currently, only CRC32
798 and Adler-32 are supported. The checksum type can be changed on
799 the fly via /proc/fs/lustre/osc/*/checksum_type.
801 --------------------------------------------------------------------------------
803 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
805 * Support for kernels:
806 2.6.5-7.283 (SLES 9),
807 2.6.9-55.EL (RHEL 4),
808 2.6.16.46-0.14 (SLES 10),
809 2.6.18.8 vanilla (kernel.org)
810 * Client support for unpatched kernels:
811 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
812 2.6.16 - 2.6.22 vanilla (kernel.org)
813 * Due to recently discovered recovery problems, we do not recommend
814 using patchless RHEL 4 clients with this or any earlier release.
815 * Recommended e2fsprogs version: 1.39.cfs8
816 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
817 * Starting with this release, the ldiskfs backing filesystem required
818 by Lustre is now in its own package, lustre-ldiskfs. This package
819 should be installed. It is versioned separately from Lustre and
820 may be released separately in future.
824 Description: block reactivating mgc import until all deactivates complete
825 Details : Fix race when failing back MDT/MGS to itself (testing)
828 Frequency : at statup only
830 Description: mds_lov_synchronize race leads to various problems
831 Details : simultaneous MDT->OST connections at startup can cause the
832 sync to abort, leaving the OSC in a bad state.
834 Severity : enhancement
836 Description: add optional extra BUILD_VERSION info
837 Details : add a new environment variable (namely LUSTRE_VERS) which allows
838 to override the lustre version.
841 Frequency : 2.6.18 servers only
843 Description: ll_kern_mount() doesn't release the module reference
844 Details : The ldiskfs module reference count never drops down to 0
845 because ll_kern_mount() doesn't release the module reference.
850 Description: server LBUG when using old ost_num_threads parameter
851 Details : Accept the old ost_num_threads parameter but warn that it
852 is deprecated, and fix an off-by-one error that caused an LBUG.
857 Description: Transient SCSI error results in persistent IO issue
858 Details : iobuf->dr_error is not reinitialized to 0 between two
862 Frequency : sometimes when underlying device returns I/O errors
864 Description: OSTs not going read-only during write failures
865 Details : OSTs are not remounted read-only when the journal commit threads
866 get I/O errors because fsfilt_ext3 calls journal_start/stop()
867 instead of the ext3 wrappers.
871 Description: poor connect scaling with increasing client count
872 Details : Don't run filter_grant_sanity_check for more than 100 exports
873 to improve scaling for large numbers of clients.
876 Frequency : SLES10 only
878 Description: sanity-quota.sh quotacheck failed: rc = -22
879 Details : Quotas cannot be enabled on SLES10.
882 Frequency : liblustre clients only
884 Description: getdirentries does not give error when run on compute nodes
885 Details : getdirentries does not fail when the size specified as an argument
886 is too small to contain at least one entry
888 Severity : enhancement
890 Description: Add LNET router traceability for debug purposes
891 Details : If a checksum failure occurs with a router as part of the
892 IO path, the NID of the last router that forwarded the bulk data
893 is printed so it can be identified.
898 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
899 Details : Due to a race condition, liblustre clients were occasionally
902 Severity : enhancement
904 Description: lfs setstripe use optional parameters instead of postional
907 Severity : enhancement
909 Description: Nanosecond timestamp support for ldiskfs
910 Details : The on-disk ldiskfs filesystem has added support for nanosecond
911 resolution timestamps. There is not yet support for this at
912 the Lustre filesystem level.
915 Frequency : during server recovery
917 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
918 Details : request with rq_no_resend flag not awake l_wait_event if they get a
922 Frequency : nfs export on patchless client
924 Description: connectathon hang when test nfs export over patchless client
925 Details : Disconnected dentry cannot be found with lookup, so we do not need
926 to unhash it or make it invalid
929 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
930 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
931 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
936 Description: open req refcounting wrong on reconnect
937 Details : If reconnect happened between getting open reply from server and
938 call to mdc_set_replay_data in ll_file_open, we will schedule
939 replay for unreferenced request that we are about to free.
940 Subsequent close will crash in variety of ways.
941 Check that request is still eligible for replay in
942 mdc_set_replay_data().
947 Description: disable writes to filesystem when reading health_check file
948 Details : the default for reading the health_check proc file has changed
949 to NOT do a journal transaction and write to disk, because this
950 can cause reads of the /proc file to hang and block HA state
951 checking on a healthy but otherwise heavily loaded system. It
952 is possible to return to the previous behaviour during configure
953 with --enable-health-write.
955 Severity : enhancement
957 Description: 64-bit inode version
958 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
959 to the inode. This will be required for version-based recovery.
964 Description: MDS fails to start if a duplicate client export is detected
965 Details : in some rare cases it was possible for a client to connect to
966 an MDS multiple times. Upon recovery the MDS would detect this
967 and fail during startup. Handle this more gracefully.
969 Severity : enhancement
971 Description: Add -o localflock option to simulate old noflock
973 Details : This will achieve local-only flock/fcntl locks
979 Description: log_commit_thread vs filter_destroy race leads to crash
980 Details : Take import reference before releasing llog record semaphore
985 Description: Wrong request locking in request set processing
986 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
987 imp_delayed_list, in this place should be used imp_lock.
990 Frequency : when reconnection
992 Description: Grant Leak when osc reconnect to OST
993 Details : When osc reconnect ost, OST(filter) should check whether it
994 should grant more space to client by comparing fed_grant and
995 cl_avail_grant, and return the granted space to client instead
996 of "new granted" space, because client will call osc_init_grant
997 to update the client grant space info.
1000 Frequency : when client reconnect to OST
1002 Description: Grant Leak when osc do resend and replay bulk write
1003 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1004 bulk write request, because the grant info while be sync between
1005 OSC and OST when reconnect, and we should ignore the grant info
1006 these of resend/replay write req.
1011 Description: Grant space more than avaiable left space sometimes.
1012 Details : When then OST is about to be full, if two bulk writing from
1013 different clients came to OST. Accord the avaliable space of the
1014 OST, the first req should be permitted, and the second one
1015 should be denied by ENOSPC. But if the seconde arrived before
1016 the first one is commited. The OST might wrongly permit second
1017 writing, which will cause grant space > avaiable space.
1020 Frequency : when client is evicted
1022 Description: Grant might be wrongly erased when osc is evicted by OST
1023 Details : when the import is evicted by server, it will fork another
1024 thread ptlrpc_invalidate_import_thread to invalidate the
1025 import, where the grant will be set to 0. While the original
1026 thread will update the grant it got when connecting. So if
1027 the former happened latter, the grant will be wrongly errased
1028 because of this race.
1033 Description: Checking Stale with correct fid
1034 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1035 to check whether it is stale, because sometimes, we want the
1036 enqueue happened anyway, and op_data.fid2 will not be initialized.
1038 Severity : enhancement
1040 Description: update patchless client
1041 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1044 Frequency : only with 2.4 kernel
1046 Description: random memory corruption
1047 Details : size of struct ll_inode_info is to big for union inode.u and this
1048 can be cause of random memory corruption.
1053 Description: Memory leak in recovery
1054 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1055 It should also check obd_fail before fsfilt_start, otherwise if
1056 fsfilt_start return -EROFS,(failover mds during mds recovery).
1057 then the req will return with repmsg->transno = 0 and rc = EROFS.
1058 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1059 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1060 be freed no matter whether fsfilt_commit success or not.
1063 Frequency : only with huge count clients
1065 Description: Prevents from taking the superblock lock in llap_from_page for
1067 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1068 for llap_from_page call in ll_removepage prevents from taking the
1069 superblock lock for a soon died page.
1074 Description: Not check open intent error before release open handle
1075 Details : in some rare cases, the open intent error is not checked before
1076 release open handle, which may cause
1077 ASSERTION(open_req->rq_transno != 0), because it tries to release
1078 the failed open handle.
1083 Description: Set cat log bitmap only after create log success.
1084 Details : in some rare cases, the cat log bitmap is set too early. and it
1085 should be set only after create log success.
1089 Description: Accessing a block bevice can re-enable I/O when Lustre is
1090 tearing down a device.
1091 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1096 Description: service threads may hog cpus when there are a lot of requests
1097 Details : Insert cond_resched to give other threads a chance to use some CPU
1102 Description: the cat log was not initialized in recovery
1103 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1104 unlink log on mds will not be initialized until mds post
1105 recovery. And also in mds post recovery, the unlink log will
1106 initialization will be done asynchronausly, so there will be race
1107 between add unlink log and unlink log initialization.
1111 Description: brw_stats were being printed incorrectly
1112 Details : brw_stats were being printed as log2 but all of them were not
1113 recorded as log2. Also remove some code duplication arising from
1114 filter_tally_{read,write}.
1118 Frequency : rare, only in recovery.
1119 Description: ASSERTION(req->rq_type != LI_POISON) failed
1120 Details : imp_lock should be held while iterating over imp_sending_list for
1121 prevent destroy request after get timeout in ptlrpc_queue_wait.
1125 Description: replay-single.sh test 52 fails
1126 Details : A lock's skiplist need to be cleanup when it being unlinked
1127 from its resource list.
1131 Description: Short directio read returns full requested size rather than
1133 Details : Direct I/O operations should return actual amount of bytes
1134 transferred rather than requested size.
1136 Severity : enhancement
1138 Description: metadata RPC reduction (e.g. for rm performance)
1139 Details : decrease the amount of synchronous RPC between clients and servers
1140 by canceling conflicing lock before the operation on the client
1141 and packing thier handles into the main operation RPC to server.
1143 Severity : enhancement
1145 Description: Async OSC create to avoid the blocking unnecessarily.
1146 Details : If a OST has no remain object, system will block on the creating
1147 when need to create a new object on this OST. Now, ways use
1148 pre-created objects when available, instead of blocking on an
1149 empty osc while others are not empty. If we must block, we block
1150 for the shortest possible period of time.
1152 Severity : enhancement
1154 Description: refine locking for avoid write wrong info into lov_objid file
1155 Details : fix possible races with add new target and write/update data in
1158 --------------------------------------------------------------------------------
1160 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1165 Frequency : on some architectures
1167 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1168 Details : Uninitialized flags sometimes cause configuration commands to
1172 Frequency : patchless clients only
1174 Description: missing __iget() symbol export
1175 Details : The __iget() symbol export is missing. To avoid the need for
1176 this on patchless clients the deathrow inode reaper is turned
1177 off, and we depend on the VM to clean up old inodes. This
1178 dependency was during via the fix for bug 12181.
1180 --------------------------------------------------------------------------------
1182 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1184 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1185 INTEROPERATE with older versions automatically. In many cases a
1186 special upgrade step is needed. Please read the
1187 user documentation before upgrading any part of a live system.
1188 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1189 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1190 * WARNING: Lustre configuration and startup changes are required with
1191 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1197 Severity : enhancement
1199 Description: MountConf
1200 Details : Lustre configuration is now managed via mkfs and mount
1201 commands instead of lmc and lconf. New obd types (MGS, MGC)
1202 are added for dynamic configuration management. See
1203 https://mail.clusterfs.com/wikis/lustre/MountConf for
1206 Severity : enhancement
1208 Description: dynamic OST addition
1209 Details : OSTs can now be added to a live filesystem
1211 Severity : enhancement
1213 Description: startup order invariance
1214 Details : MDTs and OSTs can be started in any order. Clients only
1215 require the MDT to complete startup.
1217 Severity : enhancement
1219 Description: parallel, asynchronous orphan cleanup
1220 Details : orphan cleanup is now performed in separate threads for each
1221 OST, allowing parallel non-blocking operation.
1223 Severity : enhancement
1225 Description: optimized stripe assignment
1226 Details : stripe assignments are now made based on ost space available,
1227 ost previous usage, and OSS previous usage, in order to try
1228 to optimize storage space and networking resources.
1230 Severity : enhancement
1232 Description: Permanently set tunables
1233 Details : All writable /proc/fs/lustre tunables can now be permanently
1234 set on a per-server basis, at mkfs time or on a live system.
1236 Severity : enhancement
1238 Description: Lustre message v2
1239 Details : Add lustre message format v2.
1241 Severity : enhancement
1243 Description: client OST exclusion list
1244 Details : Clients can be started with a list of OSTs that should be
1245 declared "inactive" for known non-responsive OSTs.
1249 Description: SPEC SFS validation failure on NFS v2 over lustre.
1250 Details : Changes the blocksize for regular files to be 2x RPC size,
1251 and not depend on stripe size.
1253 Severity : enhancement
1255 Description: Multiple MD RPCs in flight.
1256 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1257 To avoid overly-overloading MDS, introduce a limit on number of
1258 MDS RPCs in flight for a single client and add /proc controls
1259 to adjust this limit.
1261 Severity : enhancement
1263 Description: client read/write statistics
1264 Details : Add client read/write call usage stats for performance
1265 analysis of user processes.
1266 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1267 file access. extents_stats shows chunk size distribution.
1268 extents_stats_per_process show chunk size distribution per
1271 Severity : enhancement
1273 Description: mds statistics
1274 Details : Add detailed mds operations statistics in
1275 /proc/fs/lustre/mds/*/stats.
1279 Description: Failure of copying files with lustre special EAs.
1280 Details : Client side always return success for setxattr call for lustre
1281 special xattr (currently only "trusted.lov").
1286 Description: Refcount LNET uuids
1287 Details : The global LNET uuid list grew linearly with every startup;
1288 refcount repeated list entries instead of always adding to
1291 Severity : enhancement
1293 Description: Dynamic service threads
1294 Details : Within a small range, start extra service threads
1295 automatically when the request queue builds up.
1298 Frequency : mixed-endian client/server environments
1300 Description: mixed-endian crashes
1301 Details : The new msg_v2 system had some failures in mixed-endian
1304 Severity : enhancement
1306 Description: Easy OST removal
1307 Details : OSTs can be permanently deactivated with e.g. 'lctl
1308 conf_param lustre-OST0001.osc.active=0'
1310 Severity : enhancement
1312 Description: MGS proc entries
1313 Details : Added basic proc entries for the MGS showing what filesystems
1316 Severity : enhancement
1318 Description: provide MGS failover
1319 Details : Added config lock reacquisition after MGS server failover.
1321 Severity : enhancement
1323 Description: add Linux 2.4 support
1324 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1328 Description: a large application tries to do I/O to the same resource and dies
1329 in the middle of it.
1330 Details : Check the req->rq_arrival time after the call to
1331 ost_brw_lock_get(), but before we do anything about
1332 processing it & sending the BULK transfer request. This
1333 should help move old stale pending locks off the queue as
1334 quickly as obd_timeout.
1337 Frequency : when an incorrect nid is specified during startup
1339 Description: ptlrpc connect to non-existant node causes kernel crash
1340 Details : LNET can't be re-entered from an event callback, which
1341 happened when we expire a message after the export has been
1342 cleaned up. Instead, hand the zombie cleanup off to another
1345 Severity : enhancement
1347 Description: plain/inodebits lock performance improvement
1348 Details : Grouping plain/inodebits in granted list by their request modes
1349 and bits policy, thus improving the performance of search through
1353 Frequency : only if OST filesystem is corrupted
1355 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1356 Details : for a short time RPCs with bulk IO are in the replay list,
1357 but replay of bulk IOs is unimplemented. If the OST filesystem
1358 is corrupted due to disk cache incoherency and then replay is
1359 started it is possible to trip an assertion. Avoid putting
1360 committed RPCs into the replay list at all to avoid this issue.
1363 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1365 Description: System hang on startup
1366 Details : This bug allowed the liblustre (e.g. catamount) client to
1367 return to the app before handling all startup RPCs. This
1368 could leave the node unresponsive to lustre network traffic
1369 and manifested as a server ptllnd timeout.
1371 Severity : enhancement
1373 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1374 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1375 boolean to control whether to print peer debug info when a
1376 client's RPC times out.
1379 Frequency : only for kernels with patches from Lustre below 1.4.3
1381 Description: Remove old rdonly API
1382 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1385 Frequency : only for devices with external journals
1387 Description: Set external device read-only also
1388 Details : During a commanded failover stop, we set the disk device
1389 read-only while the server shuts down. We now also set any
1390 external journal device read-only at the same time.
1393 Frequency : when upgrading from 1.4 while trying to change parameters
1395 Description: The wrong (new) MDC name was used when setting parameters for
1396 upgraded MDT's. Also allows changing of OSC (and MDC)
1397 parameters if --writeconf is specified at tunefs upgrade time.
1400 Frequency : when setting specific ost indicies
1402 Description: QOS code breaks on skipped indicies
1403 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1404 created with --index need not be sequential.
1406 Severity : enhancement
1408 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1409 Details : The uninit_groups feature works in conjunction with the kernel
1410 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1411 up the pass1 processing of e2fsck. This is a read-only feature
1412 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1413 mount filesystems that have had this feature enabled.
1415 Severity : enhancement
1417 Description: Improve multi-block allocation algorithm to avoid fragmentation
1418 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1419 allocation locality and avoid filesystem fragmentation.
1421 ------------------------------------------------------------------------------
1423 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1425 * Support for kernels:
1426 2.4.21-47.0.1.EL (RHEL 3)
1427 2.6.5-7.283 (SLES 9)
1428 2.6.9-42.0.10.EL (RHEL 4)
1429 2.6.12.6 vanilla (kernel.org)
1430 2.6.16.27-0.9 (SLES 10)
1431 * Recommended e2fsprogs version: 1.39.cfs5
1433 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1437 Frequency : occasional, depends on client load and configuration
1438 Bugzilla : 12181, 12203
1439 Description: data loss for recently-modified files
1441 Details : In some cases it is possible that recently written or created
1442 files may not be written to disk in a timely manner (this should
1443 normally be within 30s unless client IO load is very high).
1444 The problem appears as zero-length files or files that are a
1445 multiple of 1MB in size after a client crash or client eviction
1446 that are missing data at the end of the file.
1448 This problem is more likely to be hit on clients where files are
1449 repeatedly created and unlinked in the same directory, clients
1450 have a large amount of RAM, have many CPUs, the filesystem has
1451 many OSTs, the clients are rebooted frequently, and/or the files
1452 are not accessed by other nodes after being written.
1454 The presence of the problem can be detected by looking at
1455 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1456 smaller than the second (nr_unused) then dirty files will not
1457 be flushed automatically to disk. "sync; sleep 10" should be
1458 run several times on the node before unmounting it to update
1459 Lustre (this is also safe to run on nodes without this problem).
1461 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1462 that can cause this same problem, so customers using that kernel
1463 also need to update the kernel in addition to Lustre. In order
1464 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1466 It is normal that files written just before a client crash (less
1467 than 30s) may not yet have been flushed to disk, even for local
1471 Frequency : frequent on thin XT3 nodes
1473 Description: UUID collision on thin XT3 Linux nodes
1474 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1475 randomly, since we relied on an insufficiently-seeded PRNG.
1480 Description: OSS hangs after "All ost request buffers busy"
1481 Details : A deadlock between quota and journal operations caused OSS
1482 hangs after printing "All ost request buffers busy."
1485 Frequency : always on liblustre builds
1487 Description: Cleanup compiler warnings on liblustre
1490 Frequency : always on liblustre builds on XT3
1492 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1496 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1497 > 32000 subdirectories
1498 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1499 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1500 subdirectory count crosses 32000. This will aid e2fsck to
1501 correctly handle more than 32000 subdirectories.
1504 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1506 Description: System hang on startup
1507 Details : This bug allowed the liblustre (e.g. catamount) client to
1508 return to the app before handling all startup RPCs. This
1509 could leave the node unresponsive to lustre network traffic
1510 and manifested as a server ptllnd timeout.
1512 Severity : enhancement
1514 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1515 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1516 boolean to control whether to print peer debug info when a
1517 client's RPC times out.
1522 Description: make O_SYNC working on 2.6 kernels
1523 Details : 2.6 kernels use different method for mark pages for write,
1524 so need add a code to lustre for O_SYNC work.
1529 Description: Failure to close file and release space on NFS
1530 Details : Put inode details into lock acquired in ll_intent_file_open.
1531 Use mdc_intent_lock in ll_intent_open to properly
1532 detect all kind of errors unhandled by mdc_enqueue.
1537 Description: proc file read during shutdown sometimes raced obd removal,
1539 Details : Add lock to prevent obd access after proc file removal.
1542 Frequency : Only for files larger than 4GB on 32-bit clients.
1544 Description: improperly doing page alignment of locks
1545 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1546 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1551 Description: RPCs being resent when they shouldn't be
1552 Details : Some RPCs that should not be resent are being resent. This
1553 can cause inconsistencies in the RPC state machine. Do not
1554 resend such requests.
1557 Frequency : rare, only with NFS export
1559 Description: Crash on NFS re-export node
1560 Details : under very unusual load conditions an assertion is hit in
1561 ll_intent_file_open()
1564 Frequency : only if OST filesystem is corrupted
1566 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1567 Details : for a short time RPCs with bulk IO are in the replay list,
1568 but replay of bulk IOs is unimplemented. If the OST filesystem
1569 is corrupted due to disk cache incoherency and then replay is
1570 started it is possible to trip an assertion. Avoid putting
1571 committed RPCs into the replay list at all to avoid this issue.
1576 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1577 Details : Large single O_DIRECT read and write calls can fail to allocate
1578 a sufficiently large buffer to process the request. In case of
1579 allocation failure the allocation is retried with a smaller
1580 buffer and broken into smaller requests.
1582 Severity : enhancement
1584 Description: Add -o localflock option to simulate old noflock behaviour.
1585 Details : This will achieve local-only flock/fcntl locks coherentness.
1590 Description: versioning check is incomplete
1591 Details : Checking the version difference of client vs. server, report
1592 error if the gap is too big.
1597 Description: add support PG_writeback bit
1598 Details : add support for PG_writeback bit for Lustre, for more carefull
1599 work with page cache in 2.6 kernel. This also fix some deadlocks
1600 and remove hack for work O_SYNC with 2.6 kernel.
1602 Severity : enhancement
1604 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1605 Details : The uninit_groups feature works in conjunction with the kernel
1606 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1607 up the pass1 processing of e2fsck. This is a read-only feature
1608 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1609 mount filesystems that have had this feature enabled.
1611 Severity : enhancement
1613 Description: Improve multi-block allocation algorithm to avoid fragmentation
1614 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1615 allocation locality and avoid filesystem fragmentation.
1617 ------------------------------------------------------------------------------
1619 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1621 * Support for kernels:
1622 2.6.9-42.0.3.EL (RHEL 4)
1623 2.6.5-7.276 (SLES 9)
1624 2.4.21-47.0.1.EL (RHEL 3)
1625 2.6.12.6 vanilla (kernel.org)
1626 2.6.16.21-0.8 (SLES10)
1627 * Recommended e2fsprogs version: 1.39.cfs2-0
1629 * The backwards-compatible /proc/sys/portals symlink has been removed
1630 in this release. Before upgrading, please ensure that you change
1631 any configuration scripts or /etc/sysctl.conf files that access
1632 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1633 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1634 in advance of the upgrade on any system running Lustre 1.4.6 or
1635 newer, since /proc/sys/lnet was added in that version.
1636 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1640 Frequency : only when quota is used
1642 Description: avoid scanning export list for quota master
1643 Details : Change the algorithms to avoid scanning export list in order
1644 to improve the efficiency.
1647 Frequency : MDS failover only, very rarely
1649 Description: "went back in time" messages on mds failover
1650 Details : The greatest transno may be lost when the current operation
1651 finishes with an error (transno==0) and the client's last_rcvd
1652 record is over-written. Save the greatest transno in the
1653 mds_last_transno for this case.
1656 Frequency : always for specific kernels and striping counts
1658 Description: client may get "Matching packet too big" without ACL support
1659 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1660 when trying to access files in certain configurations. The
1661 clients should in fact be denied when mounting because they do
1662 not understand ACLs.
1665 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1667 Description: many clients connecting with IO in progress causes connect timeouts
1668 Details : Avoid synchronous journal commits to avoid delays caused by many
1669 clients connecting/disconnecting when bulk IO is in progress.
1670 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1671 OST_IO_PORTAL to avoid delays behind potentially many pending
1675 Frequency : occasionally with multiple writers to a single file
1677 Description: shared writes to file may result in wrong size reported by stat()
1678 Details : Allow growing of kms when extent lock is cancelled
1681 Frequency : always with random mmap IO to multi-striped file
1683 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1684 Details : Only if the hole is at the end of OST object so that kms is too
1685 small. Fix is to increase kms accordingly in ll_nopage.
1688 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1690 Description: writes to a missing object would leak memory on the OST
1691 Details : If there is an inconsistency between the MDS and OST filesystems,
1692 such that the MDS references an object that doesn't exist, writes
1693 to that object will leak memory due to incorrect cleanup in the
1694 error handling path, eventually running out of memory on the OST.
1699 Description: Creating too long symlink causes lustre errors
1700 Details : Check symlink and name lengths before sending requests to MDS.
1703 Frequency : only if flock is enabled (not on by default)
1705 Description: posix locks not released on fd closure on 2.6.9+
1706 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1707 kernels, this caused such locks not to be released on fd close and
1708 then assertions on fs unmount about still used locks.
1711 Frequency : MDS failover only, very rarely
1713 Description: clients may get ASSERTION(granted_lock != NULL)
1714 Details : When request was taking a long time, and a client was resending
1715 a getattr by name lock request. The were multiple lock requests
1716 with the same client lock handle and
1717 mds_getattr_name->fixup_handle_for_resent_request found one of the
1718 lock handles but later failed with ASSERTION(granted_lock != NULL).
1723 Description: handle->h_buffer_credits > 0, assertion failure
1724 Details : h_buffer_credits is zero after truncate, causing assertion
1725 failure. This patch extends the transaction or creates a new
1729 Frequency : NFS re-export or patchless client
1730 Bugzilla : 11179, 10796
1731 Description: Crash on NFS re-export node (__d_move)
1732 Details : We do not want to hash the dentry if we don't have a lock.
1733 But if this dentry is later used in d_move, we'd hit uninitialised
1734 list head d_hash, so we just do this to init d_hash field but
1735 leave dentry unhashed.
1738 Frequency : NFS re-export or patchless client
1740 Description: NFS exports has problem with symbolic link
1741 Details : lustre client didn't properly install dentry when re-exported
1742 to NFS or running patchless client.
1745 Frequency : NFS re-export or patchless client
1747 Description: Various nfs/patchless fixes.
1748 Details : fixes reuse disconected alias for lookup process - this fixes
1749 warning "find_exported_dentry: npd != pd",
1750 fix permission error with open files at nfs.
1751 fix apply umask when do revalidate.
1754 Frequency : occasional
1756 Description: Crash on NFS re-export node
1757 Details : calling clear_page() on the wrong pointer triggered oops in
1758 generic_mapping_read().
1761 Frequency : rarely, using O_DIRECT IO
1763 Description: unaligned directio crashes client with LASSERT
1764 Details : check for unaligned buffers before trying any requests.
1767 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1769 Description: stale data returned from RAID cache
1770 Details : If only a small amount of IO is done to the RAID device before
1771 reading it again it is possible to get stale data from the RAID
1772 cache instead of reading it from disk.
1775 Frequency : always for sles10 kernel
1777 Description: sles10 support
1778 Details : ll_follow_link: compile fixes and using of nd_set_link
1779 under newer kernels.
1782 Frequency : depends on arch, kernel and compiler version, always on sles10
1785 Description: recursive or deep enough symlinks cause stack overflow
1786 Details : getting rid of large stack-allocated variable in
1790 Frequency : depends on hardware
1792 Description: lustre write performance loss in the SLES10 kernel
1793 Details : the performance loss is caused by using of write barriers in the
1794 ext3 code. The SLES10 kernel turns barrier support on by
1795 default. The fix is to undo that change for ldiskfs.
1797 ------------------------------------------------------------------------------
1799 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1801 * Support for kernels:
1802 2.6.9-42.0.3EL (RHEL 4)
1803 2.6.5-7.276 (SLES 9)
1804 2.4.21-47.0.1.EL (RHEL 3)
1805 2.6.12.6 vanilla (kernel.org)
1809 Frequency : quota enabled and large files being deleted
1811 Description: releasing more than 4GB of quota at once hangs OST
1812 Details : If a user deletes more than 4GB of files on a single OST it
1813 will cause the OST to spin in an infinite loop. Release
1814 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1819 Description: statfs data retrieved from /proc may be stale or zero
1820 Details : When reading per-device statfs data from /proc, in the
1821 {kbytes,files}_{total,free,avail} files, it may appear
1822 as zero or be out of date.
1825 Frequency : systems with MD RAID1 external journal devices
1827 Description: lconf's call to blkid is confused by RAID1 journal devices
1828 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1829 of returning all block devices that match the journal UUID.
1832 Frequency : always, for aggregate stripe size over 4GB
1834 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1835 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1836 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
1839 Frequency : always if "lfs find" used on a local file/directory
1841 Description: "lfs find" segfaults if used on a local file/directory
1842 Details : The case where a directory component was not specified wasn't
1843 handled correctly. Handle this properly.
1846 Frequency : always on ppc64
1848 Description: the write to an ext3 filesystem mounted with mballoc got stuck
1849 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
1850 perform endianness conversion.
1853 Frequency : rarely (truncate to non-zero file size after write under load)
1854 Bugzilla : 10730, 10687
1855 Description: Files padded with zeros to next 4K multiple
1856 Details : With filesystems mounted using the "extents" option (2.6 kernels)
1857 it is possible that files that are truncated to a non-zero size
1858 immediately after being written are filled with zero bytes beyond
1859 the truncated size. No file data is lost.
1861 Severity : enhancement
1863 Description: Allow recovery/failover for liblustre clients.
1864 Details : liblustre clients were unaware of failover configurations until
1867 Severity : enhancement
1869 Description: user file locks should fail when not mounting with flock option
1870 Details : Set up an error-returning stub in ll_file_operations.lock field
1871 to prevent incorrect behaviour when client is mounted without
1872 flock option. Also, set up properly f_op->flock field for
1876 Frequency : always on ia64
1878 Description: "lfs df" loops on printing out MDS statfs information
1879 Details : The obd_ioctl_data was not initialized and in some systems
1880 this caused a failure during the ioctl that did not return
1881 an error. Initialize the struct and return an error on failure.
1884 Frequency : SLES 9 only
1886 Description: Error of copying files with lustre special EAs as root
1887 Details : Client side always return success for setxattr call for lustre
1888 special xattr (currently only "trusted.lov").
1891 Frequency : rarely on clusters with both ia64+i386 clients
1893 Description: ia64+i686 clients doing shared IO on the same file may LBUG
1894 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
1895 clients are doing concurrent writes to the same file it is
1896 possible that the ia64 clients may LASSERT because the OST
1897 extent locks are not PAGE_SIZE aligned. Ensure that grown
1898 locks are always aligned on the request boundary.
1901 Frequency : specific use, occasional
1903 Description: Overwriting in use executable truncates on-disk binary image
1904 Details : If one node attempts to overwrite an executable in use by
1905 another node, we now correctly return ETXTBSY instead of
1906 truncating the file.
1908 Severity : enhancement
1910 Description: Async OSC create to avoid the blocking unnecessarily.
1911 Details : If a OST has no remain object, system will block on the creating
1912 when need to create a new object on this OST. Now, ways use
1913 pre-created objects when available, instead of blocking on an
1914 empty osc while others are not empty. If we must block, we block
1915 for the shortest possible period of time.
1920 Description: chmod on Lustre root is propagated to other clients
1921 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
1922 invalid by the follow_mount time.
1925 Frequency : liblustre clients only
1927 Description: Race in 'instant cancel' lock handling could lead to such locks
1928 never to be granted in case of SMP MDS
1929 Details : Do not destroy not yet granted but cbpending locks in
1933 Frequency : replay/resend of open
1935 Description: non null lock assetion failure in mds_intent_policy
1936 Details : Trying to replay/resend lockless open requests resulted in
1937 mds_open() returning 0 with no lock. Now it sets a flag if
1938 a lock is going to be returned.
1940 Severity : enhancement
1942 Description: Checksum enhancements
1943 Details : New checksum enhancements allow for resending RPCs that failed
1946 Severity : enhancement
1948 Description: Tunables on number of dirty pages in cacche
1949 Details : Allow to set limit on number of dirty pages cached.
1954 Description: client crash on unmount - lock still has references
1955 Details : In some error handling cases it was possible to leak a lock
1956 reference on a client while accessing a file. This was not
1957 harmful to the client during operation, but would cause the
1958 client to crash when the filesystem is unmounted.
1961 Frequency : specific case, rare
1963 Description: ETXTBSY on mds though file not in use by client
1964 Details : ETXTBSY is no longer incorrectly returned when attempting to
1965 chmod or chown a directory that the user previously tried to
1966 execute or a currently-executing binary.
1969 Frequency : extremely rare except on liblustre-based clients
1971 Description: Lustre space not freed when files are deleted
1972 Details : Clean up open-unlinked files after client eviction. Previously
1973 the unlink was skipped and the files remained as orphans.
1978 Description: OST failure "would be an LBUG" in waiting_locks_callback()
1979 Details : In some cases it was possible to send a blocking callback to
1980 a client doing a glimpse, even though that client didn't get
1981 a lock granted. When the glimpse lock is cancelled on the OST
1982 the freed lock is left on the waiting list and corrupted the list.
1985 Frequency : all core dumps
1987 Description: Broke core dumps to lustre
1988 Details : Negative dentry may be unhashed if parent does not have UPDATE
1989 lock, but some callers, e.g. do_coredump, expect dentry to be
1990 hashed after successful create, hash it in ll_create_it.
1992 ------------------------------------------------------------------------------
1994 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
1996 * Support for kernels:
1997 2.6.9-42.0.2.EL (RHEL 4)
1998 2.6.5-7.276 (SLES 9)
1999 2.4.21-40.EL (RHEL 3)
2000 2.6.12.6 vanilla (kernel.org)
2004 Frequency : always on RHEL 3
2006 Description: Number of open files grows over time
2007 Details : The number of open files grows over time, whether or not
2008 Lustre is started. This was due to a filp leak introduced
2009 by one of our kernel patches.
2011 ------------------------------------------------------------------------------
2013 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2015 * Support for kernels:
2016 2.6.9-42.EL (RHEL 4)
2017 2.6.5-7.276 (SLES 9)
2018 2.4.21-40.EL (RHEL 3)
2019 2.6.12.6 vanilla (kernel.org)
2024 Bugzilla : 5719, 9635, 9792, 9684
2025 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2026 Details : If a server is under heavy load and cannot reply to new
2027 connection requests before the client resends the (re)connect,
2028 the connection handling code can behave badly if two service
2029 threads are concurrently handing separate (re)connections from
2030 the same client. Add better locking to the connection handling
2031 code, and ensure that only a single connection will be processed
2032 for a given client UUID, even if the lock is dropped.
2034 Severity : enhancement
2036 Description: add TCP zero-copy support to kernel
2037 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2038 sends if the hardware supports scatter-gather and checksumming.
2039 This allows socklnd to do client-write and server-read more
2040 efficiently and reduce CPU utilization from skbuf copying.
2043 Frequency : only if NFS exporting from client
2045 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2046 Details : If setting attributes on a file created under NFS that had
2047 never been opened it would be possible to oops the client
2048 if the file had no objects.
2051 Frequency : always for liblustre
2053 Description: liblustre client does MDS+OSTs setattr RPC for each write
2054 Details : When doing a write from a liblustre client, the client
2055 incorrectly issued an RPC to the MDS and each OST the file was
2056 striped over in order to update the timestamps. When writing
2057 with small chunks and many clients this could overwhelm the MDS
2058 with RPCs. In all cases it would slow down the write because
2059 these RPCs are unnecessary.
2061 Severity : enhancement
2063 Description: allow number of MDS service threads to be changed at module load
2064 Details : It is now possible to change the number of MDS service threads
2065 running. Adding "options mds mds_num_threads={N}" to the MDS's
2066 /etc/modprobe.conf will set the number of threads for the next
2067 time Lustre is restarted (assuming the "mds" module is also
2068 reloaded at that time). The default number of threads will
2069 stay the same, 32 for most systems.
2074 Description: OST crash if filesystem is unformatted or corrupt
2075 Details : If an OST is started on a device that has never been formatted
2076 or if the filesystem is corrupt and cannot even mount then the
2077 error handling cleanup routines would dereference a NULL pointer.
2082 Description: NULL pointer deref in llap_from_page.
2083 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2084 incorrect) mapping if the page was truncated/reclaimed while it was
2085 searched for. Check for this condition and skip such pages when
2086 doing readahead. Introduce extra check to llap_from_page() to
2087 verify page->mapping->host is non-NULL (so page is not anonymous).
2090 Frequency : Sometimes when using sys_sendfile
2092 Description: "page not covered by a lock" warnings from ll_readpage
2093 Details : sendfile called ll_readpage without right page locks present.
2094 Now we introduced ll_file_sendfile that does necessary locking
2095 around call to generic_file_sendfile() much like we do in
2099 Frequency : with certain MDS communication failures at client mount time
2101 Description: NULL pointer deref after failed client mount
2102 Details : a client connection request may delayed by the network layer
2103 and not be sent until after the PTLRPC layer has timed out the
2104 request. If the client fails the mount immediately it will try
2105 to clean up before the network times out the request. Add a
2106 reference from the request import to the obd device and delay
2107 the cleanup until the network drops the request.
2110 Frequency : occasionally during client (re)connect
2112 Description: assertion failure during client (re)connect
2113 Details : processing a client connection request may be delayed by the
2114 client or server longer than the client connect timeout. This
2115 causes the client to resend the connection request. If the
2116 original connection request is replied in this interval, the
2117 client may trip an assertion failure in ptlrpc_connect_interpret()
2118 which thought it would be the only running connect process.
2121 Frequency : only with obd_echo servers and clients that are rebooted
2123 Description: kernel BUG accessing uninitialized data structure
2124 Details : When running an obd_echo server it did not start the ping_evictor
2125 thread, and when a client was evicted an uninitialized data
2126 structure was accessed. Start the ping_evictor in the RPC
2127 service startup instead of the OBD startup.
2129 Severity : enhancement
2130 Bugzilla : 10193 (patchless)
2131 Description: Remove dependency on various unexported kernel interfaces.
2132 Details : No longer need reparent_to_init, exit_mm, exit_files,
2133 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2136 Frequency : rare (only users of deprecated and unsupported LDAP config)
2138 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2139 Details : LDAP apparently contains 'lustreName' attributes instead of
2140 'name'. A simple remapping of the name is sufficient.
2143 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2145 Description: waiting_locks_callback trips kernel BUG if client is evicted
2146 Details : Running with the dump_on_timeout debug flag turned on makes
2147 it possible that the waiting_locks_callback() can try to dump
2148 the Lustre kernel debug logs from an interrupt handler. Defer
2149 this log dumping to the expired_lock_main() thread.
2151 Severity : enhancement
2153 Description: Support NFS exporting on 2.6 kernels.
2154 Details : Implement non-rawops metadata methods for NFS server to use without
2155 changing NFS server code.
2158 Frequency : very rare (synthetic metadata workload only)
2160 Description: two racing renames might cause an MDS thread to deadlock
2161 Details : Running the "racer" program may cause one MDS thread to rename
2162 a file from being the source of a rename to being the target of
2163 a rename at exactly the same time that another thread is doing
2164 so, and the second thread has already enqueued these locks after
2165 doing a lookup of the target and is trying to relock them in
2166 order. Ensure that we don't try to re-lock the same resource.
2169 Frequency : only very large systems with liblustre clients
2171 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2172 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2173 all OSTs in parallel. This allows the eviction of stale liblustre
2174 clients to proceed much faster than if they were done in series,
2175 and also offers similar improvements for other set_info RPCs.
2180 Description: excessive CPU usage during initial read phase on client
2181 Details : During the initial read phase on a client, it would agressively
2182 retry readahead on the file, consuming too much CPU and impacting
2183 performance (since 1.4.5.8). Improve the readahead algorithm
2184 to avoid this, and also improve some other common cases (read
2185 of small files in particular, where "small" is files smaller than
2186 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2191 Description: MDS crash when receiving packet with unknown intent.
2192 Details : Do not LBUG in unknown intent case, just return -EFAULT
2194 Severity : enhancement
2195 Bugzilla : 9293, 9385
2196 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2197 Details : Do not serialize getattr (non-intent version) and statfs.
2200 Frequency : occasional, when OST network is overloaded/intermittent
2202 Description: client evicted by OST after bulk IO timeout
2203 Details : If a client sends a bulk IO request (read or write) the OST
2204 may evict the client if it is unresposive to its data GET/PUT
2205 request. This is incorrect if the network is overloaded (takes
2206 too long to transfer the RPC data) or dropped the OST GET/PUT
2207 request. There is no need to evict the client at all, since
2208 the pinger and/or lock callbacks will handle this, and the
2209 client can restart the bulk request.
2212 Frequency : Always when mmapping file with no objects
2214 Description: client crashes when mmapping file with no objects
2215 Details : Check that we actually have objects in a file before doing any
2216 operations on objects in ll_vm_open, ll_vm_close and
2222 Description: Request leak when working with deleted CWD
2223 Details : Introduce advanced request refcount tracking for requests
2224 referenced from lustre intent.
2226 Severity : Enhancement
2228 Description: Cache open file handles on client.
2229 Details : MDS now will return special lock along with openhandle, if
2230 requested and client is allowed to hold openhandle, even if unused,
2231 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2232 closing files for every read/write openration.
2234 Severity : Enhancement
2236 Description: Cache open negative dentries on client when possible.
2237 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2238 negative dentries on lock revocation.
2243 Description: Remounting a client read-only wasn't possible with a zconf mount
2244 Details : It wasn't possible to remount a client read-only with llmount.
2246 Severity : enhancement
2247 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2248 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2249 MPICH is included to add Lustre-specific ADIO interfaces.
2250 This is based closely on the UFS ADIO layer and only differs
2251 in file creation, in order to allow the OST striping to be set.
2252 This is user-contributed code and not supported by CFS.
2257 Description: extended inode attributes (immutable, append-only) work improperly
2258 when 2.4 and 2.6 kernels are used on client/server or vice versa
2259 Details : Introduce kernel-independent values for these flags.
2261 Severity : enhancement
2264 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2265 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2266 granularity. Now, Fractional MB granularity is be supported,
2267 this is very useful for low memory system.
2269 Severity : enhancement
2271 Description: Getattr by fid
2272 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2273 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2274 have locks and avoids getattr rpc requests for it on every stat.
2277 Frequency : Always, for filesystems larger than 2TB
2279 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2280 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2281 filesystems larger than 2TB to be mounted with mballoc enabled.
2284 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2286 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2287 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2288 filesystems larger than 2TB is mounted then the kernel will
2289 silently corrupt the start of the filesystem. CONFIG_LBD is
2290 enabled for all CFS-supported kernels, but the possibility of
2291 this happening with a modified kernel config exists.
2293 Severity : enhancement
2295 Description: add client O_DIRECT support for 2.6 kernels
2296 Details : It is now possible to do O_DIRECT reads and writes to files
2297 in the Lustre client mountpoint on 2.6 kernel clients.
2299 Severity : enhancement
2301 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2302 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2303 parallel, not waiting for response from every OST before sending
2304 a rpc to the next OST.
2309 Description: setattr vs write race when updating file timestamps
2310 Details : Client processes that update a file timestamp into the past
2311 right after writing to the file (e.g. tar) it is possible that
2312 the updated file modification time can be reset to the current
2313 time due to a race between processing the setattr and write RPC.
2315 Severity : enhancement
2317 Description: Bring 'lfs find' closer in line with regular Linux find.
2318 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2319 -print0 options and obtains all the needed info through the lustre
2322 Severity : enhancement
2324 Description: support up to 1024 configured devices on one node
2325 Details : change obd_dev array from statically allocated to dynamically
2326 allocated structs as they are first used to reduce memory usage
2331 Description: Flush dirty partially truncated pages during truncate
2332 Details : Immediatelly flush partially truncated pages in filter_setattr,
2333 this way we completely avoid having any pages in page cache on OST
2334 and can retire ugly workarounds during writes to flush such pages.
2339 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2340 Details : protect inode from truncation within vfs_unlink() context
2341 just take a reference before calling vfs_unlink() and release it
2342 when parent's i_sem is free.
2347 Description: last_id value checked outside lock on OST caused LASSERT failure
2348 Details : If there were multiple MDS->OST object precreate requests in
2349 flight, it was possible that the OST's last object id was checked
2350 outside a lock and incorrectly tripped an assertion. Move checks
2351 inside locks, and discard old precreate requests.
2354 Frequency : always, if extents are used on OSTs
2356 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2357 Details : OSTs using the extents format would not zero the high 16 bits of
2358 the index physical block number. This is not a problem for any
2359 OST filesystems smaller than 16TB, and no kernels support ext3
2360 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2361 new/modified files) and can be fixed for existing filesystems
2362 with e2fsprogs-1.39-cfs1.
2367 Description: import connection selection may be incorrect if timer wraps
2368 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2369 import connections to be ignored if the 32-bit jiffies counter
2370 wraps. Use a 64-bit jiffies counter.
2373 Frequency : very large clusters immediately after boot
2375 Description: LNET request buffers exhausted under heavy short-term load
2376 Details : If a large number of client requests are generated on a service
2377 that has previously never seen so many requests it is possible
2378 that the request buffer growth cannot keep up with the spike in
2379 demand. Instead of dropping incoming requests, they are held in
2380 the LND until the RPC service can accept more requests.
2383 Frequency : Sometimes during replay
2385 Description: Assertion failure in ll_local_open after replay.
2386 Details : If replay happened on an open request reply before we were able
2387 to set replay handler, reply will become not swabbed tripping the
2388 assertion in ll_local_open. Now we set the handler right after
2389 recognising of open request
2392 Frequency : very rare
2394 Description: kernel reports "badness in vsnprintf"
2395 Details : Reading from the "recovery_status" /proc file in small chunks
2396 may cause a negative length in lprocfs_obd_rd_recovery_status()
2397 call to vsnprintf() (which is otherwise harmless). Exit early
2398 if there is no more space in the output buffer.
2400 Severity : enhancement
2402 Description: clear OBD RPC statistics by writing to them
2403 Details : It is now possible to clear the OBD RPC statistics by writing
2404 to the "stats" file.
2409 Description: Client mtime is not the same on different clients after utimes
2410 Details : In some cases, the client was using the utimes() syscall on
2411 a file cached on another node. The clients now validate the
2412 ctime from the MDS + OSTs to determine which one is right.
2417 Description: Inability to activate failout mode
2418 Details : lconf script incorrectly assumed that in pythong string's numeric
2419 value is used in comparisons.
2422 Frequency : always with multiple stripes per file
2424 Description: Inefficient object allocation for mutli-stripe files
2425 Details : When selecting which OSTs to stripe files over, for files with
2426 a stripe count that divides evenly into the number of OSTs,
2427 the MDS is always picking the same starting OST for each file.
2428 Return the OST selection heuristic to the original design.
2433 Description: mount failures may take full timeout to return an error
2434 Details : Under some heavy load conditions it is possible that a
2435 failed mount can wait for the full obd_timeout interval,
2436 possibly several minutes, before reporting an error.
2437 Instead return an error as soon as the status is known.
2439 Frequency : quota enabled and large files being deleted
2441 Description: releasing more than 4GB of quota at once hangs OST
2442 Details : If a user deletes more than 4GB of files on a single OST it
2443 will cause the OST to spin in an infinite loop. Release
2444 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2449 Description: statfs data retrieved from /proc may be stale or zero
2450 Details : When reading per-device statfs data from /proc, in the
2451 {kbytes,files}_{total,free,avail} files, it may appear
2452 as zero or be out of date.
2455 Frequency : systems with MD RAID1 external journal devices
2457 Description: lconf's call to blkid is confused by RAID1 journal devices
2458 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2459 of returning all block devices that match the journal UUID.
2462 Frequency : always, for aggregate stripe size over 4GB
2464 Description: assertion fails when trying to use 4GB stripe size
2465 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2466 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2469 Frequency : always on ppc64
2471 Description: the first write on an ext3 filesystem with mballoc got stuck
2472 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2473 perform endianness conversion.
2475 ------------------------------------------------------------------------------
2477 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2479 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2480 INTEROPERATE with older versions automatically. Please read the
2481 user documentation before upgrading any part of a live system.
2482 * WARNING: Lustre networking configuration changes are required with
2483 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2486 * Support for kernels:
2487 2.6.9-22.0.2.EL (RHEL 4)
2488 2.6.5-7.244 (SLES 9)
2489 2.6.12.6 vanilla (kernel.org)
2492 Severity : enhancement
2493 Bugzilla : 7981/8208
2494 Description: Introduced Lustre Networking (LNET)
2495 Details : LNET is new networking infrastructure for Lustre, it includes
2496 a reorganized network configuration mode (see the user
2497 documentation for full details) as well as support for routing
2498 between different network fabrics. Lustre Networking Devices
2499 (LNDs) for the supported network fabrics have also been
2500 created for this new infrastructure.
2502 Severity : enhancement
2503 Description: Introduced Access control lists
2504 Details : clients can set ACLs on files and directories in order to have
2505 more fine-grained permissions than the standard Unix UGO+RWX.
2506 The MDS must be started with the "-o acl" mount option.
2508 Severity : enhancement
2509 Description: Introduced filesystem quotas
2510 Details : Administrators may now establish per-user quotas on the
2513 Severity : enhancement
2515 Description: Configuration change for the XT3
2516 The PTLLND is now used to run Lustre over Portals on the XT3
2517 The configure option(s) --with-cray-portals are no longer used.
2518 Rather --with-portals=<path-to-portals-includes> is used to
2519 enable building on the XT3. In addition to enable XT3 specific
2520 features the option --enable-cray-xt3 must be used.
2525 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2526 Details : A race between allocating a new llcd and re-getting the llcd_lock
2527 allowed another thread to grab newly-allocated llcd.
2529 Severity : enhancement
2531 Description: 2.6 OST async journal commit and locking fix to improve performance
2532 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2533 2.6 kernels are now async as they already were in 2.4 kernels so
2534 that they can commit concurrently with the network bulk transfer.
2535 For block-allocated files the filter allocation semaphore is held
2536 to avoid filesystem fragmentation during allocation. BKL lock
2537 removed for 2.6 xattr operations where it is no longer needed.
2542 Description: lconf incorrectly determined whether two IP networks could talk
2543 Details : In some more complicated routing and multiple-network
2544 configurations, lconf will avoid trying to make a network
2545 connection to a disjoint part of the IP space. It was doing the
2546 math incorrectly for one set of cases.
2551 Description: Fix for potential infinite loop processing records in an llog.
2552 Details : If an llog record is corrupted/zeroed, it is possible to loop
2553 forever in llog_process(). Validate the llog record length
2554 and skip the remainder of the block on error.
2557 Frequency : occasional (liblustre only)
2559 Description: liblustre could not open files whose last component is a symlink
2560 Details : sysio_path_walk() would incorrectly pass the open intent to
2561 intermediate path components.
2564 Frequency : rare (liblustre only with non-standard tuning)
2565 Bugzilla : 7201 (7350)
2566 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2567 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2568 after it has been released, only do this for referenced locks
2570 Severity : enhancement
2572 Description: specifying an (invalid) directory default stripe_size of -1
2573 would reset the directory default striping
2574 Details : stripe_size -1 was used internally to signal directory stripe
2575 removal, now use "all default" to signal dir stripe removal
2576 as a directory striping of "all default" is not useful
2579 Frequency : common for large clusters running liblustre clients
2581 Description: doing an ls when liblustre clients are running is slow
2582 Details : sending a glimpse AST to a liblustre client waits for every AST
2583 to time out, as liblustre clients will not respond. Since they
2584 cannot cache data we refresh the OST lock LVB from disk instead.
2586 Severity : enhancement
2588 Description: doing an ls at the same time as file IO can be slow
2589 Details : enqueue and other "small" requests can be blocked behind many
2590 large IO requests. Create a new OST IO portal for non-IO
2591 requests so they can be processed faster.
2594 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2596 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2597 created file with mode 000
2598 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2599 allow an ftruncate() as a non-root user to a file with mode 000.
2600 Lustre now respects this flag to disable mode checks when
2601 truncating a file owned by the user
2604 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2607 Description: Revoking locks from clients that went dead or catatonic might take
2609 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2610 cancellation of such locks instant on servers without waiting for
2611 any reply from clients. Clients drops these locks when cancel
2612 notification from server is received without replying.
2615 Frequency : liblustre-only, when liblustre client dies or becomes busy
2617 Description: Doing ls on Linux clients can take a long time with active
2619 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2620 granting such locks to it in the first place if possible. Locks
2621 are taken by proxy on the OST during the read or write and
2622 dropped immediately afterward. Add connect flags handling, do
2623 not grant locks to liblustre clients for glimpse ASTs.
2625 Severity : enhancement
2627 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2628 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2629 sequential reads and behaves badly with applications doing random
2630 reads. Improve it by reading ahead at least read region, and
2631 avoiding excessive large RPC for small reads.
2633 Severity : enhancement
2635 Description: Creating more than 1000 files for a single job may cause a load
2636 imbalance on the OSTs if there are also a large number of OSTs.
2637 Details : qos_prep_create() uses an OST index reseed value that is an
2638 even multiple of the number of available OSTs so that if the
2639 reseed happens in the middle of the object allocation it will
2640 still utilize the OSTs as uniformly as possible.
2645 Description: OST or MDS may oops in ping_evictor_main()
2646 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2647 but doesn't restart at beginning of obd_exports_timed list
2650 Severity : enhancement
2652 Description: improve by-nid export eviction on the MDS and OST
2653 Details : allow multiple exports with the same NID to be evicted at one
2654 time without re-searching the exports list.
2657 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2659 Description: MDS may oops in groups_free()
2660 Details : in rare race conditions a newly allocated group_info struct is
2661 freed again, and this can be NULL. The 2.4 compatibility code
2662 for groups_free() checked for a NULL pointer, but 2.6 did not.
2665 Frequency : common for liblustre clients doing little filesystem IO
2666 Bugzilla : 9352, 7313
2667 Description: server may evict liblustre clients accessing contended locks
2668 Details : if a client is granted a lock or receives a completion AST
2669 with a blocking AST already set it would not reply to the AST
2670 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2673 Frequency : lfs setstripe, only systems with more than 160 OSTs
2675 Description: unable to set striping with a starting offset beyond OST 160
2676 Details : llapi_create_file() incorrectly limited the starting stripe
2677 index to the maximum single-file stripe count.
2680 Frequency : LDAP users only
2682 Description: lconf did not handle in-kernel recovery with LDAP properly
2683 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2685 Severity : enhancement
2687 Description: bind OST threads to NUMA nodes to improve performance
2688 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2689 node and do their allocations there to localize memory access
2691 Severity : enhancement
2693 Description: llmount can determine client NID directly from Myrinet (GM)
2694 Details : the client NID code from gmnalnid was moved directly into
2695 llmount, removing the need to use this or specifying the
2696 client NID explicitly when mounting GM clients with zeroconf
2699 Frequency : if client is started with down MDS
2701 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2702 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2703 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2704 ptlrpc_import_delayed_req() because we didn't check if the
2705 request was interrupted, and we also didn't break out of the
2706 event loop if there was no timeout
2711 Description: data loss during non-page-aligned writes to a single file from
2712 both multiple nodes and multiple threads on one node at same time
2713 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2714 inconsistency led to false short-reads, that were cached and later
2715 used by ->prepare_write() to fill in partially written page,
2716 leading to data loss.
2719 Frequency : always, if lconf --abort_recovery used
2721 Description: lconf --abort_recovery fails with 'Operation not supported'
2722 Details : lconf was attempting to abort recovery on the MDT device and not
2725 Severity : enhancement
2727 Description: remove cleanup logs
2728 Details : replace lconf-generated cleanup logs with lustre internal
2729 cleanup routines. Eliminates the need for client-cleanup and
2732 Severity : enhancement
2734 Description: add support for EAs (user and system) on lustre filesystems
2735 Details : it is now possible to store extended attributes in the Lustre
2736 client filesystem, and with the user_xattr mount option it
2737 is possible to allow users to store EAs on their files also
2739 Severity : enhancement
2741 Description: Add possibility (config option) to show minimal available OST free
2743 Details : When compiled with --enable-mindf configure option, statfs(2)
2744 (and so, df) will return least minimal free space available from
2745 all OSTs as amount of free space on FS, instead of summary of
2746 free spaces of all OSTs.
2748 Severity : enhancement
2750 Description: do not expand extent locks acquired on OST-side
2751 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2752 by server: they are not cached anyway.
2755 Frequency : when mmap is used/binaries executed from Lustre
2757 Description: Unmmap pages before throwing them away from read cache.
2758 Details : llap_shrink cache now attempts to unmap pages before discarding
2759 them (if unmapping failed - do not discard). SLES9 kernel has
2760 extra checks that trigger if this unmapping is not done first.
2765 Description: lconf didn't resolve symlinks before checking to see whether a
2766 given mountpoint was already in use
2769 Frequency : when migrating failover services
2770 Bugzilla : 6395, 9514
2771 Description: When migrating a subset of services from a node (e.g. failback
2772 from a failover service node) the remaining services would
2773 time out and evict clients.
2774 Details : lconf --force (implied by --failover) sets the global obd_timeout
2775 to 5 seconds in order to quickly disconnect, but this caused
2776 other RPCs to time out too quickly. Do not change the global
2777 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2779 Severity : enhancement
2780 Frequency : if MDS is started with down OST
2781 Bugzilla : 9439,5706
2782 Description: Allow startup/shutdown of an MDS without depending on the
2783 availability of the OSTs.
2784 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2785 Add appropriate locking and lov-osc refcounts for safe
2786 cleaning. Add osc abort_inflight calls in case the
2787 synchronize never started.
2790 Frequency : occasional (Cray XT3 only)
2792 Description: root not authorized to access files in CRAY_PORTALS environment
2793 Details : The client process capabilities were not honoured on the MDS in
2794 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2795 been accessed by an authorized user then root was able to access
2796 the file on the local client also. The root user capabilities
2797 are now allowed on the MDS, as this environment has secure UID.
2800 Frequency : occasional
2802 Description: ldiskfs "too long searching" message happens too often
2803 Details : A debugging message (otherwise harmless) prints too often on
2804 the OST console. This has been reduced to only happen when
2805 there are fragmentation problems on the filesystem.
2810 Description: Division by zero in statfs when all OSCs are inactive
2811 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2812 lov_statfs divides by value returned by lov_get_stripecnt().
2816 Bugzilla : 9489, 3273
2817 Description: First write from each client to each OST was only 4kB in size,
2818 to initialize client writeback cache, which caused sub-optimal
2819 RPCs and poor layout on disk for the first writen file.
2820 Details : Clients now request an initial cache grant at (re)connect time
2821 and so that they can start streaming writes to the cache right
2822 away and always do full-sized RPCs if there is enough data.
2823 If the OST is rebooted the client also re-establishes its grant
2824 so that client cached writes will be honoured under the grant.
2829 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2830 Details : Now I/O RPCs go to different portal number and (presumably) fast
2831 lock requests (and glimses) and other RPCs get their own service
2832 threads pool that should be able to service those RPCs
2835 Severity : enhancement
2837 Description: Ability to exchange lustre version between client and servers and
2838 issue warnings at client side if client is too old. Also for
2839 liblustre clients there is ability to refuse connection of too old
2841 Details : New 'version' field is added to connect data structure that is
2842 filled with version info. That info is later checked by server and
2846 Frequency : rare, liblustre only.
2847 Bugzilla : 9296, 9581
2848 Description: Two simultaneous writes from liblustre at offset within same page
2849 might proceed at the same time overwriting eachother with stale
2851 Details : I/O lock withing llu_file_prwv was released too early, before data
2852 actually was hitting the wire. Extended lock-holding time until
2853 server acknowledges receiving data.
2856 Frequency : extremely rare. Never observed in practice.
2858 Description: avoid generating lustre_handle cookie of 0.
2859 Details : class_handle_hash() generates handle cookies by incrementing
2860 global counter, and can hit 0 occasionaly (this is unlikely, but
2861 not impossible, because initial value of cookie counter is
2862 selected randonly). Value of 0 is used as a sentinel meaning
2863 "unassigned handle" --- avoid it. Also coalesce two critical
2864 sections in this function into one.
2866 Severity : enhancement
2868 Description: allow liblustre clients to delegate truncate locking to OST
2869 Details : To avoid overhead of locking, liblustre client instructs OST to
2870 take extent lock in ost_punch() on client's behalf. New connection
2871 flag is added to handle backward compatibility.
2873 Severity : enhancement
2874 Bugzilla : 4928, 7341, 9758
2875 Description: allow number of OST service threads to be specified
2876 Details : a module parameter allows the number of OST service threads
2877 to be specified via "options ost ost_num_threads={N}" in the
2878 OSS's /etc/modules.conf or /etc/modprobe.conf.
2882 Bugzilla : 6146, 9635, 9895
2883 Description: servers crash with bad pointer in target_handle_connect()
2884 Details : In rare cases when a client is reconnecting it was possible that
2885 the connection request was the last reference for that export.
2886 We would temporarily drop the export reference and get a new
2887 one, but this may have been the last reference and the export
2888 was just destroyed. Get new reference before dropping old one.
2890 Severity : enhancement
2891 Frequency : if client is started with failover MDS
2893 Description: Allow multiple MDS hostnames in the mount command
2894 Details : Try to read the configuration from all specified MDS
2895 hostnames during a client mount in case the "primary"
2898 Severity : enhancement
2900 Description: Stop sending data to evicted clients as soon as possible.
2901 Details : Check if the client we are about to send or are sending data to
2902 was evicted already. (Check is done every second of waiting,
2903 for which l_wait_event interface was extended to allow checking
2904 of exit condition at specified intervals).
2907 Frequency : rare, normally only when NFS exporting is done from client
2909 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
2911 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
2914 Frequency : always, if srand() is called before liblustre initialization
2916 Description: Liblustre uses system PRNG disturbing its usage by user application
2917 Details : Introduce internal to lustre fast and high-quality PRNG for
2918 lustre usage and make liblustre and some other places in generic
2919 lustre code to use it.
2921 Severity : enhancement
2922 Bugzilla : 9477, 9557, 9870
2923 Description: Verify that the MDS configuration logs are updated when xml is
2924 Details : Check if the .xml configuration logs are newer than the config
2925 logs stored on the MDS and report an error if this is the case.
2926 Request --write-conf, or allow starting with --old_conf.
2928 Severity : enhancement
2930 Description: Handle symlinks in the path when checking if Lustre is mounted.
2931 Details : Resolve intermediate symlinks when checking if a client has
2932 mounted a filesystem to avoid duplicate client mounts.
2937 Description: lconf can hit an error exception but still return success.
2938 Details : The lconf command catches the Command error exception at the top
2939 level script context and will exit with the associated exit
2940 status, but doesn't ensure that this exit status is non-zero.
2945 Description: failure of ptlrpc thread startup can cause oops
2946 Details : Starting a ptlrpc service thread can fail if there are a large
2947 number of threads or the server memory is very fragmented.
2948 Handle this without oopsing.
2951 Frequency : always, only if liblustre and non-default acceptor port was used
2953 Description: liblustre cannot connect to servers with non-default acceptor port
2954 Details : tcpnal_set_default_params() was not called and was therefore
2955 ignoring the environment varaible TCPNAL_PORT, as well as other
2956 TCPNAL_ environment variables
2961 Description: two objects could be created on the same OST for a single file
2962 Details : If an OST is down, in some cases it was possible to create two
2963 objects on a single OST for a single file. No problems other
2964 than potential performance impact and spurious error messages.
2968 Bugzilla : 5681, 9562
2969 Description: Client may oops in ll_unhash_aliases
2970 Details : Client dcache may become inconsistent in race condition.
2971 In some cases "getcwd" can fail if the current directory is
2977 Description: Inode refcounting problems in NFS export code
2978 Details : link_raw functions used to call d_instantiate without obtaining
2979 extra inode reference first.
2983 Bugzilla : 9942, 9903
2984 Description: Referencing freed requests leading to crash, memleaks with NFS.
2985 Details : We used to require that call to ll_revalidate_it was always
2986 followed by ll_lookup_it. Also with revalidate_special() it is
2987 possible to call ll_revalidate_it() twice for the same dentry
2988 even if first occurence returned success. This fix changes semantic
2989 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
2990 reference on a request referred from the intent.
2991 ll_intent_release() then releases such a request.
2994 Frequency : rare, normally benchmark loads only
2996 Description: unlinked inodes were kept in memory on the client
2997 Details : If a client is repeatedly creating and unlinking files it
2998 can accumulate a lot of stale inodes in the inode slab cache.
2999 If there is no other client load running this can cause the
3000 client node to run out of memory. Instead flush old inodes
3001 from client cache that have the same inode number as a new inode.
3004 Frequency : SLES9 2.6.5 kernel and long filenames only
3005 Bugzilla : 9969, 10379
3006 Description: utime reports stale NFS file handle
3007 Details : SLES9 uses out-of-dentry names in some cases, which confused
3008 the lustre dentry revalidation. Change it to always use the
3012 Frequency : rare, unless heavy write-truncate concurrency is continuous
3013 Bugzilla : 4180, 6984, 7171, 9963, 9331
3014 Description: OST becomes very slow and/or deadlocked during object unlink
3015 Details : filter_destroy() was holding onto the parent directory lock
3016 while truncating+unlinking objects. For very large objects this
3017 may block other threads for a long time and slow overall OST
3018 responsiveness. It may also be possible to get a lock ordering
3019 deadlock in this case, or run out of journal credits because of
3020 the combined truncate+unlink. Solution is to do object truncate
3021 first in one transaction without parent lock, and then do the
3022 final unlink in a new transaction with the parent lock. This
3023 reduces the lock hold time dramatically.
3026 Frequency : rare, 2.4 kernels only
3028 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3029 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3030 before unmounting their filesystem in order to avoid deadlock.
3031 The kernel_locked() function in 2.4 kernels only checks whether
3032 the kernel lock is held, not whether it is this process that is
3033 holding it as 2.6 kernels do.
3038 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3039 Details : The client ptlrpc code may be trying to reconnect to a down
3040 server before a previous connection attempt has timed out.
3041 Increase the reconnect interval to be longer than the connection
3042 timeout interval to avoid sending duplicate connections to
3046 Frequency : echo_client brw_test command
3048 Description: fix echo_client to work with OST preallocated code
3049 Details : OST preallocation code (5137) didn't take echo_client IO path
3050 into account: echo_client calls filter methods outside of any
3051 OST thread and, hence, there is no per-thread preallocated
3052 pages and buffers to use. Solution: hijack pga pages for IO. As
3053 a byproduct, this avoids unnecessary data copying.
3057 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3058 Description: Client can oops in mdc_commit_close() after open replay
3059 Details : It was possible for the MDS to return an open request with no
3060 transaction number in mds_finish_transno() if the client was
3061 evicted, but without actually returning an error. Clients
3062 would later try to replay that open and may trip an assertion
3063 Simplify the client close codepath, and always return an error
3064 from the MDS in case the open is not successful.
3067 Frequency : rare, 2.6 OSTs only
3069 Description: OST may deadlock under high load on fragmented files
3070 Details : If there was a heavy load and highly-fragmented OST filesystems
3071 it was possible to have all the OST threads deadlock waiting on
3072 allocation of biovecs, because the biovecs were not released
3073 until the entire RPC IO was completed. Instead, release biovecs
3074 as soon as they are complete to ensure forward IO progress.
3076 Severity : enhancement
3078 Description: Support for specifying external journal device at mount
3079 Details : If an OST or MDS device is formatted with an external journal
3080 device, this device major/minor is stored in the ext3 superblock
3081 and may not be valid for failover. Allow detecting and
3082 specifying the external journal at mount time.
3087 Description: Mounting an MDS with pending unlinked files may cause oops
3088 Details : target_finish_recovery() calls mds_postrecov() which returned
3089 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3090 considers this an error and immediately begins cleaning up the
3091 lov, just after starting the mds_lov process
3093 Severity : enhancement
3095 Description: Implement 'lfs df' to report actual free space on per-OST basis
3096 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3097 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3098 usage of objects; '-h' to report in human readable format.
3100 ------------------------------------------------------------------------------
3102 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3109 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3110 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3111 at startup during orphan unlinking. Instead, skip buddy update
3112 before setup, as it will be regenerated after recovery is complete.
3117 Description: If an OST is inactive, its locks might reference stale inodes.
3118 Details : lov_change_cbdata() must iterate over all namespaces, even if
3119 they are inactive to clear inode references from the lock.
3121 Severity : enhancement
3122 Frequency : occasional, if non-standard max_dirty_mb used
3124 Description: Client will block write RPCs if not enough grant
3125 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3126 then the client will block writes while waiting for another RPC
3127 to complete instead of consuming its dirty limit. With change
3128 we get improved performance when max_dirty_mb is small.
3130 Severity : enhancement
3131 Bugzilla : 3389, 6253
3132 Description: Add support for supplementary groups on the MDS.
3133 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3134 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3135 lookups for user supplementary groups into a cache.
3139 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3140 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3141 for liblustre O_EXCL is enforced by MDS.
3145 Description: readdir never returns NULL in liblustre.
3146 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3147 dentry in struct dirent.
3151 Frequency : liblustre only, depends on application IO pattern
3152 Description: liblustre clients evicted if not contacting servers
3153 Details : Don't put liblustre clients into the ping_evictor list, so
3154 they will not be evicted by the pinger ever.
3156 Severity : enhancement
3158 Description: Add ability to evict clients by NID from MDS.
3159 Details : By echoing "nid:$NID" string into
3160 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3161 $NID would be instantly evicted from this MDS and from all active
3162 OSTs connected to it.
3166 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3167 Details : lookup_it_finish() used to query file size from OSTs that was not
3172 Description: service threads change working directory to that of init
3173 Details : Starting lustre service threads may pin the working directory
3174 of the parent thread, making that filesystem busy. Threads
3175 now change to the working directory of init to avoid this.
3179 Frequency : during shutdown only
3180 Description: shutdown with a failed MDS or OST can cause unmount to hang
3181 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3186 Frequency : 2.6 only
3187 Description: chmod/chown may include an extra supplementary group
3188 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3189 supplementary group and if none is specified this is used.
3192 Bugzilla : 5479 (6816)
3194 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3195 Details : If lookup is in progress on a file that is unlinked we might try
3196 to revalidate the inode and fail in revalidate after lookup is
3197 complete and ll_file_open() enqueues the open again but
3198 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3201 Frequency : always, if lconf --abort_recovery used
3203 Description: lconf --abort_recovery fails with 'Operation not supported'
3204 Details : lconf was attempting to abort recovery on the MDT device and not
3207 ------------------------------------------------------------------------------
3209 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3214 Frequency : rare (only unsupported configurations with a node running as an
3216 Bugzilla : 6514, 5137
3217 Description: Mounting a Lustre file system on a node running as an OST could
3219 Details : OSTs now preallocates memory needed to write out data at
3220 startup, instead of when needed, to avoid having to
3221 allocate memory in possibly low memory situations.
3222 Specifically, if the file system is mounted on on OST,
3223 memory pressure could force it to try to write out data,
3224 which it needed to allocate memory to do. Due to the low
3225 memory, it would be unable to do so and the node would
3226 become unresponsive.
3228 Severity : enhancement
3230 Description: Addition of lconf --service command line option
3231 Details : lconf now accepts a '--service <arg>' option, which is
3232 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3234 Severity : enhancement
3236 Description: Failover mode is now the default for OSTs.
3237 Details : By default, OSTs will now run in failover mode. To return to
3238 the old behaviour, add '--failout' to the lmc line for OSTs.
3240 Severity : enhancement
3242 Description: Health checks are now provided for MDS and OSTs
3243 Details : Additional detailed health check information on MSD and OSTs
3244 is now provided through the procfs health_check value.
3247 Frequency : occasional, depends on IO load
3249 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3250 after numerous create/delete cycles
3251 Details : The ext3 inode allocation policy would not allocate new inodes
3252 very well on the OSTs because there are no new directories
3253 being created. Instead we look for groups with free space if
3254 the parent directories are nearly full.
3259 Description: Network or server problems during mount may cause partially
3260 mounted clients instead of returning an error.
3261 Details : The config llog parsing code may overwrite the error return
3262 code during mount error handling, returning success instead
3268 Description: MDS can fail to allocate large reply buffers
3269 Details : After long uptimes the MDS can fail to allocate large reply
3270 buffers (e.g. zconf client mount config records) due to memory
3271 fragmentation or consumption by the buffer cache. Preallocate
3272 some large reply buffers so that these replies can be sent even
3273 under memory pressure.
3277 Frequency : rare (liblustre)
3278 Description: fsx running with liblustre complained that using truncate() to
3279 extend the file doesn't work. This patch corrects that issue.
3280 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3281 ATTR_SIZE and lsm use in llu_setattr_raw.
3285 Frequency : rare, only 2.6 kernels
3286 Description: Unusual file access patterns on the MDS may result in inode
3287 data being lost in very rare circumstances.
3288 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3289 code in the 2.6 kernel caused the inode and/or EA data not to
3290 be read from disk, causing single-file corruption.
3294 Frequency : rare, only 2.6 filesystems using extents
3295 Description: Heavy concurrent write and delete load may cause data corruption.
3296 Details : It was possible under high-load situations to have an extent
3297 metadata block in the block device cache from a just-unlinked
3298 file overwrite a newly-allocated data block. We now unmap any
3299 metadata buffers that alias just-allocated data blocks.
3303 Frequency : filesystems with default stripe_count larger than 77
3304 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3305 Details : lconf specifies an inode size of 4096 bytes when the default
3306 stripe_count is larger than 77. This conflicts with the default
3307 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3308 this case to avoid pinning too much memory for large EAs.
3310 ------------------------------------------------------------------------------
3312 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3317 Frequency : rare (extremely heavy IO load with hundreds of clients)
3319 Description: Client is evicted, gets IO error writing to file
3320 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3321 caused clients to be evicted for AST timeouts. The fixes in
3322 bug 5192 mean we no longer need to have such short AST timeouts
3323 so ldlm_timeout has been increased.
3326 Frequency : occasional during --force or --failover shutdown under load
3327 Bugzilla : 5949, 4834
3328 Description: Server oops/LBUG if stopped with --force or --failover under load
3329 Details : a collection of import/export refcount and cleanup ordering
3330 issues fixed for safer force cleanup
3333 Frequency : only filesystems larger than 120 OSTs
3334 Bugzilla : 5990, 6223
3335 Description: lfs getstripe would oops on a very large filesystem
3336 Details : lov_getconfig used kfree on vmalloc'd memory
3339 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3340 Bugzilla : 6242, 6243
3341 Description: reading from files that had been truncated to a non-zero size
3342 but never opened returned no data
3343 Details : ll_file_read() reads zeros from no-object files to EOF
3348 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3349 Details : There's little harm in aborting MDS/OSS recovery and letting it
3350 try again, so I removed the LASSERT and return an error instead.
3352 Severity : enhancement
3354 Description: New debugging infrastructure for tracking down data corruption
3355 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3356 (b) cover more of the client-side code path, and (c) try to narrow
3357 down where problems occurred
3361 Bugzilla : 3819, 4364, 4397, 6313
3362 Description: Racing close and eviction MDS could cause assertion in mds_close
3363 Details : It was possible to get multiple mfd references during close and
3364 client eviction, leading to one thread referencing a freed mfd.
3366 Severity: : enhancement
3367 Bugzilla : 3262, 6359
3368 Description: Attempts to reconnect to servers are now more aggressive.
3369 Details : This builds on the enhanced upcall-less recovery that was added
3370 in 1.4.2. When trying to reconnect to servers, clients will
3371 now try each server in the failover group every 10 seconds. By
3372 default, clients would previously try one server every 25 seconds.
3377 Description: After recovery, certain operations trigger a failed
3378 assertion on a client.
3379 Details : Failing over an mds, using lconf -d --failover, while a
3380 client was doing a readdir() call would cause the client to
3381 LBUG after recovery completed and the readdir() was resent.
3383 Severity : enhancement
3385 Description: Default groups are now added by lconf
3386 Details : You can now run lconf --group <servicename> without having to
3387 manually add groups with lmc.
3390 Frequency : occasional
3392 Description: Nodes with an elan id of 0 trigger a failed assertion
3395 Frequency : always when accessing e.g. tty/console device nodes
3397 Description: tty and some other devices nodes cannot be used on lustre
3398 Details : file's private_data field is used by device data and lustre
3399 values in there got lost. New field was added to struct file to
3400 store fs-specific private data.
3403 Frequency : when exporting Lustre via NFS
3405 Description: NFSD failed occasionally when looking up a path component
3406 Details : NFSD is looking up ".." which was broken in ext3 directories
3407 that had grown large enough to become hashed.
3410 Frequency : Clusters with multiple interfaces not on the same subnet
3412 Description: Nodes will repeatedly try to reconnect to an interface which it
3413 cannot reach and report an error to the log.
3414 Details : Extra peer list entries will be created by lconf with some peers
3415 unreachable. lconf now validates the peer before adding it.
3418 Frequency : Only if a default stripe is set on the filesystem root.
3420 Description: Setting a default stripe on the filesystem root prevented the
3421 filesystem from being remounted.
3422 Details : The client was sending extra request flags in the root getattr
3423 request and did not allocate a reply buffer for the dir EA.
3426 Frequency : occasional, higher if lots of files are accessed by one client
3427 Bugzilla : 6159, 6097
3428 Description: Client trips assertion regarding lsm mismatch/magic
3429 Details : While revalidating inodes the VFS looks up inodes with ifind()
3430 and in rare cases can find an inode that is being freed.
3431 The ll_test_inode() code will free the lsm during ifind()
3432 when it finds an existing inode and then the VFS later attaches
3433 this free lsm to a new inode.
3437 Bugzilla : 6422, 7030
3438 Description: MDS deadlock between mkdir and client eviction
3439 Details : Creating a new file via mkdir or mknod (starting a transaction
3440 and getting the ns lock) can deadlock with client eviction
3441 (gets ns lock and trying to finish a synchronous transaction).
3444 Frequency : occasional
3445 Description: While starting a server, the fsfilt_ext3 module could not be
3447 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3448 kernels. Previously, lconf would still use the ext3 name
3449 when trying to load modules. Now, it will correctly use
3450 ext3 on 2.4 and ldiskfs on 2.6.
3452 Severity : enhancement
3453 Description: The default stripe count has been changed to 1
3454 Details : The interpretation of the default stripe count (0, to lfs
3455 or lmc) has been changed to mean striping across a single
3456 OST, rather than all available. For general usage we have
3457 found a stripe count of 1 or 2 works best.
3459 Severity : enhancement
3460 Description: Add support for compiling against Cray portals.
3461 Details : Conditional compiling for some areas that are different
3465 Frequency : occasional
3466 Bugzilla : 6409, 6834
3467 Description: Creating files with an explicit stripe count may lead to
3468 a failed assertion on the MDS
3469 Details : If some OSTs are full or unavailable, creating files may
3470 trigger a failed assertion on the MDS. Now, Lustre will
3471 try to use other servers or return an error to the
3475 Frequency : occasional
3477 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3478 caused lock timeout during readahead (since 1.4.2).
3479 Details : Processes doing readahead might match a lock that hasn't been
3480 granted yet if there are overlapping and conflicting lock
3481 requests. The readahead process waits on ungranted lock
3482 (original lock is CBPENDING), while OST waits for that process
3483 to cancel CBPENDING read lock and eventually evicts client.
3485 Severity : enhancement
3487 Description: Initial enabling of flock support for clients
3488 Details : Implements fcntl advisory locking and file status functions.
3489 This feature is provided as an optional mount flag (default
3490 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3491 locking are implemented yet, and those that are are not guaranteed
3492 to be completely correct in production environments.
3493 mount -t lustre -o [flock|noflock] ...
3496 Frequency : occasional
3498 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3499 assertion in the ext3 JBD (journaling) layer.
3500 Details : The b_committed_data struct is protected by the big kernel lock
3501 in 2.4 kernels, serializing journal_commit_transaction() and
3502 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3503 access to this struct. In 2.6 kernels there is finer grained
3504 locking to improve SMP performance of the JBD layer.
3508 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3513 Description: merge and cleanup kernel patches.
3516 -----------------------------------------------------------------------------
3518 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3520 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3521 versions of Lustre. Please update all servers and clients to
3522 version 1.4.2 or later at the same time. You must also run
3523 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3524 to update the configuration logs.
3526 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3527 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3528 - swab llog records of type '0' so we get proper header size/idx (5861)
3529 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3530 - fix rename of one directory over another leaking an inode (5953)
3531 - avoid SetPageDirty on 2.6 (5981)
3532 - don't re-add just-being-destroyed locks to the waiting list (5653)
3533 - when creating new directories, inherit the parent's custom
3534 striping settings if present parent (3048)
3535 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3536 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3537 - don't hold client locks on temporary worklist from l_lru (5666)
3538 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3539 - automatically evict dead clients (5921)
3540 - Update file size properly in create+truncate+fstat case (6196)
3541 - Do not unhash mountpoint dentries, do not allow removal of
3543 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3544 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3545 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3546 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3547 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3548 - Correctly handle reads of files with no objects (6243)
3549 - lctl recover will also mark a device active if deactivate used (5933)
3551 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3552 - allow --write-conf on an MDS with different nettype than client (5619)
3553 - don't write config llogs to MDS for mounts not from that MDS (5617)
3554 - lconf should create multiple TCP connections from a client (5201)
3555 - init scripts are now turned off by default; run chkconfig --on
3556 lustre and chkconfig --on lustrefs to use them
3557 - upcalls are no longer needed for clients to recover to failover
3559 - add --abort-recovery option to lconf to abort recovery on device
3561 - add support for an arbitrary number of OSTs (3026)
3562 - Quota support protocol changes.
3563 - forward compatibility changes to wire structs (6007)
3564 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3565 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3566 - improved "lustre status" script
3567 - initialize blocksize for non-regular files (6062)
3568 - added --disable-server and --disable-client configure options (5782)
3569 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3570 - Vanilla 2.4.29 support
3571 - increase maximum number of obd devices to 520 (6242)
3572 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3573 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3574 SLES 9 2.6.5 kernels
3575 - limit stripes per file to 160 (the maximum EA size) (6093)
3577 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3580 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3581 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3582 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3583 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3584 - small ext3 extents cleanups and fixes (5733)
3585 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3586 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3587 - avoid needless client->OST connect, fix handle mismatch (5317)
3588 - fix DLM error path that led to out-of-sync client, long delays (5779)
3589 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3590 - fix several locking issues related to i_size (5492,5624,5654,5672)
3591 - don't move pending lock onto export if it is already evicted (5683)
3592 - fix kernel oops when creating .foo in unlinked directory (5548)
3593 - fix deadlock in obdfilter statistics vs. object create (5811)
3594 - use time_{before,after} to avoid timer jiffies wrap (5882)
3595 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3596 - Do not leak request if server was not able to process it (5154)
3597 - If mds_open unable to find parent dir, make that negative lookup(5154)
3598 - don't create new directories with extent-mapping (5909, 5936)
3600 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3601 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3602 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3603 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3604 - support for RHEL4 kernel on i686 (5773)
3605 - provide error messages when incompatible logs are encountered (5898)
3607 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3608 * version 1.4.0.10 (1.4.1 release candidate 1)
3610 - don't keep a lock reference when lock is not granted (4238)
3611 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3612 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3613 - drop import inflight refcount on signal_completed_replay error (5255)
3614 - unlock page after async write error during send (3677)
3615 - handle missing objects in filter_preprw_read properly (5265)
3616 - no transno return for symlink open, don't save no-trasno open (3440)
3617 - don't try to complete elan receive that already failed (4012)
3618 - free RPC server reply state on error (5406)
3619 - clean up thread from ptlrpc_start_thread() on error (5160)
3620 - readahead could read extra page into cache that wasn't ejected (5388)
3621 - prevent races in class_attach/setup/cleanup/detach (5260)
3622 - don't dereference de->d_inode after l_dput of de (5458)
3623 - use "int" for stripe value returned from lock_to_stripe (5544)
3624 - mballoc allocation and error-checking fixes in 2.6 (5504)
3625 - block device patches to fix I/O request sizes in 2.6 (5482)
3626 - look up hostnames for IB nals (5602)
3627 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3628 - don't start multiple acceptors for the same port (5277)
3629 - fix incorrect LASSERT in mds_getattr_name (5635)
3630 - export a proc file for general "ping" checking (5628)
3631 - fix "lfs check" to not block when the MDS is down (5628)
3633 - service request history (4965)
3634 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3635 - create an "evict_client" /proc entry on OSTs, like the MDS has
3636 - fix mount usage message, return errors per mount(8) (5168)
3637 - change grep [] to grep "[]" in tests so they work in more UMLs
3638 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3639 - remove ancient LOV_MAGIC_V0 EA support (5047)
3640 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3641 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3642 - allow readahead allocations to fail when low on memory (5383)
3643 - mmap locking landed again, after considerable improvement (2828)
3644 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3646 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3649 - send OST transaction number in read/write reply to free req (4966)
3650 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3651 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3652 - the watchdog thread now runs as interruptible (5246)
3653 - flock/lockf fixes (but it's still disabled, pending 5135)
3654 - don't use EXT3 constants in llite code (5094)
3655 - memory shortage at startup could cause assertion (5176)
3657 - reorganization of lov code
3658 - single portals codebase
3660 - add extents/mballoc support (5025)
3661 - direct I/O reads in the obdfilter (4048)
3662 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3664 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3667 - send OST transaction number in read/write reply to free req (4966)
3668 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3669 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3670 - the watchdog thread now runs as interruptible (5246)
3671 - handle missing objects in filter_preprw_read properly (5265)
3672 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3673 - drop import inflight refcount on signal_completed_replay error (5255)
3674 - unlock page after async write error during send (3677)
3675 - return original error code on reconstructed replies (3761)
3676 - no transno return for symlink open, don't save no-trasno open (3440)
3678 - add pid to ldlm debugging output (4922)
3679 - bump the watchdog timeouts -- we can't handle 30sec yet
3680 - extra debugging for orphan dentry/inode bug (5259)
3682 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3685 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3686 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3687 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3688 - fix NULL dereference in /proc/sys/portals/routes (4827)
3689 - allow failed mdc_close() operations to be interrupted (4561)
3690 - stop precreate on OST before MDS would time out on it (4778)
3691 - don't send partial-page writes before EOF from client (4410)
3692 - discard client grant for sub-page writes on large-page clients (4520)
3693 - don't free dentries not owned by NFS code, check generation (4806)
3694 - fix lsm leak if mds_create_objects() fails (4801)
3695 - limit debug_daemon file size, always print CERROR messages (4789)
3696 - use transno after validating reply (3892)
3697 - process timed out requests if import state changes (3754)
3698 - update mtime on OST during writes, return in glimpse (4829)
3699 - add mkfsoptions to LDAP (4679)
3700 - use ->max_readahead method instead of zapping global ra (5039)
3701 - don't interrupt __l_wait_event() during strace
3703 - add software watchdogs to catch hung threads quickly (4941)
3704 - make lustrefs init script start after nfs is mounted
3705 - fix CWARN/ERROR duplication (4930)
3706 - return async write errors to application if possible (2248)
3707 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3708 - print NAL number in %x format (4645)
3709 - update barely-supported suse-2.4.21-171 series (4842)
3710 - support for sles 9 %post scripts
3711 - support for building 2.6 kernel-source packages
3712 - support for sles km_* packages
3714 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3717 - ignore -ENOENT errors in osc_destroy (3639)
3718 - notify osc create thread that OSC is being cleaned up (4600)
3719 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3720 - reconstruct ost_handle() like mds_handle() (4657)
3721 - create a new thread to do import eviction to avoid deadlock (3969)
3722 - let lconf resolve symlinked-to devices (4629)
3723 - don't unlink "objects" from directory with default EA (4554)
3724 - hold socknal file ref over connect in case target is down (4394)
3725 - allow more than 32000 subdirectories in a single directory (3244)
3726 - fix blocks count for O_DIRECT writes (3751)
3727 - OST returns ENOSPC from object create when no space left (4539)
3728 - don't send truncate RPC if file size isn't changing (4410)
3729 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3730 - bind to privileged port in socknal and tcpnal (3689)
3732 - rate limit CERROR/CWARN console message to avoid overload (4519)
3733 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3734 - basic mmap support (3918)
3735 - kernel patch series update from b1_4 (4711)
3737 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3740 - avoid crash during MDS cleanup with OST shut down (2775)
3741 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3742 - don't use bad inodes on the MDS (3744)
3743 - dynamic object preallocation to improve recovery speed (4236)
3744 - don't hold spinlock over lock dumping or change debug flags (4401)
3745 - don't zero obd_dev when it is force cleaned (3651)
3746 - print grants to console if they go negative (4431)
3747 - "lctl deactivate" will stop automatic recovery attempts (3406)
3748 - look for existing locks in ldlm_handle_enqueue() (3764)
3749 - don't resolve lock handle twice in recovery avoiding race (4401)
3750 - revalidate should check working dir is a directory (4134)
3752 - don't always mark "slow" obdfilter messages as errors (4418)
3754 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3757 - don't close LustreDB during write_conf until it is done (3860)
3758 - fix typo in lconf for_each_profile (3821)
3759 - allow dumping logs from multiple threads at one time (3820)
3760 - don't allow multiple threads in OSC recovery (3812)
3761 - fix debug_size parameters (3864)
3762 - fix mds_postrecov to initialize import for llog ctxt (3121)
3763 - replace config semaphore with spinlock (3306)
3764 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3765 - don't allow enqueue to complete on a destroyed export (3822)
3766 - down write_lock before checking llog header bitmap (3825)
3767 - recover from lock replay timeout (3764)
3768 - up llog sem before sending rpc (3652)
3769 - reduce ns lock hold times when setting kms (3267)
3770 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3771 - fix NULL deref and obd_dev leak on setup error (3312)
3772 - replace some LBUG about llog ops with error handling (3841)
3773 - don't match INVALID dentries from d_lookup and spin (3784)
3774 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3775 - fix invalid assertion in ptlrpc_set_wait (3880)
3777 - add libwrap support for the TCP acceptor (3996)
3778 - add /proc/sys/portals/routes for non-root route listing (3994)
3779 - allow setting MDS UUID in .xml (2580)
3780 - print the stack of a process that LBUGs (4228)
3782 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3785 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3786 - make sure to unset replay flag from failed open requests (3440)
3787 - if default stripe count is 0, use OST count for inode size (3636)
3788 - update parent mtime/ctime on client for create/unlink (2611)
3789 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3790 - free recovery state on server during a forced cleanup (3571)
3791 - unregister_reply for resent reqs (3063)
3792 - loop back devices mounting and status check on 2.6 (3563)
3793 - fix resource-creation race that can provoke i_size == 0 (3513)
3794 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3795 - more debugging for page-accounting assertion (3746)
3796 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3797 - avoid deadlock after precreation failure (3758)
3798 - fix race and lock order deadlock in orphan handling (3450, 3750)
3799 - add validity checks when grabbing inodes from l_ast_data (3599)
3801 - add /proc/.../recovery_status to obdfilter (3428)
3802 - lightweight CDEBUG infrastructure, debug daemon (3668)
3803 - change default OSC RPC parameters to be better on small clusters
3804 - turn off OST read cache for files smaller than 32MB
3805 - install man pages and include them in rpms (3100)
3806 - add new init script for (un)mounting lustre filesystems (2593)
3807 - run chkconfig in %post for init scripts (3701)
3808 - drop scimac NAL (unmaintained)
3810 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3813 - clean kiobufs before and after use (3485)
3814 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3815 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3816 - change init script to fail more clearly if not run as root (1528)
3817 - allow clients to reconnect during replay (1742)
3818 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3819 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3820 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3821 - fix cond_resched (3554)
3822 - only evict unfinished clients after recovery (3515)
3823 - allow bulk resend, prevent data loss (3570)
3824 - dynamic ptlrpc request buffer allocation (2102)
3825 - don't allow unlinking open directory if it isn't empty (2904)
3826 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3827 - remove extraneous obd dereference causing LASSERT failure (3334)
3828 - don't use get_cycles() when creating temp. files on the mds (3156)
3829 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3830 - handle EEXIST for set-stripe, set proper directory name (3336)
3832 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3833 - fix ksocknal_fmb_callback() error messages (2918)
3835 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3838 - don't copy lvb into (possibly NULL) reply on error (2983)
3839 - don't deref dentry after dput, don't free lvb on error (2922)
3840 - use the kms to determine writeback rpc length (2947)
3841 - increment oti_logcookies when osc is inactive (2948)
3842 - update client's i_blocks count via lvb messages (2543)
3843 - handle intent open/close of special files properly (1557)
3844 - mount MDS with errors=remount-ro, like obdfilter (2009)
3845 - initialize lock handle to avoid ASSERT on error cleanup (3057)
3846 - don't use cancelling-locks' kms values (2947)
3847 - use highest lock extent for kms, not last one (2925)
3848 - don't dereference ERR_PTR() dentry in error handling path (3107)
3849 - fix thread race in portals_debug_dumplog() (3122)
3850 - create lprocfs device entries at setup instead of at attach (1519)
3851 - common AST error handler, don't evict client on completion race (3145)
3852 - zero nameidata in detach_mnt in 2.6 (3118)
3853 - verify d_inode after revalidate_special is valid in 2.6 (3116)
3854 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
3855 - initialize RPC timeout timer earlier for 2.6 (3219)
3856 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
3857 - print nal/nid for unknown nid (3258)
3858 - additional checks for oscc recovery before doing precreate (3284)
3859 - fix ll_extent_lock() error return code for 64-bit systems (3043)
3860 - don't crash in mdc_close for bad permissions on open (3285)
3861 - zero i_rdev for non-device files (3147)
3862 - clear page->private before handing to FS, better assertion (3119)
3863 - tune the read pipeline (3236)
3864 - fix incorrect decref of invalidated dentry (2350)
3865 - provide read-ahead stats and refine rpc in flight stats (3328)
3866 - don't hold journal transaction open across create RPC (3313)
3867 - update atime on MDS at close time (3265)
3868 - close LDAP connection when recovering to avoid server load (3315)
3869 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
3870 - don't leak open file on MDS after open resend (3325)
3871 - serialize filter_precreate and filter_destroy_precreated (3329)
3872 - loop device shouldn't call sync_dev() for nul device (3092)
3873 - clear page cache after eviction (2766)
3874 - resynchronize MDS->OST in background (2824)
3875 - refuse to mount the same filesystem twice on same mountpoint (3394)
3876 - allow llmount to create routes for mounting behind routers (3320)
3877 - push lock cancellation to blocking thread for glimpse ASTs (3409)
3878 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
3879 - fix rare problem with rename on htree directories (3417)
3881 - allow default OST striping configuration per directory (1414)
3882 - fix compilation for qswnal for 2.6 kernels (3125)
3883 - increase maximum number of MDS request buffers for large systems
3884 - change liblustreapi to be useful for external progs like lfsck (3098)
3885 - increase local configuration timeout for slow disks (3353)
3886 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
3888 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3891 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
3892 - don't overwrite extent policy data in reply if lock was blocked (2901)
3893 - drop filter export grants atomically with removal from device (2663)
3894 - del obd_self_export from work_list in class_disconnect_exports (2908)
3895 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
3896 - swab reply message in mdc_close, other PPC fixes (2464)
3897 - fix destroying of named logs (2325)
3898 - overwrite old logs when running lconf --write_conf (2264)
3899 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
3900 - fix race in target_handle_connect (2898)
3901 - mds_reint_create() should take same inode create lock (2926)
3902 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
3903 - don't close files for self_export to avoid uninitialized obd (2936)
3904 - allow MDS with the same name as client node (2939)
3905 - hold dentry reference for closed log files for unlink (2325)
3906 - reserve space for all logs during transactions (2059)
3907 - don't evict page beyond end of stripe extent (2925)
3908 - don't oops on a deleted current working directory (2399)
3909 - handle hard links to targets without a parent properly (2517)
3910 - don't dereference NULL lock when racing during eviction (2867)
3911 - don't grow lock extents when lots of conflicting locks (2919)
3913 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
3916 - account for cache space usage on clients to avoid data loss (974)
3917 - lfsck support in lustre kernel code (2349)
3918 - reduce journal credits needed for BRW writes (2370)
3919 - orphan handling to avoid losing space on client/server crashes
3920 - ptlrpcd can be blocked, stopping ALL progress (2477)
3921 - use lock value blocks to assist in proper KMS, faster stat (1021)
3922 - takes i_sem instead of DLM locks internally on obdfilter (2720)
3923 - recovery for initial connections (2355)
3924 - fixes for mds_cleanup_orphans (1934)
3925 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
3926 - block all file creations until orphan recovery completes (1901)
3927 - client remove rq_connection from request struct (2423)
3928 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
3929 - recovery timer race (2670)
3930 - mdc_close recovey bug (2532)
3931 - ptlrpc cleanup bug (2710)
3932 - mds timeout on local locks (2588)
3933 - namespace lock held during RPCs (2431)
3934 - handle interrupted sync write properly (2503)
3935 - don't try to handle a message that hasn't been replied to (2699)
3936 - client assert failure during cleanup after abort recovery (2701)
3937 - leak mdc device after failed mount (2712)
3938 - ptlrpc_check_set allows timedout requests to complete (2714)
3939 - wait for inflight reqs when ptlrpcd finishes (2710)
3940 - make sure unregistered services are removed from the srv_list
3941 - reset bulk XID's when resending them (caught by 1138 test)
3942 - unregister_bulk after timeout
3943 - fix lconf error (2694)
3944 - handle write after unfinished setstripe, stripe-only getstripe (2388)
3945 - readahead locks pages, leaves pending causing memory pressure (2673)
3946 - increase OST request buffers to 4096 on large machines (2729)
3947 - fix up permission of existing directories in simple_mkdir (2661)
3948 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
3949 - don't assign transno to errored transactions (2742)
3950 - don't delete objects on OST if given a bogus objid from MDS (2751)
3951 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
3952 - if rq_no_resend, then timeout request after recovery (2432)
3953 - fix MDS llog_logid record size, 64-bit array alignment (2733)
3954 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
3955 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
3956 - MDS recovery shouldn't delete objects that it has given out (2730)
3957 - if enqueue arrives after completion, don't clobber LVB (2819)
3958 - don't unlock pages twice when trigger_group_io returns error (2814)
3959 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
3960 - don't write pages to disk if there was an error (1450)
3961 - don't ping imports that have recovery disabled (2676)
3962 - take buffered bytes into account when balancing socknal conn (2817)
3963 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
3964 - reconnect unlink llog connection after MDS reconnects to OST (2816)
3965 - remove little-endian swabbing of llog records (1987)
3966 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
3967 - retry reposting request buffers if they fail (1191)
3968 - grow extent at grant time to avoid granting a revoked lock (2809)
3969 - lock revoke doesn't evict page if covered by a second lock (2765)
3970 - disable VM readahead to avoid reading outside lock extents (2805)
3972 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
3973 - updated kernel patches for hp-2.4.20 kernel (2681)
3975 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
3978 - fix truncated write corruption (2366)
3979 - fix for failed assertion in iopen_connect_dentry (1792,2517)
3981 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
3982 - ensure len(uuid) < 37 in lmc (1171)
3983 - fix ia64 OOPS in llog_test (2255)
3984 - zero end of page at obdfilter for partial page writes (2648)
3985 - don't leave stale dentries around after renames (bug 2428)
3986 - fix timeouts when evicting a client with a single lock held (2642)
3987 - set deadline for the initial HELLO message to drain (2634)
3988 - print out dotted-quad IP addresses in the socknal (2302)
3990 - additional debugging for MDS client eviction problem (2443)
3991 - fix mkfsoptions support for osts (2603, 2604)
3993 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
3996 - add series for the vanilla 2.6.0 kernel
3997 - add series for the vanilla 2.4.24 kernel
3998 - add series for a cray x86/64 UL kernel drop
3999 - fix xattr patches for the vanilla 2.4.19 series
4001 - generate true UUIDs in lmc (1171)
4002 - have portals stack dumping break in UML (2466)
4003 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4004 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4005 - wake callback waiting threads on client eviction (2460)
4006 - Add --ptldebug and --subsystem to lmc (1719)
4007 - update assertion to allow safe interrupt allocation
4008 - set rq_no_resend for cancel requests (2432)
4009 - recalculate ptlrpcd timeout after resend (2494)
4010 - call vfs_rmdir when removing pending directories (2368)
4011 - fix renaming a file to itself (2429)
4012 - lmc creates a default one-stripe lov (2454)
4013 - expand procfs space to handle large clusters (2326)
4014 - increase UML stack to avoid overflow
4015 - update lconf's list of debug and subsystem masks
4016 - fix lfs find --obd (2510)
4017 - /proc tunable for disabling filter read caching (2591)
4018 - stop rpm packages from altering slapd.conf (2301)
4019 - disable nagle in the socknal under 0conf (2578)
4020 - choose mds inode size based on stripe count (2572)
4021 - fix kernel-source rpm problems (2516)
4023 - add --disable-doc to avoid pdf generation (2421)
4024 - update documentation, tests, type-os, comments
4025 - avoid format warnings on ia64
4026 - remove the TOE NAL
4027 - tiny code cleanups by removing unused fields
4029 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4032 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4033 - remove the most blatant lies from BUILDING file (2371)
4034 - change default debug level to reasonable production setting
4035 - reduce client side cache size to reduce cache flush time
4036 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4037 - make TCP zerocopy and pinger support enabled by default (2476)
4038 - sync writes completed after process exits caused crashes (2319)
4039 - maintain correct mount count on the MDS (2356)
4040 - backout 1557, because 2316 wasn't really fixed
4041 - better file I/O statistics gathering in /proc
4042 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4043 - another kernel patch to fix zero-copy TCP function export
4044 - don't take duplicate lock when processing re-sent getattr (2420)
4045 - lctl uses obd_self_export instead of creating new conn (2353)
4046 - MDS/OST recovery case which requires object creation asserted (2425)
4047 - move lfs from /usr/sbin to /usr/bin in packages
4048 - fix race between mds_client_add and mds_client_free (2417)
4049 - use kmalloc instead of slabs in portals (2430)
4050 - don't create duplicate records when a failover MDS is present (2442)
4051 - remove unnecessary mount age check (2332)
4052 - don't remove directory inodes from locks prematurely (2451)
4053 - don't break if MDS service name is the same as hostname (2103)
4054 - fix races in client write RPC generation when cache full (2482)
4056 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4059 - remove now-unused request->rq_obd (278)
4060 - if an allocation fails, print out how much memory we've used (1933)
4061 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4062 - add the "configurable stack size" patch to most series files (1256)
4063 - ability to write large log records, for 100+ OST configs (2306)
4064 - fix NULL deref when filter_prep fails (2314)
4065 - fix operator precedence error in filter_sync
4066 - dynamic allocation of socknal TX descriptors (2315)
4067 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4068 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4069 - ensure that CWARN messages in Portals always get to the syslog
4070 - __init/__exit are not for prototype decls (ldlm_init/exit)
4071 - x86-64 compile warning fixes
4072 - fix gateway LMC keyword conflict (2318)
4073 - fix MDS lock inversions in getattr/reint paths (1844)
4074 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4075 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4076 - DLM race condition prevented some lock evictions (2328)
4077 - ENOMEM detection and retry on socknal sends (2230)
4078 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4079 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4080 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4081 - better error messages when a client is rejected during recovery (1505)
4082 - avoid cancelling locks which were never granted, after failure (2330)
4083 - fix i_sem/journal inversion in mds_client_add (2333)
4084 - fix truncate/getattr lock cycle deadlock (2334)
4085 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4086 - fix two rare exit paths which could leak an l_lock() ref (2321)
4087 - fencepost error in MDS/OST orphan recovery (2226)
4088 - make log record alignment 8 bytes (1988)
4089 - lstripe now fails when requested offset > ost_count (2237)
4090 - ensure that all kernel series have a complete list.h (1607)
4091 - fix crashes in special-file operations (2316)
4092 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4094 - allow configurable automake binary, for testing new versions
4095 - small update to the lfs documentation
4097 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4099 * fix negative export reference count in fsfilt_sync (2312)
4101 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4102 * release candidate 0.9.1
4104 - orphans are moved into the PENDING directory for possible recovery
4105 - replayed opens now open by fid for orphan/rename safety (1042)
4106 - last close of an orphan inode generates a transno (683)
4107 - chdir() and mount() now pin the directory entry (1020)
4108 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4109 - discard very old requests without processing them (1502)
4110 - remove some common, well-understood CERRORs (1505)
4111 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4112 - clear "grant" flags in OST replies until OST grant code lands (1644)
4113 - fix read performance by not clobbering i_blksize on client (1598)
4114 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4115 - make LRU size a /proc tunable, clears locks when reduced (707)
4116 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4117 - server locks take references on exports now (1558)
4118 - build fixes for 2.4.20-rh trees (1663)
4119 - return an error from lov_create if all OSCs are inactive (1751)
4120 - fix import levels when a reconnect happens without a timeout (1597)
4121 - exit early from mds_open if we get a lookup error (1749)
4122 - partial page read at EOF wouldn't wait for disk before sending (1642)
4123 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4124 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4125 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4126 - write/truncate lock inversion (1639)
4127 - Don't auto-load obdclass, portals modules during cleanup (1495)
4128 - fix timestamps from jumping to "now" (1763)
4129 - extra journal assertions (1648)
4130 - add an extra multiunlink test (1771)
4131 - fix read_record/write_record API (1776)
4132 - fix leak of offset_extent, possible incorrect i_size later (1772)
4133 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4134 - Debugging for the kqswnal_get_idle_tx problems (1820)
4135 - Allow recovery to be attempted multiple times (1536)
4136 - Write out MDS last_rcvd file after it is first created (1600)
4137 - Fix tx_descriptor leak in failed transmit situations (1827)
4138 - ext3 journaling fixes for assertion failure after IO error (1871)
4139 - class_export_put() on freed export after completion AST error (1896)
4140 - Fix revalidate looping in VFS (1322)
4141 - Don't access a freed export during MDS_REINT timeout (1521)
4142 - Add open-unlink recovery support on the MDS (1673,1764)
4143 - Return an error if no MDS data was read from last_rcvd (1946)
4144 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4145 - Don't setup a disk device that doesn't match exported UUID (317)
4146 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4147 - avoid committing NULL handle in force close
4148 - local.sh is now a one-stripe LOV configuration
4149 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4150 - fix ext3 htree duplicate directory entry corruption (1516)
4151 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4152 - update to new LOV EA format (2097)
4153 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4155 2003-06-15 Phil Schwan <phil@clusterfs.com>
4158 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4159 - per-import/export recovery handling (958, 931, 959)
4160 - multiple last-rcvd slots, for serving multiple FSes (949)
4161 - connections are again shared between multiple imp/exports (963, 964)
4162 - "umount -f" would hang if any requests needed to be sent (393, 978)
4163 - avoid pinning large req buffer by copying for queued messages (989)
4164 - add "uuid" to "lctl device" command to help upcalls (991)
4165 - "open" RPCs with transnos would confuse recovery counters (1037)
4166 - do proper endian conversion of all wire messages (288, 340, 891)
4167 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4168 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4169 - fix ptlrpc_connection leak in target_handle_connect (1174)
4170 - fix import refcounting bug in OST and MDS cleanup (1134)
4171 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4172 - fix very unlikely obd_types race condition (501)
4173 - remove osc_open hack for echo_client (1187)
4174 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4175 - a failure in read_inode2 leads to deadlock (1139)
4176 - cancel ack-locks as soon as transaction is committed (1072)
4177 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4178 - make sure to commitrw after any preprw to avoid deadlock (1162)
4179 - failing to execute a file in a lustre FS would lock inode (1203)
4180 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4181 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4182 - obd_uuid2tgt would walk off the end of the list (1255)
4183 - on IA64 the osc would give portals incorrect bulk size (1258)
4184 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4185 - fix lock inversion caused by new llite matching code (1282)
4186 - limit the number of dirty pages on a client to 10MB (1286)
4187 - timed out locks were not being corrected cancelled (1289)
4188 - fix O_DIRECT above 4GB on IA-32 (1292)
4189 * major user-visible changes
4190 - fail out/fail over policy now controlled by the upcall (993)
4192 - add OBD_PING to check server availability and failure (954)
4193 - lustre messages are now sent in sending host order (288, 340, 891)
4194 - add eadatalen to MDS getattr reply (340)
4195 - OST read replies may contain second buffer, with per-page status (593)
4197 2003-03-11 Phil Schwan <phil@clusterfs.com>
4200 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4201 - failed open()s could cause deadlock; fixed (867, 869)
4202 - stop cancelling OST locks when files are closed (481)
4203 - overlapping XID spaces caused network corruption (851, 853)
4204 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4205 - setattr_raw added, to do single-RPC, server-side setattrs
4206 - lmc/lconf syntax change for OST UUIDs
4207 - fix crashy race condition between ptlrpc_free_req and osc_close
4208 - don't use request in mdc_enqueue if we hit a timeout (889)
4209 - don't set the inode i_size for regular files from the MDS (896)
4210 - handle out of order completion AST (842)
4211 - don't LBUG if a lock request times out after receiving AST (913)
4212 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4213 - if a bad lock AST arrives, send an error instead of dropping entirely
4214 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4215 - fix leak in bulk IO when only partially completed (899, 900, 926)
4216 - fix O_DIRECT for ia64 (55)
4217 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4218 - C-z after timeout could hang a process forever; fixed (977)
4220 - client-side I/O cache (678, 924, 929, 941, 970)
4222 - READPAGE and SETATTRs which don't take server-side locks get
4225 2003-02-11 Phil Schwan <phil@clusterfs.com>
4228 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4229 - Fix fsfilt_extN_readpage() to read a full page of directory
4230 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4231 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4232 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4233 - Most intent operations were removed, in favour of a new RPC mode
4234 that does a single RPC to the server and bypasses most of the VFS
4235 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4236 - Aggressively cancel local locks on DLM servers
4237 - mds_reint_unlink sends EA to the client if it's the last nlink.
4238 client uses that EA to unlink OST objects.
4239 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4240 - recursive symlinks were fixed (439)
4241 - fixed NULL deref in DEBUG_REQ
4242 - filter_update_lastobjid no longer calls sync, which annoyed extN
4243 - fixed multi-client small-writes to a single file problem (445)
4244 - fixed mtime updates during file writes (607)
4245 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4246 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4247 - fixed Linux OST size reporting problem (444, 656)
4248 - OST now updates object mtime with writes or setattr (607, 619)
4249 - client verifies file size before zeroing page past EOF (445)
4250 - OST now writes last allocated objid to disk with allocation (108)
4251 - LOV on echo now works (409)
4253 - mds_reint_unlink sends a new buffer, with the EA included. this
4254 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4255 set if a regular file was being unlinked, and it was the last link
4256 - use PtlGet from the target for bulk writes (315)
4257 - OST now updates object mtime with writes or setattr (607, 619)
4258 - LDLM now has a grant-time callback to revalidate locked items, if
4260 - Many MDS operations were reorganized to combat race conditions
4262 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4263 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4265 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4268 - Fully reactivate OST imports after reconnection (512, others)
4269 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4270 - More graceful error handling for truncating on dead OST (515)
4271 - Don't error out unless we're actually accessing dead stripes (474)
4272 - Fix garbage sizes when stripes are missing (410)
4273 - LRU counters were broken, causing constant lock purge (433, 432)
4274 - garbage on read from stripes with failed OSTs (441)
4275 - mark OSCs as active before reconnecting during recovery (438)
4276 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4277 - lfind did not preserve OST order in output (443)
4278 - symlinks cause hung clients, incorrect data (439)
4279 - stop dereferencing request after dropping refcount (457)
4280 - don't LASSERT(spin_is_locked) on non-SMP (455)
4281 - fixes for many rename() bugs
4282 - fstat didn't correctly synchronize attributes (399)
4283 - server must handle lock cancellation during blocking AST prep (487)
4284 - bulk descriptors were free()d too soon (511)
4285 - fix paths in lconf, which would load incorrect modules (451, 507)
4286 - fix confusing lconf 'host not found' error message (386)
4287 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4288 - fix race condition in mdc_blocking_ast() for inode access (526)
4289 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4290 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4291 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4292 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4293 - fix lstripe in several ways (kernel side) (527)
4294 - fix request leak in ldlm_cli_enqueue (262)
4295 - incorrect OSC was marked inactive after OST failure
4296 - call mds_fs_cleanup before unmounting filesystem (524)
4297 - fix races between taking ns_lock and ldlm_lock_change_resource
4298 - fix races updating LOV export open file list
4299 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4300 - fix recovery NULL deref in ldlm_cli_cancel_unused
4301 - fix some DLM races by using new hash table for lock handles (419)
4302 - permit the client to specify desired inodes, at replay
4303 - duplicate requests when we queue them for replay reintegration
4304 - fix last_rcvd offset calculation
4305 - sync after each recovered transaction, so we always make progress
4306 - never, not always, ERESTART requests without transnos
4307 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4309 - skip replay if the MDS says that the client is already connected
4310 - don't check for a recovery-enabled export to match lctl's UUID
4311 - don't INC_USE_COUNT for phantom exports
4312 - don't crash when cleaning up phantom exports (567)
4313 - don't double-finish or set replay data for errored mdc_open requests
4314 - abort requests when they time out, so we don't get old replies
4315 - send/receive replies for AST messages again
4316 - if the client says that it doesn't have the lock, cancel it on the
4318 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4319 mark it as destroyed, it will all work out when decref is called
4320 - fix module use counts (22, 581)
4322 - ASTs now expect a reply (server cancels lock on error reply)
4324 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4327 - fix many simultaneous client startup (392)
4328 - fix dentry->d_it clobbering
4329 - credentials weren't being shipped for readdir/getattr operations
4330 - remove invalid assertions triggered during some concurrent MD
4332 - proper Lustre versions added (336, 389)
4333 - fix memory leak for create error case (398)
4334 - fix LOV locking bug that would get cli/srv out of sync
4335 - fix echo client over LOV (409)
4336 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4337 - fix double-O_EXCL intent crash (424)
4338 - avoid sending multiple lock CANCELs (352)
4340 - MDS can do multi-client recovery (modulo bugs in new code)
4342 - many updates, edits, cleanups
4344 2002-11-18 Phil Schwan <phil@clusterfs.com>
4347 - fix null d_it dereference (346)
4348 - fix full OST/dbench hang (333)
4349 - fix permission problem with file removal (286)
4350 - fix removal of OSCs from LOV when they fail
4351 - fix NULL deref during bulk timeout (214)
4352 - fix problems related to multiple filesystems on one MDS (241)
4353 - fixed serious subtle metadata locking bugs
4354 - free locks on clients when inodes are removed due to memory
4356 - fix inode pointer in lock data (285)
4357 - partial support for multiple MDS on a single host (241)
4358 - data locks weren't cancelled at clear_inode time (290, 311)
4359 - intent locks could lead to unbounded lock growth (205)
4360 - added a maximum lock count, an LRU list, and a flusher
4361 - fix multiple rename (365)
4362 - properly abstracted the echo client
4363 - OSC locked 1 byte too many; fixed
4364 - rewrote brw callback code:
4365 - fixed recovery bugs related to LOVs (306)
4366 - fixed too-many-pages-in-one-write crash (191)
4367 - fixed (again) crash in sync_io_timeout (214)
4368 - probably fixed callback-related race (385)
4370 - Add capability to MDS protocol
4371 - LDLM cancellations and callbacks on different portals
4373 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4376 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4377 - defer open object destruction to close time (601981, 138)
4378 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4379 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4380 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4381 - don't call obd_setattr in ll_file_release for destroyed objects
4382 * protocol change to lustre_msg: move |version| and add |flags|
4383 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4384 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4385 * added replay of create, unlink, link and rename operations during
4386 MDS failover; recovery should be much more robust now
4387 * remove failed OSCs from LOVs (only lov_create uses this so far)
4388 * the lustre-HOWTO was brought (more) up to date (582544)
4390 2002-10-23 Phil Schwan <phil@clusterfs.com>
4393 - in-use dentries weren't being reused properly (617851)
4394 - prevent multiple LDLM setup (599178)
4395 - fix LOV size calculations for truncate (617853)
4396 - fix client handling of MDS intent errors (POSIX)
4397 - fix permission bug in lovstripe.c test (624321)
4398 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4399 - truncate past end of file could corrupt data
4400 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4401 - a race in recovery could return ETIMEDOUT to apps (623947)
4402 - building outside the source directory was fixed
4403 * the lustre-HOWTO was brought (more) up to date (582544)
4404 * major progress was made on recovery functionality
4406 2002-10-10 Phil Schwan <phil@clusterfs.com>
4409 - recovery deadlock fix
4410 - rm -rf causes LBUG fix (617817)
4411 - file open by multiple tasks fix (618962)
4412 - directory permissions bugs (602707 and 620007)
4413 - journal_stop fixed with locking (611313)
4414 - O_APPEND failures resolved (618273, perhaps 614459)
4415 - lconf PATH fix (619770)
4416 - IA64 build fix (621450)
4417 - RPC buffer sizes scale with amount of memory
4419 2002-10-01 Phil Schwan <phil@clusterfs.com>
4422 - locks would be cancelled without throwing away data pages,
4423 resulting in inconsistent data (605627)
4424 - inode attributes were not always being refreshed (605627, 612449)
4425 - lconf now continues to cleanup after lctl reports an error
4426 - MDS now enforces user permissions (602707)
4427 - lprocfs cleanup fixed, but not yet enabled (614157)
4428 - fixed infinite server hang, should a client not respond to an AST
4429 - avoid going into recovery if user calls readlink() with a buffer
4430 that's too small (613941)
4431 - AST RPCs no longer require replies (614867) -- this may be changed
4432 - don't crash server if client sends an IOV that's too big (611336)
4433 - fixed lock conversion deadlock (611892)
4434 - fixed the following of symlinks (614622)
4435 * recovery: the server can remove locks from a client that dies, other
4436 clients can make progress
4437 * more extN patch fixes
4438 * compile-time configurable ptlrpc buffer allocations
4440 - collaborative read cache document
4441 - Lustre Lite Performance CDR document-in-progress
4443 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4446 - fix typo in patch-2.4.18
4448 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4451 - clear ptlrpc request each time in handle_incoming_request()
4452 - unlink of files now destroys the object on the OST
4454 2002-09-19 Peter Braam <braam@clusterfs.com>
4456 * add hard link support
4457 * change obdfile creation method
4458 * kernel patch changed
4460 2002-09-19 Peter Braam <braam@clusterfs.com>
4463 - stack overflow bug in extN fixed
4465 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4467 * documentation updates
4468 - add man pages for config tools
4469 - update tests/README to describe testing with new config tools
4470 - finish metadata API descriptions
4471 * bug fixes and cleanups
4472 - statfs workaround for 16TB limit
4473 - LOV stripe allocation improved, can stripe on subset of OSTs
4474 - LOV file size/IO offset was wrong for files > 4GB in size
4475 - object EA data was being dropped, caused files to be unreadable
4476 - memory overflow with non-LOV OST caused memory corruption
4477 - fixed regression tests to work with new config tools, obdfilter
4478 - fixed bug when directory size became larger than 1 block
4479 - fixed bug (for single client case) when PWD was deleted
4480 - invalidate local directory pages when doing intent-based ops
4481 - avoid LDLM oops when lock callback contained bad data
4483 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4485 * documentation updates
4486 * bug fixes and cleanups
4487 - configuration tools
4490 - 64-bit compile warnings
4491 - 64-bit internal statfs data
4493 * test_brw on persistent OST devices
4495 * lprocfs (disabled)
4497 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4499 * documentation updates
4500 * bug fixes and cleanups
4501 * configuration tools
4503 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4506 * many small fixes to 0_5_4
4507 * io/network handling
4508 * thinkos in MDS operations
4510 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4513 * crucial basic fixes to 0.5.3
4514 * IOR, Iozone work over Elan
4517 2002-08-07 Phil Schwan <phil@clusterfs.com>
4518 * version 0_5_3, our first alpha
4519 * we use the new Portals iovs
4520 * documentation updates
4521 * bug fixes and cleanups
4522 * small changes in the DLM wire protocol
4524 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4525 * version 0_5_1 with some initial stability,
4526 * locking on MD and file I/O.
4527 * documentation updates
4528 * several bug fixes since 0.5.0
4529 * small changes in wire protocol
4531 2002-07-18 Phil Schwan <phil@clusterfs.com>
4533 * delivered as Lustre Light Alpha
4534 * fixed a crash after handling invalid MDS requests
4535 * fixed directory pages for architectures with non-4k pages sizes
4537 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4538 * release version v0_4_4
4539 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4541 * Fixes a number of interruption problems with OST operations.
4542 * Update documentation for portals header changes
4543 * Move all wire protocol structs/defines to lustre_idl.h
4544 * Fixes symlink length bug.
4545 * Add tcpdump to repository.
4547 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4548 * release version v0_4_3
4549 * Fixes statfs for inodes on extN.
4550 * Fixes bug in runtests which would delete /etc/hosts.
4551 * Use 64-bit object IDs wherever possible (not into VFS though)
4552 Remove ost_get_info, which is unused by lustre, and out of date.
4554 2002-07-03 Peter Braam <braam@clusterfs.com>
4555 * release version v0_4_2 Fixes a lookup error (type not passed)
4556 * move forward to head of Portals
4557 * move forward to latest Lustre kernel
4559 2002-06-25 Peter Braam <braam@clusterfs.com>
4560 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob