1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: fix build on ppc32
18 Details : compile code with -m64 flag produce wrong object file for ppc32.
22 Description: add message levels for liblustreapi
26 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
27 Details : Move the 'good_osts' check before the 'total_bavail' check. This
28 will result in an -EAGAIN and in the exit call path we call
29 alloc_rr() which will with increasing aggressiveness attempt to
30 aquire precreated objects on the minimum number of required OSCs.
34 Description: Use old size assignment to avoid deadlock
35 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
36 the scheduling while holding a spinlock. We do not need locking
37 for size in ll_update_inode() because size is only updated from
38 the MDS for directories or files without objects, so there is no
39 other place to do the update, and concurrent access to such inodes
40 are protected by the inode lock.
44 Description: resolve "_IOWR redefined" build error on SLES10
48 Description: dump the memory debugging after all modules are unloaded to
49 suppress false negative in conf_sanity test 39
51 Severity : enhancement
53 Description: build kernel-ib packages for OFED 1.3 in our release cycle
58 Description: fix SLES kernel versioning
59 Details : the kernel version for our SLES 10 kernel did not include a "-"
60 before the "smp" at the end. while this was not a problem in
61 general, it did mean that software trying to use the kernel
62 version to try to detect a vendor specific kernel would fail.
63 this was most evident by the OFED build scripts.
67 Description: Don't update lov_desc members until making sure they are valid
68 Details : When updating lov_desc members via proc fs, need fix their
69 validities before doing the real update.
73 Description: don't put request into delay list while invalidate in flight.
74 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
75 import in flight. this produce timeout for invalidate and sometimes
80 Frequency : on ppc only
81 Description: not convert ost objects for directory because it's not exist.
82 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
86 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
88 Severity : enhancement
90 Description: Update to support 2.6.22.14 vanilla kernel.
94 Frequency : rare, on recovery
95 Description: read procfs can produce deadlock in some situation
96 Details : Holding lprocfs lock which send rpc can produce block for destroy
97 obd objects and this also block reconnect with -EALREADY. This isn't
98 fix all lprocfs bugs - but make it rare.
100 Severity : enhancement
102 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
105 Frequency : frequent on X2 node
107 Description: mdc_set_open_replay_data LBUG
108 Details : Set replay data for requests that are eligible for replay.
112 Description: lustre_mgs: operation 101 on unconnected MGS
113 Details : When MGC is disconnected from MGS long enough, MGS will evict the
114 MGC, and late on MGC cannot successfully connect to MGS and a lot
115 of the error messages complaining that MGS is not connected.
118 Frequency : on start mds
120 Description: Implement get_info(last_id) in obdfilter.
123 Frequency : occasional
125 Description: Correctly check stale fid, not start epoch if ost not support SOM
126 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
127 overwrited with new generated fid, but mds can anwer with one of these
128 two fids and both is not stale. setattr incorectly start epoch and
129 assume will be called done_writeting, but without SOM done_writing
133 Frequency : rare, depends on device drivers and load
135 Description: MDS or OSS nodes crash due to stack overflow
136 Details : Code changes in 1.8.0 increased the stack usage of some functions.
137 In some cases, in conjunction with device drivers that use a lot
138 of stack the MDS (or possibly OSS) service threads could overflow
139 the stack. One change which was identified to consume additional
140 stack has been reworked to avoid the extra stack usage.
143 Frequency : occasional
145 Description: Do not fail import if osc_interpret_create gets -EAGAIN
146 Details : If osc_interpret_create got -EAGAIN it immediately exits and
147 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
148 oscc_has_objects and see OSC has no objests and call
149 oscc_internal_create to resend create request.
151 Severity : enhancement
153 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
155 Severity : enhancement
157 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
160 Frequency : very rare
162 Description: Fix replay if there is an un-replied request and open
163 Details : In some cases, older replay request will revert the
164 mcd->mcd_last_xid on MDS which is used to record the client's
167 Severity : enhancement
169 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
171 Severity : enhancement
173 Description: Add rhel5 support to HEAD.
175 Serverity : enhancement
177 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
182 Description: Don't allow skipping OSTs if index has been specified.
183 Details : Don't allow skipping OSTs if index has been specified, make locking
184 in internal create lots better.
188 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
189 Details : access to bitfield in structure is always rounded to long
190 and this produce problem with not atomic change any bit.
194 Description: Lustre make rpms failed.
195 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
196 builds Lustre rpms from tarball.
199 Frequency : rare on shutdown ost
201 Description: If llog cancel was not send before clean_exports phase, this can
202 produce deadlock in llog code.
203 Details : If llog thread has last reference to obd and call class_import_put
204 this produce deadlock because llog_cleanup_commit_master wait when
205 last llog_commit_thread exited, but this never success because was
206 called from llog_commit_thread.
210 Description: allow userland application know is lost one of stripes.
211 Details : fill lvb_blocks with error code on ost and return it to
212 application if error flag found.
216 Description: NULL lov_tgts causing MDS oops
217 Details : more safe checks for NULL lov_tgts for avoid oops.
219 Severity : enhancement
221 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
225 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
226 Details : If an OST is down the MDS will hang indefinitely in
227 obd_statfs_rqset() waiting for the statfs data. While for
228 MDS QOS usage of statfs, it should not stuck in waiting.
230 Severity : enhancement
232 Description: remote_acl support
233 Details : Support ACL-based permission check for remote user.
234 Support setfacl/getfacl for remote user with the utils
235 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
236 the system "{s,g}etfacl" utils.
238 Severity : enhancement
240 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
242 Severity : enhancement
244 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
248 Description: make mgs_setparam() handle fsname containing dash
249 Details : fsname containing a dash does not work with lctl conf_param
251 Severity : enhancement
253 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
255 Severity : enhancement
257 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
261 Frequency : rare, at shutdown
262 Description: access already free / zero obd_namespace.
263 Details : if client_disconnect_export was called without force flag set,
264 and exist connect request in flight, this can produce access to
265 NULL pointer (or already free pointer) when connect_interpret
266 store ocd flags in obd_namespace.
270 Frequency : only at startup
271 Description: not alloc memory with spinlock held.
272 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
273 if any spinlock held.
275 Severity : enhancement
277 Description: make lustre randomly fail allocating memory
278 Details : Make lustre randomly failed allocating memory for testing purpose.
280 Severity : enhancement
282 Description: lost problems with lov objid file
283 Details : Fixes some scability and access to not inited memory problems
284 in work with lov objdid file.
289 Description: lfs find does not continue on file error
290 Details : Continue other files processing when a file/dir is absent.
294 Description: Inconsistent usage of lustre_pack_reply()
295 Details : Standardize the usage of lustre_pack_reply() such that it
296 always generate a CERROR on failure.
299 Frequency : occasional
301 Description: MDS hang or stay in waiting lock
302 Details : If client receive lock with CBPENDING flag ldlm need send lock
303 cancel as separate rpc, to avoid situation when cancel request
304 can't processed due all i/o threads stay in wait lock.
308 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
312 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
316 Description: Update to SLES9 kernel-2.6.5-7.287.3.
320 Description: Update to RHEL4 latest kernel.
322 Severity : enhancement
324 Description: Build SLES10 patchless client fails
325 Details : The configure was broken by run ./configure with
326 --with-linux-obj=.... argument for patchless client. When the
327 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
328 can't search header adequately. Use absolute path such as
329 -I($LINUX)/include instead.
333 Description: interrupt oig_wait produce painc on resend.
334 Details : brw_redo_request can be used for resend requests from ptlrpcd and
335 private set, and this produce situation when rq_ptlrpcd_data not
336 copyed to new allocated request and triggered LBUG on assert
337 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
338 ptlrpcd set if request is changed and can be safety changed to use
343 Description: LASSERT_{REQ,REP}SWAB macros are buggy
344 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
345 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
346 these macros and replace them with their difinitions instead.
350 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
351 Details : Update which_patch & target file for SLES10 latest kernel.
354 Description: Remove client patches from SLES 10 kernel.
355 Details : This causes SLES 10 clients to behave as patchless clients
356 even on a Lustre-patched (server) kernel.
358 Severity : enhancement
360 Description: self-adjustable client's lru lists
361 Details : use adaptive algorithm for managing client cached locks lru
362 lists according to current server load, other client's work
363 pattern, memory activities, etc. Both, server and client
364 side namespaces provide number of proc tunables for controlling
367 Severity : enhancement
369 Description: light-weight GSS support
370 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
371 and reduce performance overhead.
373 Severity : enhancement
375 Description: Linux keyring support
376 Details : Support using service of Linux keyring for Lustre GSS internal
377 context refresh/cache mechanism.
381 Description: Fix errors in lfs documentation
382 Details : Fixes man pages
386 Description: don't use GFP_* in generic Lustre code.
387 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
391 Description: obdclass is limited by single OBD_ALLOC(idarray)
392 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
396 Description: warnings with build patchless client with vanila 2.6.19 and up
397 Details : change old ctl_table style and replace ctl_table/ctl_table_header
398 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
402 Frequency : Only for SLES
403 Description: sanity_quota fail test_1
404 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
405 file, which makes TSTID[2] inunique.
409 Frequency : Only for SLES9
410 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
411 Details : Update target/ChangeLog/which_patch .
415 Frequency : Only for test_10 in sanity-quota.sh
416 Description: a bug in quota test code
417 Details : checking a test flag in a wrong place
421 Frequency : Only for test_13 in sanity-quota.sh
422 Description: a bug in quota test script
423 Details : shouldn't check overfull allocated quota
427 Description: jbd statistics
428 Details : Port older jbd statistics patch for sles10
430 Severity : enhancement
432 Description: Add printing inode info into message about error in writepage.
437 Description: reply_lock_interpret crash due to race with it and lock cancel.
438 Details : Do not replay locks that are being cancelled. Do not reference
439 locks by their address during replay, just by their handle.
443 Frequency : When flocks are used.
444 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
445 Details : Flock locks might destroy just granted lock if it could be merged
446 with another existing flock, this is done in completion handler,
447 so teach ldlm_cli_enquque_fini that this is a valid case for
453 Description: Oops in read and write path when failing to allocate lock.
454 Details : Check if lock allocation failed and return error back.
458 Description: lstripe command fails for valid OST index
459 Details : The stripe offset is compared to lov->desc.ld_tgt_count
460 instead of lov->desc.ld_active_tgt_count.
464 Description: sanity.sh failed test 103
465 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
470 Description: df doesn't work properly if diskfs blocksize != 4K
471 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
475 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
476 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
480 Description: obd_health_check_timeout too short
481 Details : set obd_health_check_timeout as 1.5x of obd_timeout
485 Description: llapi_file_create() does not allow some changes
486 Details : add llapi_file_open() that allows specifying the mode and
487 open flags, and also returns an open file handle.
491 Description: merge and cleanup kernel patches.
492 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
496 Description: Add journal checksum support.(Kernel part)
497 Details : The journal checksum feature adds two new flags i.e
498 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
499 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
500 indicates that the commit block contains the checksum for
501 the blocks described by the descriptor blocks. Now commit
502 record can be sent to disk without waiting for descriptor
503 blocks to be written to disk. This behavior is controlled
504 using JBD2_FEATURE_ASYNC_COMMIT flag.
508 Description: OSS needs mutliple precreate threads
509 Details : Add ability to start more than one create thread per OSS.
513 Description: Sanity.sh test_65k failed: setstripe should have succeeded
514 Details : execute lfs setstripe on client
518 Description: mds_obd_create error creating tmp object
519 Details : When the user sets quota on root, llog will be affected and can't
520 create files and write files.
523 Frequency : Always on ia64 patchless client, and possibly others.
525 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
526 Details : This allows the patchless client to be loaded on architectures
531 Description: RedHat Update kernel for RHEL5
532 Details : Add the kernel config file for RHEL5.
536 Description: RedHat Update kernel for RHEL5
537 Details : Modify the kernel config file more closer RHEL5.
541 Description: Build failure against Centos5 (RHEL5)
542 Details : Define PAGE_SIZE when it isn't present.
546 Description: test_7 on sanity-quota.sh can't be used on separate machines
547 Details : add facet to handle it
551 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
552 Details : with NFS, the anon dentry's parent was set to itself in
553 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
554 corresponding dentry other than use rec->ur_name.
556 Severity : enhancement
558 Description: lfs setstripe enhancement
559 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
563 Description: enable data checksumming by default
564 Details : enable checksum by default, allow --disable-checksum
565 configure option and "-o nochecksum" mount option. Checksums
566 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
570 Description: lustre support for RHEL5
571 Details : Add support for RHEL5.
575 Description: Client eviction due to failover config
576 Details : after a connection loss, the lustre client should attempt to
577 reconnect to the last active server first before trying the
578 other potential connections.
582 Description: when mds and osts use different quota unit(32bit and 64bit),
583 quota will be released repeatly.
584 Details : void sending multiple quota reqs to mds, which will keep the
585 status between the reqs.
589 Description: osts not allocated evenly to files
590 Details : change the condition to increase offset_idx
593 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
595 Description: Data corruption for OSTs that are formatted larger than 2TB
597 Details : When generating the bio request for lustre file writes the
598 sector number would overflow a temporary variable before being
599 used for the IO. The data reads correctly from Lustre (which
600 will overflow in a similar manner) but other file data or
601 filesystem metadata may be corrupted in some cases.
605 Description: Tune the kernel for good SCSI performance.
606 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
607 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
612 Description: rewrite ext2-derived code in obdclass/uuid.c
613 Details : rewrite inherited code (uuid parsing code from ext2 utils)
614 from scratch preserving functionality.
618 Description: "lfs find -obd UUID" prints directories
619 Details : "lfs find -obd UUID" will return all directory names instead
620 of just file names. It is incorrect because the directories
621 do not reside on the OSTs.
625 Description: use i_size_read and i_size_write in 2.6 port
626 Details : replace inode->i_size access with i_size_read/write()
629 Frequency : when removing large files
631 Description: scheduling issue during removal of large Lustre files
632 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
633 It causes scheduling issues when removing large files (17TB in the
637 Frequency : only with liblustre clients on XT3
639 Description: evictions taking too long
640 Details : allow llrd to evict clients directly on OSTs
643 Frequency : only on ppc
645 Description: /proc/fs/lustre/devices broken on ppc
646 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
647 We should make sure the type of 'index' is loff_t and then cast
648 explicitly as needed below. Do not assign an explicitly cast
652 Frequency : only for rhel5
654 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
655 Details : Modify the target file & which_kernel.
657 Severity : enhancement
659 Description: omit set fsid for export NFS
660 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
661 as FS_REQUIRES_DEV for avoid problems with generate fsid.
665 Frequency : Always, for kernels after 2.6.16
666 Description: Fix warning idr_remove called for id=.. which is not allocated.
667 Details : Last kernels save old s_dev before kill super and not allow
668 to restore from callback - restore it before call kill_anon_super.
672 Description: buffer overruns could theoretically occur
673 Details : llapi_semantic_traverse() modifies the "path" argument by
674 appending values to the end of the origin string, and a
675 overrun may occur. Adding buffer overrun check in liblustreapi.
679 Description: Fix error on 'ls .' at the top of the Lustre mount.
680 Details : Don't revalidate dentry if it is a root dentry.
684 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
685 Details : Modify vm-tunables-rhel4.patch.
689 Description: Kernel config for 2.6.18-vanilla.
690 Details : Modify targets/2.6-vanilla.target.in.
691 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
692 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
693 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
694 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
698 Description: improve handling recoverable errors
699 Details : if request processig with error which can be recoverable on server
700 request should be resend, otherwise page released from cache and
705 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
706 Details : Modify target file & which_patch.
707 A flaw was found in the IA32 system call emulation provided
708 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
709 value could be stored in the %RAX register, which could trigger an
710 out-of-bounds system call table access. An untrusted local user
711 could exploit this flaw to run code in the kernel
712 (ie a root privilege escalation). (CVE-2007-4573).
716 Description: O_DIRECT bypasses client statistics.
717 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
718 read_ahead_stats were updated. copied the stats section from
719 osc_send_oap_rpc() into async_internal().
723 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
727 Description: change order of libsysio includes
728 Details : '#include sysio.h' should always come before '#include xtio.h'
732 Description: handle "serious error: objid * already exists" more gracefully
733 Details : If LAST_ID value on disk is smaller than the objects existing in
734 the O/0/d* directories, it indicates disk corruption and causes an
735 LBUG(). If the object is 0-length, then we should use the existing
736 object. This will help to avoid a full fsck in most cases.
738 Severity : enhancement
740 Description: adapt the lustre_config script to support the upgrade case
741 Details : Add "-u" option for lustre_config script to support upgrading 1.4
742 server targets to 1.6 in parallel.
746 Description: To avoid grant space > avaible space when the disk is almost
747 full. Without this patch you might see the error "grant XXXX >
748 available" or some LBUG about grant, when the disk is almost
750 Details : In filter_check_grant, for non_grant cache write, we should
751 check the left space by if (*left > ungranted + bytes), instead
752 of (*left > ungranted), because only we are sure the left space
753 is enough for another "bytes", then the ungrant space should be
754 increase. In client, we should update cl_avail_grant only there
755 is OBD_MD_FLGRANT in the reply.
759 Description: Update RHEL 4 kernel to fix local root privilege escalation.
760 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
761 described in CVE-2007-4573. This problem could allow untrusted
762 local users to gain root access.
765 Frequency : when using O_DIRECT and quotas
767 Description: Incorrect file ownership on O_DIRECT output files
768 Details : block usage reported by 'lfs quota' does not take into account
769 files that have been written with O_DIRECT.
774 Description: touch file failed when fs is not full
775 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
776 otherwise we can get ENOSP while fs is not full.
780 Description: parallel lock callbacks
781 Details : Instead of sending blocking and completion callbacks as separated
782 requests, adding them to a set and sending in parallel.
785 Frequency : only for Cray XT3
786 Bugzilla : 12829/13455
787 Description: Changing primary group doesn't change the group lustre assigns to
789 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
790 overridden with the primary group provided by the group upcall,
791 whereas the supplied fsgid can be trusted if it is in the list of
792 supplementary groups returned by the group upcall.
794 Severity : enhancement
796 Description: Allow masking D_WARNING, D_ERROR messages from console
797 Details : Console messages can now be disabled via lnet.printk.
801 Description: User code with malformed file open parameter crashes client node
802 Details : Before packing join_file req, all the related reference should be
803 checked carefully in case some malformed flags cause fake
804 join_file req on client.
808 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
809 Details : ldlm_completion_ast() assumes that a lock is granted when the req
810 mode is equal to the granted mode. However, it should also check
811 that LDLM_FL_CP_REQD is not set.
815 Description: Heavy nfs access might result in deadlocks
816 Details : After ELC code landed, it is now improper to enqueue any mds
817 locks under och_sem, because enqueue might want to decide to
818 cancel open locks for same inode we are holding och_sem for.
822 Description: Client eviction while running blogbench
823 Details : A lot of unlink operations with concurrent I/O can lead to a
824 deadlock causing evictions. To address the problem, the number of
825 oustanding OST_DESTROY requests is now throttled to
826 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
827 callbacks are processed in priority.
831 Description: enable ACLs on MDS by default
832 Details : ACLs must be enabled on MDS by default.
835 Frequency : PPC/PPC64 only
837 Description: conflicts between asm-ppc64/types.h and lustre_types.h
838 Details : fix duplicated definitions between asm-ppc64/types.h and
839 lustre_types.h on PPC.
842 Frequency : PPC/PPC64 only
844 Description: asm-ppc/segment.h does not exist
845 Details : fix compile issue on PPC.
849 Description: better handle error messages in extents code
852 Frequency : RHEL4 only
854 Description: mkfs is very slow on IA64/RHEL4
855 Details : A performance regression has been discovered in the MPT Fusion
856 driver between versions 3.02.73rh and 3.02.99.00rh. As a
857 consequence, we have downgraded the MPT Fusion driver in the RHEL4
858 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
860 Severity : enhancement
862 Description: SNMP support enhancement
863 Details : Adding total number of sampled request for an MDS node in snmp
866 Severity : enhancement
868 Description: Optimize ldlm waiting list processing for PR extent locks
869 Details : When processing waiting list for read extent lock and meeting read
870 lock that is same or wider to it that is not contended, skip
871 processing rest of the list and immediatelly return current
872 status of conflictness, since we are guaranteed there are no
873 conflicting locks in the rest of the list.
877 Description: Time out and refuse to reconnect
878 Details : When the failover node is the primary node, it is possible
879 to have two identical connections in imp_conn_list. We must
880 compare not conn's pointers but NIDs, otherwise we can defeat
881 connection throttling.
885 Description: port llog fixes from b1_6 into HEAD
886 Details : Port llog reference couting and some llog cleanups from b1_6
887 (bug 10800) into HEAD, for protect from panic and access to already
888 free llog structures.
892 Description: Detect stride IO mode in read-ahead
893 Details : When a client does stride read, read-ahead should detect that and
894 read-ahead pages according to the detected stride pattern.
898 Description: data checksumming impacts single node performance
899 Details : add support for several checksum algorithm. Currently, only CRC32
900 and Adler-32 are supported. The checksum type can be changed on
901 the fly via /proc/fs/lustre/osc/*/checksum_type.
905 Description: use adler32 for page checksums
906 Details : when available, use the Adler-32 algorithm instead of CRC32 for
911 Description: build for x2 fails
912 Details : fix compile issue on Cray systems.
916 Description: Properly match for duplicate locks
917 Details : Due to different lock order from skiplists code, we need to
918 traverse entire list for now
921 Frequency : only on PPC/SLES10
923 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
924 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
925 be defined. Add a hack in mkfs_lustre.c to work around this compile
930 Description: LASSERT on MDS when client holding flock lock dies
931 Details : ldlm pool logic depends on number of granted locks equal to
932 number of released locks which is not true for flock locks, so
933 just exclude such locks from consideration.
937 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
938 Details : Use fsfilt_sync() for both the whole filesystem sync and
939 individual file sync to eliminate dangerous inode locking
940 with I_LOCK that can lead to a deadlock.
944 Description: performance in 1.6.3
945 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
946 sure that 1MB requests are not fragmented by the block layer.
948 Severity : enhancement
950 Description: organize the server-side client stats on per-nid basis
951 Details : Change the structure of stats under obdfilter and mds to
959 The "uuid"s file would list the uuids of _active_ exports.
960 And the clear entry is to clear all stats and stale nids.
962 --------------------------------------------------------------------------------
964 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
966 * Support for kernels:
967 2.6.5-7.283 (SLES 9),
968 2.6.9-55.EL (RHEL 4),
969 2.6.16.46-0.14 (SLES 10),
970 2.6.18.8 vanilla (kernel.org)
971 * Client support for unpatched kernels:
972 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
973 2.6.16 - 2.6.22 vanilla (kernel.org)
974 * Due to recently discovered recovery problems, we do not recommend
975 using patchless RHEL 4 clients with this or any earlier release.
976 * Recommended e2fsprogs version: 1.39.cfs8
977 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
978 * Starting with this release, the ldiskfs backing filesystem required
979 by Lustre is now in its own package, lustre-ldiskfs. This package
980 should be installed. It is versioned separately from Lustre and
981 may be released separately in future.
985 Description: block reactivating mgc import until all deactivates complete
986 Details : Fix race when failing back MDT/MGS to itself (testing)
989 Frequency : at statup only
991 Description: mds_lov_synchronize race leads to various problems
992 Details : simultaneous MDT->OST connections at startup can cause the
993 sync to abort, leaving the OSC in a bad state.
995 Severity : enhancement
997 Description: add optional extra BUILD_VERSION info
998 Details : add a new environment variable (namely LUSTRE_VERS) which allows
999 to override the lustre version.
1002 Frequency : 2.6.18 servers only
1004 Description: ll_kern_mount() doesn't release the module reference
1005 Details : The ldiskfs module reference count never drops down to 0
1006 because ll_kern_mount() doesn't release the module reference.
1011 Description: server LBUG when using old ost_num_threads parameter
1012 Details : Accept the old ost_num_threads parameter but warn that it
1013 is deprecated, and fix an off-by-one error that caused an LBUG.
1018 Description: Transient SCSI error results in persistent IO issue
1019 Details : iobuf->dr_error is not reinitialized to 0 between two
1023 Frequency : sometimes when underlying device returns I/O errors
1025 Description: OSTs not going read-only during write failures
1026 Details : OSTs are not remounted read-only when the journal commit threads
1027 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1028 instead of the ext3 wrappers.
1032 Description: poor connect scaling with increasing client count
1033 Details : Don't run filter_grant_sanity_check for more than 100 exports
1034 to improve scaling for large numbers of clients.
1037 Frequency : SLES10 only
1039 Description: sanity-quota.sh quotacheck failed: rc = -22
1040 Details : Quotas cannot be enabled on SLES10.
1043 Frequency : liblustre clients only
1045 Description: getdirentries does not give error when run on compute nodes
1046 Details : getdirentries does not fail when the size specified as an argument
1047 is too small to contain at least one entry
1049 Severity : enhancement
1051 Description: Add LNET router traceability for debug purposes
1052 Details : If a checksum failure occurs with a router as part of the
1053 IO path, the NID of the last router that forwarded the bulk data
1054 is printed so it can be identified.
1059 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1060 Details : Due to a race condition, liblustre clients were occasionally
1061 evicted incorrectly.
1063 Severity : enhancement
1065 Description: lfs setstripe use optional parameters instead of postional
1068 Severity : enhancement
1070 Description: Nanosecond timestamp support for ldiskfs
1071 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1072 resolution timestamps. There is not yet support for this at
1073 the Lustre filesystem level.
1076 Frequency : during server recovery
1078 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1079 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1083 Frequency : nfs export on patchless client
1085 Description: connectathon hang when test nfs export over patchless client
1086 Details : Disconnected dentry cannot be found with lookup, so we do not need
1087 to unhash it or make it invalid
1090 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1091 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1092 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1097 Description: open req refcounting wrong on reconnect
1098 Details : If reconnect happened between getting open reply from server and
1099 call to mdc_set_replay_data in ll_file_open, we will schedule
1100 replay for unreferenced request that we are about to free.
1101 Subsequent close will crash in variety of ways.
1102 Check that request is still eligible for replay in
1103 mdc_set_replay_data().
1108 Description: disable writes to filesystem when reading health_check file
1109 Details : the default for reading the health_check proc file has changed
1110 to NOT do a journal transaction and write to disk, because this
1111 can cause reads of the /proc file to hang and block HA state
1112 checking on a healthy but otherwise heavily loaded system. It
1113 is possible to return to the previous behaviour during configure
1114 with --enable-health-write.
1116 Severity : enhancement
1118 Description: 64-bit inode version
1119 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1120 to the inode. This will be required for version-based recovery.
1125 Description: MDS fails to start if a duplicate client export is detected
1126 Details : in some rare cases it was possible for a client to connect to
1127 an MDS multiple times. Upon recovery the MDS would detect this
1128 and fail during startup. Handle this more gracefully.
1130 Severity : enhancement
1132 Description: Add -o localflock option to simulate old noflock
1134 Details : This will achieve local-only flock/fcntl locks
1140 Description: log_commit_thread vs filter_destroy race leads to crash
1141 Details : Take import reference before releasing llog record semaphore
1146 Description: Wrong request locking in request set processing
1147 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1148 imp_delayed_list, in this place should be used imp_lock.
1151 Frequency : when reconnection
1153 Description: Grant Leak when osc reconnect to OST
1154 Details : When osc reconnect ost, OST(filter) should check whether it
1155 should grant more space to client by comparing fed_grant and
1156 cl_avail_grant, and return the granted space to client instead
1157 of "new granted" space, because client will call osc_init_grant
1158 to update the client grant space info.
1161 Frequency : when client reconnect to OST
1163 Description: Grant Leak when osc do resend and replay bulk write
1164 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1165 bulk write request, because the grant info while be sync between
1166 OSC and OST when reconnect, and we should ignore the grant info
1167 these of resend/replay write req.
1172 Description: Grant space more than avaiable left space sometimes.
1173 Details : When then OST is about to be full, if two bulk writing from
1174 different clients came to OST. Accord the avaliable space of the
1175 OST, the first req should be permitted, and the second one
1176 should be denied by ENOSPC. But if the seconde arrived before
1177 the first one is commited. The OST might wrongly permit second
1178 writing, which will cause grant space > avaiable space.
1181 Frequency : when client is evicted
1183 Description: Grant might be wrongly erased when osc is evicted by OST
1184 Details : when the import is evicted by server, it will fork another
1185 thread ptlrpc_invalidate_import_thread to invalidate the
1186 import, where the grant will be set to 0. While the original
1187 thread will update the grant it got when connecting. So if
1188 the former happened latter, the grant will be wrongly errased
1189 because of this race.
1194 Description: Checking Stale with correct fid
1195 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1196 to check whether it is stale, because sometimes, we want the
1197 enqueue happened anyway, and op_data.fid2 will not be initialized.
1199 Severity : enhancement
1201 Description: update patchless client
1202 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1205 Frequency : only with 2.4 kernel
1207 Description: random memory corruption
1208 Details : size of struct ll_inode_info is to big for union inode.u and this
1209 can be cause of random memory corruption.
1214 Description: Memory leak in recovery
1215 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1216 It should also check obd_fail before fsfilt_start, otherwise if
1217 fsfilt_start return -EROFS,(failover mds during mds recovery).
1218 then the req will return with repmsg->transno = 0 and rc = EROFS.
1219 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1220 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1221 be freed no matter whether fsfilt_commit success or not.
1224 Frequency : only with huge count clients
1226 Description: Prevents from taking the superblock lock in llap_from_page for
1228 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1229 for llap_from_page call in ll_removepage prevents from taking the
1230 superblock lock for a soon died page.
1235 Description: Not check open intent error before release open handle
1236 Details : in some rare cases, the open intent error is not checked before
1237 release open handle, which may cause
1238 ASSERTION(open_req->rq_transno != 0), because it tries to release
1239 the failed open handle.
1244 Description: Set cat log bitmap only after create log success.
1245 Details : in some rare cases, the cat log bitmap is set too early. and it
1246 should be set only after create log success.
1250 Description: Accessing a block bevice can re-enable I/O when Lustre is
1251 tearing down a device.
1252 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1257 Description: service threads may hog cpus when there are a lot of requests
1258 Details : Insert cond_resched to give other threads a chance to use some CPU
1263 Description: the cat log was not initialized in recovery
1264 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1265 unlink log on mds will not be initialized until mds post
1266 recovery. And also in mds post recovery, the unlink log will
1267 initialization will be done asynchronausly, so there will be race
1268 between add unlink log and unlink log initialization.
1272 Description: brw_stats were being printed incorrectly
1273 Details : brw_stats were being printed as log2 but all of them were not
1274 recorded as log2. Also remove some code duplication arising from
1275 filter_tally_{read,write}.
1279 Frequency : rare, only in recovery.
1280 Description: ASSERTION(req->rq_type != LI_POISON) failed
1281 Details : imp_lock should be held while iterating over imp_sending_list for
1282 prevent destroy request after get timeout in ptlrpc_queue_wait.
1286 Description: replay-single.sh test 52 fails
1287 Details : A lock's skiplist need to be cleanup when it being unlinked
1288 from its resource list.
1292 Description: Short directio read returns full requested size rather than
1294 Details : Direct I/O operations should return actual amount of bytes
1295 transferred rather than requested size.
1297 Severity : enhancement
1299 Description: metadata RPC reduction (e.g. for rm performance)
1300 Details : decrease the amount of synchronous RPC between clients and servers
1301 by canceling conflicing lock before the operation on the client
1302 and packing thier handles into the main operation RPC to server.
1304 Severity : enhancement
1306 Description: Async OSC create to avoid the blocking unnecessarily.
1307 Details : If a OST has no remain object, system will block on the creating
1308 when need to create a new object on this OST. Now, ways use
1309 pre-created objects when available, instead of blocking on an
1310 empty osc while others are not empty. If we must block, we block
1311 for the shortest possible period of time.
1313 Severity : enhancement
1315 Description: refine locking for avoid write wrong info into lov_objid file
1316 Details : fix possible races with add new target and write/update data in
1319 --------------------------------------------------------------------------------
1321 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1326 Frequency : on some architectures
1328 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1329 Details : Uninitialized flags sometimes cause configuration commands to
1333 Frequency : patchless clients only
1335 Description: missing __iget() symbol export
1336 Details : The __iget() symbol export is missing. To avoid the need for
1337 this on patchless clients the deathrow inode reaper is turned
1338 off, and we depend on the VM to clean up old inodes. This
1339 dependency was during via the fix for bug 12181.
1341 --------------------------------------------------------------------------------
1343 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1345 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1346 INTEROPERATE with older versions automatically. In many cases a
1347 special upgrade step is needed. Please read the
1348 user documentation before upgrading any part of a live system.
1349 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1350 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1351 * WARNING: Lustre configuration and startup changes are required with
1352 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1358 Severity : enhancement
1360 Description: MountConf
1361 Details : Lustre configuration is now managed via mkfs and mount
1362 commands instead of lmc and lconf. New obd types (MGS, MGC)
1363 are added for dynamic configuration management. See
1364 https://mail.clusterfs.com/wikis/lustre/MountConf for
1367 Severity : enhancement
1369 Description: dynamic OST addition
1370 Details : OSTs can now be added to a live filesystem
1372 Severity : enhancement
1374 Description: startup order invariance
1375 Details : MDTs and OSTs can be started in any order. Clients only
1376 require the MDT to complete startup.
1378 Severity : enhancement
1380 Description: parallel, asynchronous orphan cleanup
1381 Details : orphan cleanup is now performed in separate threads for each
1382 OST, allowing parallel non-blocking operation.
1384 Severity : enhancement
1386 Description: optimized stripe assignment
1387 Details : stripe assignments are now made based on ost space available,
1388 ost previous usage, and OSS previous usage, in order to try
1389 to optimize storage space and networking resources.
1391 Severity : enhancement
1393 Description: Permanently set tunables
1394 Details : All writable /proc/fs/lustre tunables can now be permanently
1395 set on a per-server basis, at mkfs time or on a live system.
1397 Severity : enhancement
1399 Description: Lustre message v2
1400 Details : Add lustre message format v2.
1402 Severity : enhancement
1404 Description: client OST exclusion list
1405 Details : Clients can be started with a list of OSTs that should be
1406 declared "inactive" for known non-responsive OSTs.
1410 Description: SPEC SFS validation failure on NFS v2 over lustre.
1411 Details : Changes the blocksize for regular files to be 2x RPC size,
1412 and not depend on stripe size.
1414 Severity : enhancement
1416 Description: Multiple MD RPCs in flight.
1417 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1418 To avoid overly-overloading MDS, introduce a limit on number of
1419 MDS RPCs in flight for a single client and add /proc controls
1420 to adjust this limit.
1422 Severity : enhancement
1424 Description: client read/write statistics
1425 Details : Add client read/write call usage stats for performance
1426 analysis of user processes.
1427 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1428 file access. extents_stats shows chunk size distribution.
1429 extents_stats_per_process show chunk size distribution per
1432 Severity : enhancement
1434 Description: mds statistics
1435 Details : Add detailed mds operations statistics in
1436 /proc/fs/lustre/mds/*/stats.
1440 Description: Failure of copying files with lustre special EAs.
1441 Details : Client side always return success for setxattr call for lustre
1442 special xattr (currently only "trusted.lov").
1447 Description: Refcount LNET uuids
1448 Details : The global LNET uuid list grew linearly with every startup;
1449 refcount repeated list entries instead of always adding to
1452 Severity : enhancement
1454 Description: Dynamic service threads
1455 Details : Within a small range, start extra service threads
1456 automatically when the request queue builds up.
1459 Frequency : mixed-endian client/server environments
1461 Description: mixed-endian crashes
1462 Details : The new msg_v2 system had some failures in mixed-endian
1465 Severity : enhancement
1467 Description: Easy OST removal
1468 Details : OSTs can be permanently deactivated with e.g. 'lctl
1469 conf_param lustre-OST0001.osc.active=0'
1471 Severity : enhancement
1473 Description: MGS proc entries
1474 Details : Added basic proc entries for the MGS showing what filesystems
1477 Severity : enhancement
1479 Description: provide MGS failover
1480 Details : Added config lock reacquisition after MGS server failover.
1482 Severity : enhancement
1484 Description: add Linux 2.4 support
1485 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1489 Description: a large application tries to do I/O to the same resource and dies
1490 in the middle of it.
1491 Details : Check the req->rq_arrival time after the call to
1492 ost_brw_lock_get(), but before we do anything about
1493 processing it & sending the BULK transfer request. This
1494 should help move old stale pending locks off the queue as
1495 quickly as obd_timeout.
1498 Frequency : when an incorrect nid is specified during startup
1500 Description: ptlrpc connect to non-existant node causes kernel crash
1501 Details : LNET can't be re-entered from an event callback, which
1502 happened when we expire a message after the export has been
1503 cleaned up. Instead, hand the zombie cleanup off to another
1506 Severity : enhancement
1508 Description: plain/inodebits lock performance improvement
1509 Details : Grouping plain/inodebits in granted list by their request modes
1510 and bits policy, thus improving the performance of search through
1514 Frequency : only if OST filesystem is corrupted
1516 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1517 Details : for a short time RPCs with bulk IO are in the replay list,
1518 but replay of bulk IOs is unimplemented. If the OST filesystem
1519 is corrupted due to disk cache incoherency and then replay is
1520 started it is possible to trip an assertion. Avoid putting
1521 committed RPCs into the replay list at all to avoid this issue.
1524 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1526 Description: System hang on startup
1527 Details : This bug allowed the liblustre (e.g. catamount) client to
1528 return to the app before handling all startup RPCs. This
1529 could leave the node unresponsive to lustre network traffic
1530 and manifested as a server ptllnd timeout.
1532 Severity : enhancement
1534 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1535 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1536 boolean to control whether to print peer debug info when a
1537 client's RPC times out.
1540 Frequency : only for kernels with patches from Lustre below 1.4.3
1542 Description: Remove old rdonly API
1543 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1546 Frequency : only for devices with external journals
1548 Description: Set external device read-only also
1549 Details : During a commanded failover stop, we set the disk device
1550 read-only while the server shuts down. We now also set any
1551 external journal device read-only at the same time.
1554 Frequency : when upgrading from 1.4 while trying to change parameters
1556 Description: The wrong (new) MDC name was used when setting parameters for
1557 upgraded MDT's. Also allows changing of OSC (and MDC)
1558 parameters if --writeconf is specified at tunefs upgrade time.
1561 Frequency : when setting specific ost indicies
1563 Description: QOS code breaks on skipped indicies
1564 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1565 created with --index need not be sequential.
1567 Severity : enhancement
1569 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1570 Details : The uninit_groups feature works in conjunction with the kernel
1571 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1572 up the pass1 processing of e2fsck. This is a read-only feature
1573 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1574 mount filesystems that have had this feature enabled.
1576 Severity : enhancement
1578 Description: Improve multi-block allocation algorithm to avoid fragmentation
1579 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1580 allocation locality and avoid filesystem fragmentation.
1582 ------------------------------------------------------------------------------
1584 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1586 * Support for kernels:
1587 2.4.21-47.0.1.EL (RHEL 3)
1588 2.6.5-7.283 (SLES 9)
1589 2.6.9-42.0.10.EL (RHEL 4)
1590 2.6.12.6 vanilla (kernel.org)
1591 2.6.16.27-0.9 (SLES 10)
1592 * Recommended e2fsprogs version: 1.39.cfs5
1594 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1598 Frequency : occasional, depends on client load and configuration
1599 Bugzilla : 12181, 12203
1600 Description: data loss for recently-modified files
1602 Details : In some cases it is possible that recently written or created
1603 files may not be written to disk in a timely manner (this should
1604 normally be within 30s unless client IO load is very high).
1605 The problem appears as zero-length files or files that are a
1606 multiple of 1MB in size after a client crash or client eviction
1607 that are missing data at the end of the file.
1609 This problem is more likely to be hit on clients where files are
1610 repeatedly created and unlinked in the same directory, clients
1611 have a large amount of RAM, have many CPUs, the filesystem has
1612 many OSTs, the clients are rebooted frequently, and/or the files
1613 are not accessed by other nodes after being written.
1615 The presence of the problem can be detected by looking at
1616 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1617 smaller than the second (nr_unused) then dirty files will not
1618 be flushed automatically to disk. "sync; sleep 10" should be
1619 run several times on the node before unmounting it to update
1620 Lustre (this is also safe to run on nodes without this problem).
1622 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1623 that can cause this same problem, so customers using that kernel
1624 also need to update the kernel in addition to Lustre. In order
1625 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1627 It is normal that files written just before a client crash (less
1628 than 30s) may not yet have been flushed to disk, even for local
1632 Frequency : frequent on thin XT3 nodes
1634 Description: UUID collision on thin XT3 Linux nodes
1635 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1636 randomly, since we relied on an insufficiently-seeded PRNG.
1641 Description: OSS hangs after "All ost request buffers busy"
1642 Details : A deadlock between quota and journal operations caused OSS
1643 hangs after printing "All ost request buffers busy."
1646 Frequency : always on liblustre builds
1648 Description: Cleanup compiler warnings on liblustre
1651 Frequency : always on liblustre builds on XT3
1653 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1657 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1658 > 32000 subdirectories
1659 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1660 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1661 subdirectory count crosses 32000. This will aid e2fsck to
1662 correctly handle more than 32000 subdirectories.
1665 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1667 Description: System hang on startup
1668 Details : This bug allowed the liblustre (e.g. catamount) client to
1669 return to the app before handling all startup RPCs. This
1670 could leave the node unresponsive to lustre network traffic
1671 and manifested as a server ptllnd timeout.
1673 Severity : enhancement
1675 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1676 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1677 boolean to control whether to print peer debug info when a
1678 client's RPC times out.
1683 Description: make O_SYNC working on 2.6 kernels
1684 Details : 2.6 kernels use different method for mark pages for write,
1685 so need add a code to lustre for O_SYNC work.
1690 Description: Failure to close file and release space on NFS
1691 Details : Put inode details into lock acquired in ll_intent_file_open.
1692 Use mdc_intent_lock in ll_intent_open to properly
1693 detect all kind of errors unhandled by mdc_enqueue.
1698 Description: proc file read during shutdown sometimes raced obd removal,
1700 Details : Add lock to prevent obd access after proc file removal.
1703 Frequency : Only for files larger than 4GB on 32-bit clients.
1705 Description: improperly doing page alignment of locks
1706 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1707 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1712 Description: RPCs being resent when they shouldn't be
1713 Details : Some RPCs that should not be resent are being resent. This
1714 can cause inconsistencies in the RPC state machine. Do not
1715 resend such requests.
1718 Frequency : rare, only with NFS export
1720 Description: Crash on NFS re-export node
1721 Details : under very unusual load conditions an assertion is hit in
1722 ll_intent_file_open()
1725 Frequency : only if OST filesystem is corrupted
1727 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1728 Details : for a short time RPCs with bulk IO are in the replay list,
1729 but replay of bulk IOs is unimplemented. If the OST filesystem
1730 is corrupted due to disk cache incoherency and then replay is
1731 started it is possible to trip an assertion. Avoid putting
1732 committed RPCs into the replay list at all to avoid this issue.
1737 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1738 Details : Large single O_DIRECT read and write calls can fail to allocate
1739 a sufficiently large buffer to process the request. In case of
1740 allocation failure the allocation is retried with a smaller
1741 buffer and broken into smaller requests.
1743 Severity : enhancement
1745 Description: Add -o localflock option to simulate old noflock behaviour.
1746 Details : This will achieve local-only flock/fcntl locks coherentness.
1751 Description: versioning check is incomplete
1752 Details : Checking the version difference of client vs. server, report
1753 error if the gap is too big.
1758 Description: add support PG_writeback bit
1759 Details : add support for PG_writeback bit for Lustre, for more carefull
1760 work with page cache in 2.6 kernel. This also fix some deadlocks
1761 and remove hack for work O_SYNC with 2.6 kernel.
1763 Severity : enhancement
1765 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1766 Details : The uninit_groups feature works in conjunction with the kernel
1767 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1768 up the pass1 processing of e2fsck. This is a read-only feature
1769 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1770 mount filesystems that have had this feature enabled.
1772 Severity : enhancement
1774 Description: Improve multi-block allocation algorithm to avoid fragmentation
1775 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1776 allocation locality and avoid filesystem fragmentation.
1778 ------------------------------------------------------------------------------
1780 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1782 * Support for kernels:
1783 2.6.9-42.0.3.EL (RHEL 4)
1784 2.6.5-7.276 (SLES 9)
1785 2.4.21-47.0.1.EL (RHEL 3)
1786 2.6.12.6 vanilla (kernel.org)
1787 2.6.16.21-0.8 (SLES10)
1788 * Recommended e2fsprogs version: 1.39.cfs2-0
1790 * The backwards-compatible /proc/sys/portals symlink has been removed
1791 in this release. Before upgrading, please ensure that you change
1792 any configuration scripts or /etc/sysctl.conf files that access
1793 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1794 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1795 in advance of the upgrade on any system running Lustre 1.4.6 or
1796 newer, since /proc/sys/lnet was added in that version.
1797 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1801 Frequency : only when quota is used
1803 Description: avoid scanning export list for quota master
1804 Details : Change the algorithms to avoid scanning export list in order
1805 to improve the efficiency.
1808 Frequency : MDS failover only, very rarely
1810 Description: "went back in time" messages on mds failover
1811 Details : The greatest transno may be lost when the current operation
1812 finishes with an error (transno==0) and the client's last_rcvd
1813 record is over-written. Save the greatest transno in the
1814 mds_last_transno for this case.
1817 Frequency : always for specific kernels and striping counts
1819 Description: client may get "Matching packet too big" without ACL support
1820 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1821 when trying to access files in certain configurations. The
1822 clients should in fact be denied when mounting because they do
1823 not understand ACLs.
1826 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1828 Description: many clients connecting with IO in progress causes connect timeouts
1829 Details : Avoid synchronous journal commits to avoid delays caused by many
1830 clients connecting/disconnecting when bulk IO is in progress.
1831 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1832 OST_IO_PORTAL to avoid delays behind potentially many pending
1836 Frequency : occasionally with multiple writers to a single file
1838 Description: shared writes to file may result in wrong size reported by stat()
1839 Details : Allow growing of kms when extent lock is cancelled
1842 Frequency : always with random mmap IO to multi-striped file
1844 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1845 Details : Only if the hole is at the end of OST object so that kms is too
1846 small. Fix is to increase kms accordingly in ll_nopage.
1849 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1851 Description: writes to a missing object would leak memory on the OST
1852 Details : If there is an inconsistency between the MDS and OST filesystems,
1853 such that the MDS references an object that doesn't exist, writes
1854 to that object will leak memory due to incorrect cleanup in the
1855 error handling path, eventually running out of memory on the OST.
1860 Description: Creating too long symlink causes lustre errors
1861 Details : Check symlink and name lengths before sending requests to MDS.
1864 Frequency : only if flock is enabled (not on by default)
1866 Description: posix locks not released on fd closure on 2.6.9+
1867 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1868 kernels, this caused such locks not to be released on fd close and
1869 then assertions on fs unmount about still used locks.
1872 Frequency : MDS failover only, very rarely
1874 Description: clients may get ASSERTION(granted_lock != NULL)
1875 Details : When request was taking a long time, and a client was resending
1876 a getattr by name lock request. The were multiple lock requests
1877 with the same client lock handle and
1878 mds_getattr_name->fixup_handle_for_resent_request found one of the
1879 lock handles but later failed with ASSERTION(granted_lock != NULL).
1884 Description: handle->h_buffer_credits > 0, assertion failure
1885 Details : h_buffer_credits is zero after truncate, causing assertion
1886 failure. This patch extends the transaction or creates a new
1890 Frequency : NFS re-export or patchless client
1891 Bugzilla : 11179, 10796
1892 Description: Crash on NFS re-export node (__d_move)
1893 Details : We do not want to hash the dentry if we don't have a lock.
1894 But if this dentry is later used in d_move, we'd hit uninitialised
1895 list head d_hash, so we just do this to init d_hash field but
1896 leave dentry unhashed.
1899 Frequency : NFS re-export or patchless client
1901 Description: NFS exports has problem with symbolic link
1902 Details : lustre client didn't properly install dentry when re-exported
1903 to NFS or running patchless client.
1906 Frequency : NFS re-export or patchless client
1908 Description: Various nfs/patchless fixes.
1909 Details : fixes reuse disconected alias for lookup process - this fixes
1910 warning "find_exported_dentry: npd != pd",
1911 fix permission error with open files at nfs.
1912 fix apply umask when do revalidate.
1915 Frequency : occasional
1917 Description: Crash on NFS re-export node
1918 Details : calling clear_page() on the wrong pointer triggered oops in
1919 generic_mapping_read().
1922 Frequency : rarely, using O_DIRECT IO
1924 Description: unaligned directio crashes client with LASSERT
1925 Details : check for unaligned buffers before trying any requests.
1928 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1930 Description: stale data returned from RAID cache
1931 Details : If only a small amount of IO is done to the RAID device before
1932 reading it again it is possible to get stale data from the RAID
1933 cache instead of reading it from disk.
1936 Frequency : always for sles10 kernel
1938 Description: sles10 support
1939 Details : ll_follow_link: compile fixes and using of nd_set_link
1940 under newer kernels.
1943 Frequency : depends on arch, kernel and compiler version, always on sles10
1946 Description: recursive or deep enough symlinks cause stack overflow
1947 Details : getting rid of large stack-allocated variable in
1951 Frequency : depends on hardware
1953 Description: lustre write performance loss in the SLES10 kernel
1954 Details : the performance loss is caused by using of write barriers in the
1955 ext3 code. The SLES10 kernel turns barrier support on by
1956 default. The fix is to undo that change for ldiskfs.
1958 ------------------------------------------------------------------------------
1960 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1962 * Support for kernels:
1963 2.6.9-42.0.3EL (RHEL 4)
1964 2.6.5-7.276 (SLES 9)
1965 2.4.21-47.0.1.EL (RHEL 3)
1966 2.6.12.6 vanilla (kernel.org)
1970 Frequency : quota enabled and large files being deleted
1972 Description: releasing more than 4GB of quota at once hangs OST
1973 Details : If a user deletes more than 4GB of files on a single OST it
1974 will cause the OST to spin in an infinite loop. Release
1975 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1980 Description: statfs data retrieved from /proc may be stale or zero
1981 Details : When reading per-device statfs data from /proc, in the
1982 {kbytes,files}_{total,free,avail} files, it may appear
1983 as zero or be out of date.
1986 Frequency : systems with MD RAID1 external journal devices
1988 Description: lconf's call to blkid is confused by RAID1 journal devices
1989 Details : Use the "blkid -l" flag to locate the MD RAID device instead
1990 of returning all block devices that match the journal UUID.
1993 Frequency : always, for aggregate stripe size over 4GB
1995 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
1996 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
1997 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2000 Frequency : always if "lfs find" used on a local file/directory
2002 Description: "lfs find" segfaults if used on a local file/directory
2003 Details : The case where a directory component was not specified wasn't
2004 handled correctly. Handle this properly.
2007 Frequency : always on ppc64
2009 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2010 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2011 perform endianness conversion.
2014 Frequency : rarely (truncate to non-zero file size after write under load)
2015 Bugzilla : 10730, 10687
2016 Description: Files padded with zeros to next 4K multiple
2017 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2018 it is possible that files that are truncated to a non-zero size
2019 immediately after being written are filled with zero bytes beyond
2020 the truncated size. No file data is lost.
2022 Severity : enhancement
2024 Description: Allow recovery/failover for liblustre clients.
2025 Details : liblustre clients were unaware of failover configurations until
2028 Severity : enhancement
2030 Description: user file locks should fail when not mounting with flock option
2031 Details : Set up an error-returning stub in ll_file_operations.lock field
2032 to prevent incorrect behaviour when client is mounted without
2033 flock option. Also, set up properly f_op->flock field for
2037 Frequency : always on ia64
2039 Description: "lfs df" loops on printing out MDS statfs information
2040 Details : The obd_ioctl_data was not initialized and in some systems
2041 this caused a failure during the ioctl that did not return
2042 an error. Initialize the struct and return an error on failure.
2045 Frequency : SLES 9 only
2047 Description: Error of copying files with lustre special EAs as root
2048 Details : Client side always return success for setxattr call for lustre
2049 special xattr (currently only "trusted.lov").
2052 Frequency : rarely on clusters with both ia64+i386 clients
2054 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2055 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2056 clients are doing concurrent writes to the same file it is
2057 possible that the ia64 clients may LASSERT because the OST
2058 extent locks are not PAGE_SIZE aligned. Ensure that grown
2059 locks are always aligned on the request boundary.
2062 Frequency : specific use, occasional
2064 Description: Overwriting in use executable truncates on-disk binary image
2065 Details : If one node attempts to overwrite an executable in use by
2066 another node, we now correctly return ETXTBSY instead of
2067 truncating the file.
2069 Severity : enhancement
2071 Description: Async OSC create to avoid the blocking unnecessarily.
2072 Details : If a OST has no remain object, system will block on the creating
2073 when need to create a new object on this OST. Now, ways use
2074 pre-created objects when available, instead of blocking on an
2075 empty osc while others are not empty. If we must block, we block
2076 for the shortest possible period of time.
2081 Description: chmod on Lustre root is propagated to other clients
2082 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2083 invalid by the follow_mount time.
2086 Frequency : liblustre clients only
2088 Description: Race in 'instant cancel' lock handling could lead to such locks
2089 never to be granted in case of SMP MDS
2090 Details : Do not destroy not yet granted but cbpending locks in
2094 Frequency : replay/resend of open
2096 Description: non null lock assetion failure in mds_intent_policy
2097 Details : Trying to replay/resend lockless open requests resulted in
2098 mds_open() returning 0 with no lock. Now it sets a flag if
2099 a lock is going to be returned.
2101 Severity : enhancement
2103 Description: Checksum enhancements
2104 Details : New checksum enhancements allow for resending RPCs that failed
2107 Severity : enhancement
2109 Description: Tunables on number of dirty pages in cacche
2110 Details : Allow to set limit on number of dirty pages cached.
2115 Description: client crash on unmount - lock still has references
2116 Details : In some error handling cases it was possible to leak a lock
2117 reference on a client while accessing a file. This was not
2118 harmful to the client during operation, but would cause the
2119 client to crash when the filesystem is unmounted.
2122 Frequency : specific case, rare
2124 Description: ETXTBSY on mds though file not in use by client
2125 Details : ETXTBSY is no longer incorrectly returned when attempting to
2126 chmod or chown a directory that the user previously tried to
2127 execute or a currently-executing binary.
2130 Frequency : extremely rare except on liblustre-based clients
2132 Description: Lustre space not freed when files are deleted
2133 Details : Clean up open-unlinked files after client eviction. Previously
2134 the unlink was skipped and the files remained as orphans.
2139 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2140 Details : In some cases it was possible to send a blocking callback to
2141 a client doing a glimpse, even though that client didn't get
2142 a lock granted. When the glimpse lock is cancelled on the OST
2143 the freed lock is left on the waiting list and corrupted the list.
2146 Frequency : all core dumps
2148 Description: Broke core dumps to lustre
2149 Details : Negative dentry may be unhashed if parent does not have UPDATE
2150 lock, but some callers, e.g. do_coredump, expect dentry to be
2151 hashed after successful create, hash it in ll_create_it.
2153 ------------------------------------------------------------------------------
2155 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2157 * Support for kernels:
2158 2.6.9-42.0.2.EL (RHEL 4)
2159 2.6.5-7.276 (SLES 9)
2160 2.4.21-40.EL (RHEL 3)
2161 2.6.12.6 vanilla (kernel.org)
2165 Frequency : always on RHEL 3
2167 Description: Number of open files grows over time
2168 Details : The number of open files grows over time, whether or not
2169 Lustre is started. This was due to a filp leak introduced
2170 by one of our kernel patches.
2172 ------------------------------------------------------------------------------
2174 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2176 * Support for kernels:
2177 2.6.9-42.EL (RHEL 4)
2178 2.6.5-7.276 (SLES 9)
2179 2.4.21-40.EL (RHEL 3)
2180 2.6.12.6 vanilla (kernel.org)
2185 Bugzilla : 5719, 9635, 9792, 9684
2186 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2187 Details : If a server is under heavy load and cannot reply to new
2188 connection requests before the client resends the (re)connect,
2189 the connection handling code can behave badly if two service
2190 threads are concurrently handing separate (re)connections from
2191 the same client. Add better locking to the connection handling
2192 code, and ensure that only a single connection will be processed
2193 for a given client UUID, even if the lock is dropped.
2195 Severity : enhancement
2197 Description: add TCP zero-copy support to kernel
2198 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2199 sends if the hardware supports scatter-gather and checksumming.
2200 This allows socklnd to do client-write and server-read more
2201 efficiently and reduce CPU utilization from skbuf copying.
2204 Frequency : only if NFS exporting from client
2206 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2207 Details : If setting attributes on a file created under NFS that had
2208 never been opened it would be possible to oops the client
2209 if the file had no objects.
2212 Frequency : always for liblustre
2214 Description: liblustre client does MDS+OSTs setattr RPC for each write
2215 Details : When doing a write from a liblustre client, the client
2216 incorrectly issued an RPC to the MDS and each OST the file was
2217 striped over in order to update the timestamps. When writing
2218 with small chunks and many clients this could overwhelm the MDS
2219 with RPCs. In all cases it would slow down the write because
2220 these RPCs are unnecessary.
2222 Severity : enhancement
2224 Description: allow number of MDS service threads to be changed at module load
2225 Details : It is now possible to change the number of MDS service threads
2226 running. Adding "options mds mds_num_threads={N}" to the MDS's
2227 /etc/modprobe.conf will set the number of threads for the next
2228 time Lustre is restarted (assuming the "mds" module is also
2229 reloaded at that time). The default number of threads will
2230 stay the same, 32 for most systems.
2235 Description: OST crash if filesystem is unformatted or corrupt
2236 Details : If an OST is started on a device that has never been formatted
2237 or if the filesystem is corrupt and cannot even mount then the
2238 error handling cleanup routines would dereference a NULL pointer.
2243 Description: NULL pointer deref in llap_from_page.
2244 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2245 incorrect) mapping if the page was truncated/reclaimed while it was
2246 searched for. Check for this condition and skip such pages when
2247 doing readahead. Introduce extra check to llap_from_page() to
2248 verify page->mapping->host is non-NULL (so page is not anonymous).
2251 Frequency : Sometimes when using sys_sendfile
2253 Description: "page not covered by a lock" warnings from ll_readpage
2254 Details : sendfile called ll_readpage without right page locks present.
2255 Now we introduced ll_file_sendfile that does necessary locking
2256 around call to generic_file_sendfile() much like we do in
2260 Frequency : with certain MDS communication failures at client mount time
2262 Description: NULL pointer deref after failed client mount
2263 Details : a client connection request may delayed by the network layer
2264 and not be sent until after the PTLRPC layer has timed out the
2265 request. If the client fails the mount immediately it will try
2266 to clean up before the network times out the request. Add a
2267 reference from the request import to the obd device and delay
2268 the cleanup until the network drops the request.
2271 Frequency : occasionally during client (re)connect
2273 Description: assertion failure during client (re)connect
2274 Details : processing a client connection request may be delayed by the
2275 client or server longer than the client connect timeout. This
2276 causes the client to resend the connection request. If the
2277 original connection request is replied in this interval, the
2278 client may trip an assertion failure in ptlrpc_connect_interpret()
2279 which thought it would be the only running connect process.
2282 Frequency : only with obd_echo servers and clients that are rebooted
2284 Description: kernel BUG accessing uninitialized data structure
2285 Details : When running an obd_echo server it did not start the ping_evictor
2286 thread, and when a client was evicted an uninitialized data
2287 structure was accessed. Start the ping_evictor in the RPC
2288 service startup instead of the OBD startup.
2290 Severity : enhancement
2291 Bugzilla : 10193 (patchless)
2292 Description: Remove dependency on various unexported kernel interfaces.
2293 Details : No longer need reparent_to_init, exit_mm, exit_files,
2294 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2297 Frequency : rare (only users of deprecated and unsupported LDAP config)
2299 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2300 Details : LDAP apparently contains 'lustreName' attributes instead of
2301 'name'. A simple remapping of the name is sufficient.
2304 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2306 Description: waiting_locks_callback trips kernel BUG if client is evicted
2307 Details : Running with the dump_on_timeout debug flag turned on makes
2308 it possible that the waiting_locks_callback() can try to dump
2309 the Lustre kernel debug logs from an interrupt handler. Defer
2310 this log dumping to the expired_lock_main() thread.
2312 Severity : enhancement
2314 Description: Support NFS exporting on 2.6 kernels.
2315 Details : Implement non-rawops metadata methods for NFS server to use without
2316 changing NFS server code.
2319 Frequency : very rare (synthetic metadata workload only)
2321 Description: two racing renames might cause an MDS thread to deadlock
2322 Details : Running the "racer" program may cause one MDS thread to rename
2323 a file from being the source of a rename to being the target of
2324 a rename at exactly the same time that another thread is doing
2325 so, and the second thread has already enqueued these locks after
2326 doing a lookup of the target and is trying to relock them in
2327 order. Ensure that we don't try to re-lock the same resource.
2330 Frequency : only very large systems with liblustre clients
2332 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2333 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2334 all OSTs in parallel. This allows the eviction of stale liblustre
2335 clients to proceed much faster than if they were done in series,
2336 and also offers similar improvements for other set_info RPCs.
2341 Description: excessive CPU usage during initial read phase on client
2342 Details : During the initial read phase on a client, it would agressively
2343 retry readahead on the file, consuming too much CPU and impacting
2344 performance (since 1.4.5.8). Improve the readahead algorithm
2345 to avoid this, and also improve some other common cases (read
2346 of small files in particular, where "small" is files smaller than
2347 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2352 Description: MDS crash when receiving packet with unknown intent.
2353 Details : Do not LBUG in unknown intent case, just return -EFAULT
2355 Severity : enhancement
2356 Bugzilla : 9293, 9385
2357 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2358 Details : Do not serialize getattr (non-intent version) and statfs.
2361 Frequency : occasional, when OST network is overloaded/intermittent
2363 Description: client evicted by OST after bulk IO timeout
2364 Details : If a client sends a bulk IO request (read or write) the OST
2365 may evict the client if it is unresposive to its data GET/PUT
2366 request. This is incorrect if the network is overloaded (takes
2367 too long to transfer the RPC data) or dropped the OST GET/PUT
2368 request. There is no need to evict the client at all, since
2369 the pinger and/or lock callbacks will handle this, and the
2370 client can restart the bulk request.
2373 Frequency : Always when mmapping file with no objects
2375 Description: client crashes when mmapping file with no objects
2376 Details : Check that we actually have objects in a file before doing any
2377 operations on objects in ll_vm_open, ll_vm_close and
2383 Description: Request leak when working with deleted CWD
2384 Details : Introduce advanced request refcount tracking for requests
2385 referenced from lustre intent.
2387 Severity : Enhancement
2389 Description: Cache open file handles on client.
2390 Details : MDS now will return special lock along with openhandle, if
2391 requested and client is allowed to hold openhandle, even if unused,
2392 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2393 closing files for every read/write openration.
2395 Severity : Enhancement
2397 Description: Cache open negative dentries on client when possible.
2398 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2399 negative dentries on lock revocation.
2404 Description: Remounting a client read-only wasn't possible with a zconf mount
2405 Details : It wasn't possible to remount a client read-only with llmount.
2407 Severity : enhancement
2408 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2409 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2410 MPICH is included to add Lustre-specific ADIO interfaces.
2411 This is based closely on the UFS ADIO layer and only differs
2412 in file creation, in order to allow the OST striping to be set.
2413 This is user-contributed code and not supported by CFS.
2418 Description: extended inode attributes (immutable, append-only) work improperly
2419 when 2.4 and 2.6 kernels are used on client/server or vice versa
2420 Details : Introduce kernel-independent values for these flags.
2422 Severity : enhancement
2425 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2426 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2427 granularity. Now, Fractional MB granularity is be supported,
2428 this is very useful for low memory system.
2430 Severity : enhancement
2432 Description: Getattr by fid
2433 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2434 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2435 have locks and avoids getattr rpc requests for it on every stat.
2438 Frequency : Always, for filesystems larger than 2TB
2440 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2441 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2442 filesystems larger than 2TB to be mounted with mballoc enabled.
2445 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2447 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2448 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2449 filesystems larger than 2TB is mounted then the kernel will
2450 silently corrupt the start of the filesystem. CONFIG_LBD is
2451 enabled for all CFS-supported kernels, but the possibility of
2452 this happening with a modified kernel config exists.
2454 Severity : enhancement
2456 Description: add client O_DIRECT support for 2.6 kernels
2457 Details : It is now possible to do O_DIRECT reads and writes to files
2458 in the Lustre client mountpoint on 2.6 kernel clients.
2460 Severity : enhancement
2462 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2463 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2464 parallel, not waiting for response from every OST before sending
2465 a rpc to the next OST.
2470 Description: setattr vs write race when updating file timestamps
2471 Details : Client processes that update a file timestamp into the past
2472 right after writing to the file (e.g. tar) it is possible that
2473 the updated file modification time can be reset to the current
2474 time due to a race between processing the setattr and write RPC.
2476 Severity : enhancement
2478 Description: Bring 'lfs find' closer in line with regular Linux find.
2479 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2480 -print0 options and obtains all the needed info through the lustre
2483 Severity : enhancement
2485 Description: support up to 1024 configured devices on one node
2486 Details : change obd_dev array from statically allocated to dynamically
2487 allocated structs as they are first used to reduce memory usage
2492 Description: Flush dirty partially truncated pages during truncate
2493 Details : Immediatelly flush partially truncated pages in filter_setattr,
2494 this way we completely avoid having any pages in page cache on OST
2495 and can retire ugly workarounds during writes to flush such pages.
2500 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2501 Details : protect inode from truncation within vfs_unlink() context
2502 just take a reference before calling vfs_unlink() and release it
2503 when parent's i_sem is free.
2508 Description: last_id value checked outside lock on OST caused LASSERT failure
2509 Details : If there were multiple MDS->OST object precreate requests in
2510 flight, it was possible that the OST's last object id was checked
2511 outside a lock and incorrectly tripped an assertion. Move checks
2512 inside locks, and discard old precreate requests.
2515 Frequency : always, if extents are used on OSTs
2517 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2518 Details : OSTs using the extents format would not zero the high 16 bits of
2519 the index physical block number. This is not a problem for any
2520 OST filesystems smaller than 16TB, and no kernels support ext3
2521 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2522 new/modified files) and can be fixed for existing filesystems
2523 with e2fsprogs-1.39-cfs1.
2528 Description: import connection selection may be incorrect if timer wraps
2529 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2530 import connections to be ignored if the 32-bit jiffies counter
2531 wraps. Use a 64-bit jiffies counter.
2534 Frequency : very large clusters immediately after boot
2536 Description: LNET request buffers exhausted under heavy short-term load
2537 Details : If a large number of client requests are generated on a service
2538 that has previously never seen so many requests it is possible
2539 that the request buffer growth cannot keep up with the spike in
2540 demand. Instead of dropping incoming requests, they are held in
2541 the LND until the RPC service can accept more requests.
2544 Frequency : Sometimes during replay
2546 Description: Assertion failure in ll_local_open after replay.
2547 Details : If replay happened on an open request reply before we were able
2548 to set replay handler, reply will become not swabbed tripping the
2549 assertion in ll_local_open. Now we set the handler right after
2550 recognising of open request
2553 Frequency : very rare
2555 Description: kernel reports "badness in vsnprintf"
2556 Details : Reading from the "recovery_status" /proc file in small chunks
2557 may cause a negative length in lprocfs_obd_rd_recovery_status()
2558 call to vsnprintf() (which is otherwise harmless). Exit early
2559 if there is no more space in the output buffer.
2561 Severity : enhancement
2563 Description: clear OBD RPC statistics by writing to them
2564 Details : It is now possible to clear the OBD RPC statistics by writing
2565 to the "stats" file.
2570 Description: Client mtime is not the same on different clients after utimes
2571 Details : In some cases, the client was using the utimes() syscall on
2572 a file cached on another node. The clients now validate the
2573 ctime from the MDS + OSTs to determine which one is right.
2578 Description: Inability to activate failout mode
2579 Details : lconf script incorrectly assumed that in pythong string's numeric
2580 value is used in comparisons.
2583 Frequency : always with multiple stripes per file
2585 Description: Inefficient object allocation for mutli-stripe files
2586 Details : When selecting which OSTs to stripe files over, for files with
2587 a stripe count that divides evenly into the number of OSTs,
2588 the MDS is always picking the same starting OST for each file.
2589 Return the OST selection heuristic to the original design.
2594 Description: mount failures may take full timeout to return an error
2595 Details : Under some heavy load conditions it is possible that a
2596 failed mount can wait for the full obd_timeout interval,
2597 possibly several minutes, before reporting an error.
2598 Instead return an error as soon as the status is known.
2600 Frequency : quota enabled and large files being deleted
2602 Description: releasing more than 4GB of quota at once hangs OST
2603 Details : If a user deletes more than 4GB of files on a single OST it
2604 will cause the OST to spin in an infinite loop. Release
2605 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2610 Description: statfs data retrieved from /proc may be stale or zero
2611 Details : When reading per-device statfs data from /proc, in the
2612 {kbytes,files}_{total,free,avail} files, it may appear
2613 as zero or be out of date.
2616 Frequency : systems with MD RAID1 external journal devices
2618 Description: lconf's call to blkid is confused by RAID1 journal devices
2619 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2620 of returning all block devices that match the journal UUID.
2623 Frequency : always, for aggregate stripe size over 4GB
2625 Description: assertion fails when trying to use 4GB stripe size
2626 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2627 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2630 Frequency : always on ppc64
2632 Description: the first write on an ext3 filesystem with mballoc got stuck
2633 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2634 perform endianness conversion.
2636 ------------------------------------------------------------------------------
2638 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2640 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2641 INTEROPERATE with older versions automatically. Please read the
2642 user documentation before upgrading any part of a live system.
2643 * WARNING: Lustre networking configuration changes are required with
2644 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2647 * Support for kernels:
2648 2.6.9-22.0.2.EL (RHEL 4)
2649 2.6.5-7.244 (SLES 9)
2650 2.6.12.6 vanilla (kernel.org)
2653 Severity : enhancement
2654 Bugzilla : 7981/8208
2655 Description: Introduced Lustre Networking (LNET)
2656 Details : LNET is new networking infrastructure for Lustre, it includes
2657 a reorganized network configuration mode (see the user
2658 documentation for full details) as well as support for routing
2659 between different network fabrics. Lustre Networking Devices
2660 (LNDs) for the supported network fabrics have also been
2661 created for this new infrastructure.
2663 Severity : enhancement
2664 Description: Introduced Access control lists
2665 Details : clients can set ACLs on files and directories in order to have
2666 more fine-grained permissions than the standard Unix UGO+RWX.
2667 The MDS must be started with the "-o acl" mount option.
2669 Severity : enhancement
2670 Description: Introduced filesystem quotas
2671 Details : Administrators may now establish per-user quotas on the
2674 Severity : enhancement
2676 Description: Configuration change for the XT3
2677 The PTLLND is now used to run Lustre over Portals on the XT3
2678 The configure option(s) --with-cray-portals are no longer used.
2679 Rather --with-portals=<path-to-portals-includes> is used to
2680 enable building on the XT3. In addition to enable XT3 specific
2681 features the option --enable-cray-xt3 must be used.
2686 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2687 Details : A race between allocating a new llcd and re-getting the llcd_lock
2688 allowed another thread to grab newly-allocated llcd.
2690 Severity : enhancement
2692 Description: 2.6 OST async journal commit and locking fix to improve performance
2693 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2694 2.6 kernels are now async as they already were in 2.4 kernels so
2695 that they can commit concurrently with the network bulk transfer.
2696 For block-allocated files the filter allocation semaphore is held
2697 to avoid filesystem fragmentation during allocation. BKL lock
2698 removed for 2.6 xattr operations where it is no longer needed.
2703 Description: lconf incorrectly determined whether two IP networks could talk
2704 Details : In some more complicated routing and multiple-network
2705 configurations, lconf will avoid trying to make a network
2706 connection to a disjoint part of the IP space. It was doing the
2707 math incorrectly for one set of cases.
2712 Description: Fix for potential infinite loop processing records in an llog.
2713 Details : If an llog record is corrupted/zeroed, it is possible to loop
2714 forever in llog_process(). Validate the llog record length
2715 and skip the remainder of the block on error.
2718 Frequency : occasional (liblustre only)
2720 Description: liblustre could not open files whose last component is a symlink
2721 Details : sysio_path_walk() would incorrectly pass the open intent to
2722 intermediate path components.
2725 Frequency : rare (liblustre only with non-standard tuning)
2726 Bugzilla : 7201 (7350)
2727 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2728 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2729 after it has been released, only do this for referenced locks
2731 Severity : enhancement
2733 Description: specifying an (invalid) directory default stripe_size of -1
2734 would reset the directory default striping
2735 Details : stripe_size -1 was used internally to signal directory stripe
2736 removal, now use "all default" to signal dir stripe removal
2737 as a directory striping of "all default" is not useful
2740 Frequency : common for large clusters running liblustre clients
2742 Description: doing an ls when liblustre clients are running is slow
2743 Details : sending a glimpse AST to a liblustre client waits for every AST
2744 to time out, as liblustre clients will not respond. Since they
2745 cannot cache data we refresh the OST lock LVB from disk instead.
2747 Severity : enhancement
2749 Description: doing an ls at the same time as file IO can be slow
2750 Details : enqueue and other "small" requests can be blocked behind many
2751 large IO requests. Create a new OST IO portal for non-IO
2752 requests so they can be processed faster.
2755 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2757 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2758 created file with mode 000
2759 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2760 allow an ftruncate() as a non-root user to a file with mode 000.
2761 Lustre now respects this flag to disable mode checks when
2762 truncating a file owned by the user
2765 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2768 Description: Revoking locks from clients that went dead or catatonic might take
2770 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2771 cancellation of such locks instant on servers without waiting for
2772 any reply from clients. Clients drops these locks when cancel
2773 notification from server is received without replying.
2776 Frequency : liblustre-only, when liblustre client dies or becomes busy
2778 Description: Doing ls on Linux clients can take a long time with active
2780 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2781 granting such locks to it in the first place if possible. Locks
2782 are taken by proxy on the OST during the read or write and
2783 dropped immediately afterward. Add connect flags handling, do
2784 not grant locks to liblustre clients for glimpse ASTs.
2786 Severity : enhancement
2788 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2789 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2790 sequential reads and behaves badly with applications doing random
2791 reads. Improve it by reading ahead at least read region, and
2792 avoiding excessive large RPC for small reads.
2794 Severity : enhancement
2796 Description: Creating more than 1000 files for a single job may cause a load
2797 imbalance on the OSTs if there are also a large number of OSTs.
2798 Details : qos_prep_create() uses an OST index reseed value that is an
2799 even multiple of the number of available OSTs so that if the
2800 reseed happens in the middle of the object allocation it will
2801 still utilize the OSTs as uniformly as possible.
2806 Description: OST or MDS may oops in ping_evictor_main()
2807 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2808 but doesn't restart at beginning of obd_exports_timed list
2811 Severity : enhancement
2813 Description: improve by-nid export eviction on the MDS and OST
2814 Details : allow multiple exports with the same NID to be evicted at one
2815 time without re-searching the exports list.
2818 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2820 Description: MDS may oops in groups_free()
2821 Details : in rare race conditions a newly allocated group_info struct is
2822 freed again, and this can be NULL. The 2.4 compatibility code
2823 for groups_free() checked for a NULL pointer, but 2.6 did not.
2826 Frequency : common for liblustre clients doing little filesystem IO
2827 Bugzilla : 9352, 7313
2828 Description: server may evict liblustre clients accessing contended locks
2829 Details : if a client is granted a lock or receives a completion AST
2830 with a blocking AST already set it would not reply to the AST
2831 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2834 Frequency : lfs setstripe, only systems with more than 160 OSTs
2836 Description: unable to set striping with a starting offset beyond OST 160
2837 Details : llapi_create_file() incorrectly limited the starting stripe
2838 index to the maximum single-file stripe count.
2841 Frequency : LDAP users only
2843 Description: lconf did not handle in-kernel recovery with LDAP properly
2844 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2846 Severity : enhancement
2848 Description: bind OST threads to NUMA nodes to improve performance
2849 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2850 node and do their allocations there to localize memory access
2852 Severity : enhancement
2854 Description: llmount can determine client NID directly from Myrinet (GM)
2855 Details : the client NID code from gmnalnid was moved directly into
2856 llmount, removing the need to use this or specifying the
2857 client NID explicitly when mounting GM clients with zeroconf
2860 Frequency : if client is started with down MDS
2862 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2863 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2864 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2865 ptlrpc_import_delayed_req() because we didn't check if the
2866 request was interrupted, and we also didn't break out of the
2867 event loop if there was no timeout
2872 Description: data loss during non-page-aligned writes to a single file from
2873 both multiple nodes and multiple threads on one node at same time
2874 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2875 inconsistency led to false short-reads, that were cached and later
2876 used by ->prepare_write() to fill in partially written page,
2877 leading to data loss.
2880 Frequency : always, if lconf --abort_recovery used
2882 Description: lconf --abort_recovery fails with 'Operation not supported'
2883 Details : lconf was attempting to abort recovery on the MDT device and not
2886 Severity : enhancement
2888 Description: remove cleanup logs
2889 Details : replace lconf-generated cleanup logs with lustre internal
2890 cleanup routines. Eliminates the need for client-cleanup and
2893 Severity : enhancement
2895 Description: add support for EAs (user and system) on lustre filesystems
2896 Details : it is now possible to store extended attributes in the Lustre
2897 client filesystem, and with the user_xattr mount option it
2898 is possible to allow users to store EAs on their files also
2900 Severity : enhancement
2902 Description: Add possibility (config option) to show minimal available OST free
2904 Details : When compiled with --enable-mindf configure option, statfs(2)
2905 (and so, df) will return least minimal free space available from
2906 all OSTs as amount of free space on FS, instead of summary of
2907 free spaces of all OSTs.
2909 Severity : enhancement
2911 Description: do not expand extent locks acquired on OST-side
2912 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2913 by server: they are not cached anyway.
2916 Frequency : when mmap is used/binaries executed from Lustre
2918 Description: Unmmap pages before throwing them away from read cache.
2919 Details : llap_shrink cache now attempts to unmap pages before discarding
2920 them (if unmapping failed - do not discard). SLES9 kernel has
2921 extra checks that trigger if this unmapping is not done first.
2926 Description: lconf didn't resolve symlinks before checking to see whether a
2927 given mountpoint was already in use
2930 Frequency : when migrating failover services
2931 Bugzilla : 6395, 9514
2932 Description: When migrating a subset of services from a node (e.g. failback
2933 from a failover service node) the remaining services would
2934 time out and evict clients.
2935 Details : lconf --force (implied by --failover) sets the global obd_timeout
2936 to 5 seconds in order to quickly disconnect, but this caused
2937 other RPCs to time out too quickly. Do not change the global
2938 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2940 Severity : enhancement
2941 Frequency : if MDS is started with down OST
2942 Bugzilla : 9439,5706
2943 Description: Allow startup/shutdown of an MDS without depending on the
2944 availability of the OSTs.
2945 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2946 Add appropriate locking and lov-osc refcounts for safe
2947 cleaning. Add osc abort_inflight calls in case the
2948 synchronize never started.
2951 Frequency : occasional (Cray XT3 only)
2953 Description: root not authorized to access files in CRAY_PORTALS environment
2954 Details : The client process capabilities were not honoured on the MDS in
2955 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2956 been accessed by an authorized user then root was able to access
2957 the file on the local client also. The root user capabilities
2958 are now allowed on the MDS, as this environment has secure UID.
2961 Frequency : occasional
2963 Description: ldiskfs "too long searching" message happens too often
2964 Details : A debugging message (otherwise harmless) prints too often on
2965 the OST console. This has been reduced to only happen when
2966 there are fragmentation problems on the filesystem.
2971 Description: Division by zero in statfs when all OSCs are inactive
2972 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2973 lov_statfs divides by value returned by lov_get_stripecnt().
2977 Bugzilla : 9489, 3273
2978 Description: First write from each client to each OST was only 4kB in size,
2979 to initialize client writeback cache, which caused sub-optimal
2980 RPCs and poor layout on disk for the first writen file.
2981 Details : Clients now request an initial cache grant at (re)connect time
2982 and so that they can start streaming writes to the cache right
2983 away and always do full-sized RPCs if there is enough data.
2984 If the OST is rebooted the client also re-establishes its grant
2985 so that client cached writes will be honoured under the grant.
2990 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
2991 Details : Now I/O RPCs go to different portal number and (presumably) fast
2992 lock requests (and glimses) and other RPCs get their own service
2993 threads pool that should be able to service those RPCs
2996 Severity : enhancement
2998 Description: Ability to exchange lustre version between client and servers and
2999 issue warnings at client side if client is too old. Also for
3000 liblustre clients there is ability to refuse connection of too old
3002 Details : New 'version' field is added to connect data structure that is
3003 filled with version info. That info is later checked by server and
3007 Frequency : rare, liblustre only.
3008 Bugzilla : 9296, 9581
3009 Description: Two simultaneous writes from liblustre at offset within same page
3010 might proceed at the same time overwriting eachother with stale
3012 Details : I/O lock withing llu_file_prwv was released too early, before data
3013 actually was hitting the wire. Extended lock-holding time until
3014 server acknowledges receiving data.
3017 Frequency : extremely rare. Never observed in practice.
3019 Description: avoid generating lustre_handle cookie of 0.
3020 Details : class_handle_hash() generates handle cookies by incrementing
3021 global counter, and can hit 0 occasionaly (this is unlikely, but
3022 not impossible, because initial value of cookie counter is
3023 selected randonly). Value of 0 is used as a sentinel meaning
3024 "unassigned handle" --- avoid it. Also coalesce two critical
3025 sections in this function into one.
3027 Severity : enhancement
3029 Description: allow liblustre clients to delegate truncate locking to OST
3030 Details : To avoid overhead of locking, liblustre client instructs OST to
3031 take extent lock in ost_punch() on client's behalf. New connection
3032 flag is added to handle backward compatibility.
3034 Severity : enhancement
3035 Bugzilla : 4928, 7341, 9758
3036 Description: allow number of OST service threads to be specified
3037 Details : a module parameter allows the number of OST service threads
3038 to be specified via "options ost ost_num_threads={N}" in the
3039 OSS's /etc/modules.conf or /etc/modprobe.conf.
3043 Bugzilla : 6146, 9635, 9895
3044 Description: servers crash with bad pointer in target_handle_connect()
3045 Details : In rare cases when a client is reconnecting it was possible that
3046 the connection request was the last reference for that export.
3047 We would temporarily drop the export reference and get a new
3048 one, but this may have been the last reference and the export
3049 was just destroyed. Get new reference before dropping old one.
3051 Severity : enhancement
3052 Frequency : if client is started with failover MDS
3054 Description: Allow multiple MDS hostnames in the mount command
3055 Details : Try to read the configuration from all specified MDS
3056 hostnames during a client mount in case the "primary"
3059 Severity : enhancement
3061 Description: Stop sending data to evicted clients as soon as possible.
3062 Details : Check if the client we are about to send or are sending data to
3063 was evicted already. (Check is done every second of waiting,
3064 for which l_wait_event interface was extended to allow checking
3065 of exit condition at specified intervals).
3068 Frequency : rare, normally only when NFS exporting is done from client
3070 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3072 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3075 Frequency : always, if srand() is called before liblustre initialization
3077 Description: Liblustre uses system PRNG disturbing its usage by user application
3078 Details : Introduce internal to lustre fast and high-quality PRNG for
3079 lustre usage and make liblustre and some other places in generic
3080 lustre code to use it.
3082 Severity : enhancement
3083 Bugzilla : 9477, 9557, 9870
3084 Description: Verify that the MDS configuration logs are updated when xml is
3085 Details : Check if the .xml configuration logs are newer than the config
3086 logs stored on the MDS and report an error if this is the case.
3087 Request --write-conf, or allow starting with --old_conf.
3089 Severity : enhancement
3091 Description: Handle symlinks in the path when checking if Lustre is mounted.
3092 Details : Resolve intermediate symlinks when checking if a client has
3093 mounted a filesystem to avoid duplicate client mounts.
3098 Description: lconf can hit an error exception but still return success.
3099 Details : The lconf command catches the Command error exception at the top
3100 level script context and will exit with the associated exit
3101 status, but doesn't ensure that this exit status is non-zero.
3106 Description: failure of ptlrpc thread startup can cause oops
3107 Details : Starting a ptlrpc service thread can fail if there are a large
3108 number of threads or the server memory is very fragmented.
3109 Handle this without oopsing.
3112 Frequency : always, only if liblustre and non-default acceptor port was used
3114 Description: liblustre cannot connect to servers with non-default acceptor port
3115 Details : tcpnal_set_default_params() was not called and was therefore
3116 ignoring the environment varaible TCPNAL_PORT, as well as other
3117 TCPNAL_ environment variables
3122 Description: two objects could be created on the same OST for a single file
3123 Details : If an OST is down, in some cases it was possible to create two
3124 objects on a single OST for a single file. No problems other
3125 than potential performance impact and spurious error messages.
3129 Bugzilla : 5681, 9562
3130 Description: Client may oops in ll_unhash_aliases
3131 Details : Client dcache may become inconsistent in race condition.
3132 In some cases "getcwd" can fail if the current directory is
3138 Description: Inode refcounting problems in NFS export code
3139 Details : link_raw functions used to call d_instantiate without obtaining
3140 extra inode reference first.
3144 Bugzilla : 9942, 9903
3145 Description: Referencing freed requests leading to crash, memleaks with NFS.
3146 Details : We used to require that call to ll_revalidate_it was always
3147 followed by ll_lookup_it. Also with revalidate_special() it is
3148 possible to call ll_revalidate_it() twice for the same dentry
3149 even if first occurence returned success. This fix changes semantic
3150 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3151 reference on a request referred from the intent.
3152 ll_intent_release() then releases such a request.
3155 Frequency : rare, normally benchmark loads only
3157 Description: unlinked inodes were kept in memory on the client
3158 Details : If a client is repeatedly creating and unlinking files it
3159 can accumulate a lot of stale inodes in the inode slab cache.
3160 If there is no other client load running this can cause the
3161 client node to run out of memory. Instead flush old inodes
3162 from client cache that have the same inode number as a new inode.
3165 Frequency : SLES9 2.6.5 kernel and long filenames only
3166 Bugzilla : 9969, 10379
3167 Description: utime reports stale NFS file handle
3168 Details : SLES9 uses out-of-dentry names in some cases, which confused
3169 the lustre dentry revalidation. Change it to always use the
3173 Frequency : rare, unless heavy write-truncate concurrency is continuous
3174 Bugzilla : 4180, 6984, 7171, 9963, 9331
3175 Description: OST becomes very slow and/or deadlocked during object unlink
3176 Details : filter_destroy() was holding onto the parent directory lock
3177 while truncating+unlinking objects. For very large objects this
3178 may block other threads for a long time and slow overall OST
3179 responsiveness. It may also be possible to get a lock ordering
3180 deadlock in this case, or run out of journal credits because of
3181 the combined truncate+unlink. Solution is to do object truncate
3182 first in one transaction without parent lock, and then do the
3183 final unlink in a new transaction with the parent lock. This
3184 reduces the lock hold time dramatically.
3187 Frequency : rare, 2.4 kernels only
3189 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3190 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3191 before unmounting their filesystem in order to avoid deadlock.
3192 The kernel_locked() function in 2.4 kernels only checks whether
3193 the kernel lock is held, not whether it is this process that is
3194 holding it as 2.6 kernels do.
3199 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3200 Details : The client ptlrpc code may be trying to reconnect to a down
3201 server before a previous connection attempt has timed out.
3202 Increase the reconnect interval to be longer than the connection
3203 timeout interval to avoid sending duplicate connections to
3207 Frequency : echo_client brw_test command
3209 Description: fix echo_client to work with OST preallocated code
3210 Details : OST preallocation code (5137) didn't take echo_client IO path
3211 into account: echo_client calls filter methods outside of any
3212 OST thread and, hence, there is no per-thread preallocated
3213 pages and buffers to use. Solution: hijack pga pages for IO. As
3214 a byproduct, this avoids unnecessary data copying.
3218 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3219 Description: Client can oops in mdc_commit_close() after open replay
3220 Details : It was possible for the MDS to return an open request with no
3221 transaction number in mds_finish_transno() if the client was
3222 evicted, but without actually returning an error. Clients
3223 would later try to replay that open and may trip an assertion
3224 Simplify the client close codepath, and always return an error
3225 from the MDS in case the open is not successful.
3228 Frequency : rare, 2.6 OSTs only
3230 Description: OST may deadlock under high load on fragmented files
3231 Details : If there was a heavy load and highly-fragmented OST filesystems
3232 it was possible to have all the OST threads deadlock waiting on
3233 allocation of biovecs, because the biovecs were not released
3234 until the entire RPC IO was completed. Instead, release biovecs
3235 as soon as they are complete to ensure forward IO progress.
3237 Severity : enhancement
3239 Description: Support for specifying external journal device at mount
3240 Details : If an OST or MDS device is formatted with an external journal
3241 device, this device major/minor is stored in the ext3 superblock
3242 and may not be valid for failover. Allow detecting and
3243 specifying the external journal at mount time.
3248 Description: Mounting an MDS with pending unlinked files may cause oops
3249 Details : target_finish_recovery() calls mds_postrecov() which returned
3250 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3251 considers this an error and immediately begins cleaning up the
3252 lov, just after starting the mds_lov process
3254 Severity : enhancement
3256 Description: Implement 'lfs df' to report actual free space on per-OST basis
3257 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3258 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3259 usage of objects; '-h' to report in human readable format.
3261 ------------------------------------------------------------------------------
3263 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3270 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3271 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3272 at startup during orphan unlinking. Instead, skip buddy update
3273 before setup, as it will be regenerated after recovery is complete.
3278 Description: If an OST is inactive, its locks might reference stale inodes.
3279 Details : lov_change_cbdata() must iterate over all namespaces, even if
3280 they are inactive to clear inode references from the lock.
3282 Severity : enhancement
3283 Frequency : occasional, if non-standard max_dirty_mb used
3285 Description: Client will block write RPCs if not enough grant
3286 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3287 then the client will block writes while waiting for another RPC
3288 to complete instead of consuming its dirty limit. With change
3289 we get improved performance when max_dirty_mb is small.
3291 Severity : enhancement
3292 Bugzilla : 3389, 6253
3293 Description: Add support for supplementary groups on the MDS.
3294 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3295 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3296 lookups for user supplementary groups into a cache.
3300 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3301 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3302 for liblustre O_EXCL is enforced by MDS.
3306 Description: readdir never returns NULL in liblustre.
3307 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3308 dentry in struct dirent.
3312 Frequency : liblustre only, depends on application IO pattern
3313 Description: liblustre clients evicted if not contacting servers
3314 Details : Don't put liblustre clients into the ping_evictor list, so
3315 they will not be evicted by the pinger ever.
3317 Severity : enhancement
3319 Description: Add ability to evict clients by NID from MDS.
3320 Details : By echoing "nid:$NID" string into
3321 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3322 $NID would be instantly evicted from this MDS and from all active
3323 OSTs connected to it.
3327 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3328 Details : lookup_it_finish() used to query file size from OSTs that was not
3333 Description: service threads change working directory to that of init
3334 Details : Starting lustre service threads may pin the working directory
3335 of the parent thread, making that filesystem busy. Threads
3336 now change to the working directory of init to avoid this.
3340 Frequency : during shutdown only
3341 Description: shutdown with a failed MDS or OST can cause unmount to hang
3342 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3347 Frequency : 2.6 only
3348 Description: chmod/chown may include an extra supplementary group
3349 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3350 supplementary group and if none is specified this is used.
3353 Bugzilla : 5479 (6816)
3355 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3356 Details : If lookup is in progress on a file that is unlinked we might try
3357 to revalidate the inode and fail in revalidate after lookup is
3358 complete and ll_file_open() enqueues the open again but
3359 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3362 Frequency : always, if lconf --abort_recovery used
3364 Description: lconf --abort_recovery fails with 'Operation not supported'
3365 Details : lconf was attempting to abort recovery on the MDT device and not
3368 ------------------------------------------------------------------------------
3370 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3375 Frequency : rare (only unsupported configurations with a node running as an
3377 Bugzilla : 6514, 5137
3378 Description: Mounting a Lustre file system on a node running as an OST could
3380 Details : OSTs now preallocates memory needed to write out data at
3381 startup, instead of when needed, to avoid having to
3382 allocate memory in possibly low memory situations.
3383 Specifically, if the file system is mounted on on OST,
3384 memory pressure could force it to try to write out data,
3385 which it needed to allocate memory to do. Due to the low
3386 memory, it would be unable to do so and the node would
3387 become unresponsive.
3389 Severity : enhancement
3391 Description: Addition of lconf --service command line option
3392 Details : lconf now accepts a '--service <arg>' option, which is
3393 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3395 Severity : enhancement
3397 Description: Failover mode is now the default for OSTs.
3398 Details : By default, OSTs will now run in failover mode. To return to
3399 the old behaviour, add '--failout' to the lmc line for OSTs.
3401 Severity : enhancement
3403 Description: Health checks are now provided for MDS and OSTs
3404 Details : Additional detailed health check information on MSD and OSTs
3405 is now provided through the procfs health_check value.
3408 Frequency : occasional, depends on IO load
3410 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3411 after numerous create/delete cycles
3412 Details : The ext3 inode allocation policy would not allocate new inodes
3413 very well on the OSTs because there are no new directories
3414 being created. Instead we look for groups with free space if
3415 the parent directories are nearly full.
3420 Description: Network or server problems during mount may cause partially
3421 mounted clients instead of returning an error.
3422 Details : The config llog parsing code may overwrite the error return
3423 code during mount error handling, returning success instead
3429 Description: MDS can fail to allocate large reply buffers
3430 Details : After long uptimes the MDS can fail to allocate large reply
3431 buffers (e.g. zconf client mount config records) due to memory
3432 fragmentation or consumption by the buffer cache. Preallocate
3433 some large reply buffers so that these replies can be sent even
3434 under memory pressure.
3438 Frequency : rare (liblustre)
3439 Description: fsx running with liblustre complained that using truncate() to
3440 extend the file doesn't work. This patch corrects that issue.
3441 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3442 ATTR_SIZE and lsm use in llu_setattr_raw.
3446 Frequency : rare, only 2.6 kernels
3447 Description: Unusual file access patterns on the MDS may result in inode
3448 data being lost in very rare circumstances.
3449 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3450 code in the 2.6 kernel caused the inode and/or EA data not to
3451 be read from disk, causing single-file corruption.
3455 Frequency : rare, only 2.6 filesystems using extents
3456 Description: Heavy concurrent write and delete load may cause data corruption.
3457 Details : It was possible under high-load situations to have an extent
3458 metadata block in the block device cache from a just-unlinked
3459 file overwrite a newly-allocated data block. We now unmap any
3460 metadata buffers that alias just-allocated data blocks.
3464 Frequency : filesystems with default stripe_count larger than 77
3465 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3466 Details : lconf specifies an inode size of 4096 bytes when the default
3467 stripe_count is larger than 77. This conflicts with the default
3468 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3469 this case to avoid pinning too much memory for large EAs.
3471 ------------------------------------------------------------------------------
3473 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3478 Frequency : rare (extremely heavy IO load with hundreds of clients)
3480 Description: Client is evicted, gets IO error writing to file
3481 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3482 caused clients to be evicted for AST timeouts. The fixes in
3483 bug 5192 mean we no longer need to have such short AST timeouts
3484 so ldlm_timeout has been increased.
3487 Frequency : occasional during --force or --failover shutdown under load
3488 Bugzilla : 5949, 4834
3489 Description: Server oops/LBUG if stopped with --force or --failover under load
3490 Details : a collection of import/export refcount and cleanup ordering
3491 issues fixed for safer force cleanup
3494 Frequency : only filesystems larger than 120 OSTs
3495 Bugzilla : 5990, 6223
3496 Description: lfs getstripe would oops on a very large filesystem
3497 Details : lov_getconfig used kfree on vmalloc'd memory
3500 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3501 Bugzilla : 6242, 6243
3502 Description: reading from files that had been truncated to a non-zero size
3503 but never opened returned no data
3504 Details : ll_file_read() reads zeros from no-object files to EOF
3509 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3510 Details : There's little harm in aborting MDS/OSS recovery and letting it
3511 try again, so I removed the LASSERT and return an error instead.
3513 Severity : enhancement
3515 Description: New debugging infrastructure for tracking down data corruption
3516 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3517 (b) cover more of the client-side code path, and (c) try to narrow
3518 down where problems occurred
3522 Bugzilla : 3819, 4364, 4397, 6313
3523 Description: Racing close and eviction MDS could cause assertion in mds_close
3524 Details : It was possible to get multiple mfd references during close and
3525 client eviction, leading to one thread referencing a freed mfd.
3527 Severity: : enhancement
3528 Bugzilla : 3262, 6359
3529 Description: Attempts to reconnect to servers are now more aggressive.
3530 Details : This builds on the enhanced upcall-less recovery that was added
3531 in 1.4.2. When trying to reconnect to servers, clients will
3532 now try each server in the failover group every 10 seconds. By
3533 default, clients would previously try one server every 25 seconds.
3538 Description: After recovery, certain operations trigger a failed
3539 assertion on a client.
3540 Details : Failing over an mds, using lconf -d --failover, while a
3541 client was doing a readdir() call would cause the client to
3542 LBUG after recovery completed and the readdir() was resent.
3544 Severity : enhancement
3546 Description: Default groups are now added by lconf
3547 Details : You can now run lconf --group <servicename> without having to
3548 manually add groups with lmc.
3551 Frequency : occasional
3553 Description: Nodes with an elan id of 0 trigger a failed assertion
3556 Frequency : always when accessing e.g. tty/console device nodes
3558 Description: tty and some other devices nodes cannot be used on lustre
3559 Details : file's private_data field is used by device data and lustre
3560 values in there got lost. New field was added to struct file to
3561 store fs-specific private data.
3564 Frequency : when exporting Lustre via NFS
3566 Description: NFSD failed occasionally when looking up a path component
3567 Details : NFSD is looking up ".." which was broken in ext3 directories
3568 that had grown large enough to become hashed.
3571 Frequency : Clusters with multiple interfaces not on the same subnet
3573 Description: Nodes will repeatedly try to reconnect to an interface which it
3574 cannot reach and report an error to the log.
3575 Details : Extra peer list entries will be created by lconf with some peers
3576 unreachable. lconf now validates the peer before adding it.
3579 Frequency : Only if a default stripe is set on the filesystem root.
3581 Description: Setting a default stripe on the filesystem root prevented the
3582 filesystem from being remounted.
3583 Details : The client was sending extra request flags in the root getattr
3584 request and did not allocate a reply buffer for the dir EA.
3587 Frequency : occasional, higher if lots of files are accessed by one client
3588 Bugzilla : 6159, 6097
3589 Description: Client trips assertion regarding lsm mismatch/magic
3590 Details : While revalidating inodes the VFS looks up inodes with ifind()
3591 and in rare cases can find an inode that is being freed.
3592 The ll_test_inode() code will free the lsm during ifind()
3593 when it finds an existing inode and then the VFS later attaches
3594 this free lsm to a new inode.
3598 Bugzilla : 6422, 7030
3599 Description: MDS deadlock between mkdir and client eviction
3600 Details : Creating a new file via mkdir or mknod (starting a transaction
3601 and getting the ns lock) can deadlock with client eviction
3602 (gets ns lock and trying to finish a synchronous transaction).
3605 Frequency : occasional
3606 Description: While starting a server, the fsfilt_ext3 module could not be
3608 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3609 kernels. Previously, lconf would still use the ext3 name
3610 when trying to load modules. Now, it will correctly use
3611 ext3 on 2.4 and ldiskfs on 2.6.
3613 Severity : enhancement
3614 Description: The default stripe count has been changed to 1
3615 Details : The interpretation of the default stripe count (0, to lfs
3616 or lmc) has been changed to mean striping across a single
3617 OST, rather than all available. For general usage we have
3618 found a stripe count of 1 or 2 works best.
3620 Severity : enhancement
3621 Description: Add support for compiling against Cray portals.
3622 Details : Conditional compiling for some areas that are different
3626 Frequency : occasional
3627 Bugzilla : 6409, 6834
3628 Description: Creating files with an explicit stripe count may lead to
3629 a failed assertion on the MDS
3630 Details : If some OSTs are full or unavailable, creating files may
3631 trigger a failed assertion on the MDS. Now, Lustre will
3632 try to use other servers or return an error to the
3636 Frequency : occasional
3638 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3639 caused lock timeout during readahead (since 1.4.2).
3640 Details : Processes doing readahead might match a lock that hasn't been
3641 granted yet if there are overlapping and conflicting lock
3642 requests. The readahead process waits on ungranted lock
3643 (original lock is CBPENDING), while OST waits for that process
3644 to cancel CBPENDING read lock and eventually evicts client.
3646 Severity : enhancement
3648 Description: Initial enabling of flock support for clients
3649 Details : Implements fcntl advisory locking and file status functions.
3650 This feature is provided as an optional mount flag (default
3651 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3652 locking are implemented yet, and those that are are not guaranteed
3653 to be completely correct in production environments.
3654 mount -t lustre -o [flock|noflock] ...
3657 Frequency : occasional
3659 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3660 assertion in the ext3 JBD (journaling) layer.
3661 Details : The b_committed_data struct is protected by the big kernel lock
3662 in 2.4 kernels, serializing journal_commit_transaction() and
3663 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3664 access to this struct. In 2.6 kernels there is finer grained
3665 locking to improve SMP performance of the JBD layer.
3669 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3674 Description: merge and cleanup kernel patches.
3677 -----------------------------------------------------------------------------
3679 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3681 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3682 versions of Lustre. Please update all servers and clients to
3683 version 1.4.2 or later at the same time. You must also run
3684 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3685 to update the configuration logs.
3687 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3688 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3689 - swab llog records of type '0' so we get proper header size/idx (5861)
3690 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3691 - fix rename of one directory over another leaking an inode (5953)
3692 - avoid SetPageDirty on 2.6 (5981)
3693 - don't re-add just-being-destroyed locks to the waiting list (5653)
3694 - when creating new directories, inherit the parent's custom
3695 striping settings if present parent (3048)
3696 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3697 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3698 - don't hold client locks on temporary worklist from l_lru (5666)
3699 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3700 - automatically evict dead clients (5921)
3701 - Update file size properly in create+truncate+fstat case (6196)
3702 - Do not unhash mountpoint dentries, do not allow removal of
3704 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3705 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3706 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3707 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3708 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3709 - Correctly handle reads of files with no objects (6243)
3710 - lctl recover will also mark a device active if deactivate used (5933)
3712 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3713 - allow --write-conf on an MDS with different nettype than client (5619)
3714 - don't write config llogs to MDS for mounts not from that MDS (5617)
3715 - lconf should create multiple TCP connections from a client (5201)
3716 - init scripts are now turned off by default; run chkconfig --on
3717 lustre and chkconfig --on lustrefs to use them
3718 - upcalls are no longer needed for clients to recover to failover
3720 - add --abort-recovery option to lconf to abort recovery on device
3722 - add support for an arbitrary number of OSTs (3026)
3723 - Quota support protocol changes.
3724 - forward compatibility changes to wire structs (6007)
3725 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3726 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3727 - improved "lustre status" script
3728 - initialize blocksize for non-regular files (6062)
3729 - added --disable-server and --disable-client configure options (5782)
3730 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3731 - Vanilla 2.4.29 support
3732 - increase maximum number of obd devices to 520 (6242)
3733 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3734 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3735 SLES 9 2.6.5 kernels
3736 - limit stripes per file to 160 (the maximum EA size) (6093)
3738 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3741 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3742 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3743 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3744 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3745 - small ext3 extents cleanups and fixes (5733)
3746 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3747 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3748 - avoid needless client->OST connect, fix handle mismatch (5317)
3749 - fix DLM error path that led to out-of-sync client, long delays (5779)
3750 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3751 - fix several locking issues related to i_size (5492,5624,5654,5672)
3752 - don't move pending lock onto export if it is already evicted (5683)
3753 - fix kernel oops when creating .foo in unlinked directory (5548)
3754 - fix deadlock in obdfilter statistics vs. object create (5811)
3755 - use time_{before,after} to avoid timer jiffies wrap (5882)
3756 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3757 - Do not leak request if server was not able to process it (5154)
3758 - If mds_open unable to find parent dir, make that negative lookup(5154)
3759 - don't create new directories with extent-mapping (5909, 5936)
3761 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3762 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3763 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3764 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3765 - support for RHEL4 kernel on i686 (5773)
3766 - provide error messages when incompatible logs are encountered (5898)
3768 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3769 * version 1.4.0.10 (1.4.1 release candidate 1)
3771 - don't keep a lock reference when lock is not granted (4238)
3772 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3773 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3774 - drop import inflight refcount on signal_completed_replay error (5255)
3775 - unlock page after async write error during send (3677)
3776 - handle missing objects in filter_preprw_read properly (5265)
3777 - no transno return for symlink open, don't save no-trasno open (3440)
3778 - don't try to complete elan receive that already failed (4012)
3779 - free RPC server reply state on error (5406)
3780 - clean up thread from ptlrpc_start_thread() on error (5160)
3781 - readahead could read extra page into cache that wasn't ejected (5388)
3782 - prevent races in class_attach/setup/cleanup/detach (5260)
3783 - don't dereference de->d_inode after l_dput of de (5458)
3784 - use "int" for stripe value returned from lock_to_stripe (5544)
3785 - mballoc allocation and error-checking fixes in 2.6 (5504)
3786 - block device patches to fix I/O request sizes in 2.6 (5482)
3787 - look up hostnames for IB nals (5602)
3788 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3789 - don't start multiple acceptors for the same port (5277)
3790 - fix incorrect LASSERT in mds_getattr_name (5635)
3791 - export a proc file for general "ping" checking (5628)
3792 - fix "lfs check" to not block when the MDS is down (5628)
3794 - service request history (4965)
3795 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3796 - create an "evict_client" /proc entry on OSTs, like the MDS has
3797 - fix mount usage message, return errors per mount(8) (5168)
3798 - change grep [] to grep "[]" in tests so they work in more UMLs
3799 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3800 - remove ancient LOV_MAGIC_V0 EA support (5047)
3801 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3802 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3803 - allow readahead allocations to fail when low on memory (5383)
3804 - mmap locking landed again, after considerable improvement (2828)
3805 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3807 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3810 - send OST transaction number in read/write reply to free req (4966)
3811 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3812 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3813 - the watchdog thread now runs as interruptible (5246)
3814 - flock/lockf fixes (but it's still disabled, pending 5135)
3815 - don't use EXT3 constants in llite code (5094)
3816 - memory shortage at startup could cause assertion (5176)
3818 - reorganization of lov code
3819 - single portals codebase
3821 - add extents/mballoc support (5025)
3822 - direct I/O reads in the obdfilter (4048)
3823 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3825 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3828 - send OST transaction number in read/write reply to free req (4966)
3829 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3830 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3831 - the watchdog thread now runs as interruptible (5246)
3832 - handle missing objects in filter_preprw_read properly (5265)
3833 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3834 - drop import inflight refcount on signal_completed_replay error (5255)
3835 - unlock page after async write error during send (3677)
3836 - return original error code on reconstructed replies (3761)
3837 - no transno return for symlink open, don't save no-trasno open (3440)
3839 - add pid to ldlm debugging output (4922)
3840 - bump the watchdog timeouts -- we can't handle 30sec yet
3841 - extra debugging for orphan dentry/inode bug (5259)
3843 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3846 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3847 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3848 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3849 - fix NULL dereference in /proc/sys/portals/routes (4827)
3850 - allow failed mdc_close() operations to be interrupted (4561)
3851 - stop precreate on OST before MDS would time out on it (4778)
3852 - don't send partial-page writes before EOF from client (4410)
3853 - discard client grant for sub-page writes on large-page clients (4520)
3854 - don't free dentries not owned by NFS code, check generation (4806)
3855 - fix lsm leak if mds_create_objects() fails (4801)
3856 - limit debug_daemon file size, always print CERROR messages (4789)
3857 - use transno after validating reply (3892)
3858 - process timed out requests if import state changes (3754)
3859 - update mtime on OST during writes, return in glimpse (4829)
3860 - add mkfsoptions to LDAP (4679)
3861 - use ->max_readahead method instead of zapping global ra (5039)
3862 - don't interrupt __l_wait_event() during strace
3864 - add software watchdogs to catch hung threads quickly (4941)
3865 - make lustrefs init script start after nfs is mounted
3866 - fix CWARN/ERROR duplication (4930)
3867 - return async write errors to application if possible (2248)
3868 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3869 - print NAL number in %x format (4645)
3870 - update barely-supported suse-2.4.21-171 series (4842)
3871 - support for sles 9 %post scripts
3872 - support for building 2.6 kernel-source packages
3873 - support for sles km_* packages
3875 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3878 - ignore -ENOENT errors in osc_destroy (3639)
3879 - notify osc create thread that OSC is being cleaned up (4600)
3880 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3881 - reconstruct ost_handle() like mds_handle() (4657)
3882 - create a new thread to do import eviction to avoid deadlock (3969)
3883 - let lconf resolve symlinked-to devices (4629)
3884 - don't unlink "objects" from directory with default EA (4554)
3885 - hold socknal file ref over connect in case target is down (4394)
3886 - allow more than 32000 subdirectories in a single directory (3244)
3887 - fix blocks count for O_DIRECT writes (3751)
3888 - OST returns ENOSPC from object create when no space left (4539)
3889 - don't send truncate RPC if file size isn't changing (4410)
3890 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3891 - bind to privileged port in socknal and tcpnal (3689)
3893 - rate limit CERROR/CWARN console message to avoid overload (4519)
3894 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3895 - basic mmap support (3918)
3896 - kernel patch series update from b1_4 (4711)
3898 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3901 - avoid crash during MDS cleanup with OST shut down (2775)
3902 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3903 - don't use bad inodes on the MDS (3744)
3904 - dynamic object preallocation to improve recovery speed (4236)
3905 - don't hold spinlock over lock dumping or change debug flags (4401)
3906 - don't zero obd_dev when it is force cleaned (3651)
3907 - print grants to console if they go negative (4431)
3908 - "lctl deactivate" will stop automatic recovery attempts (3406)
3909 - look for existing locks in ldlm_handle_enqueue() (3764)
3910 - don't resolve lock handle twice in recovery avoiding race (4401)
3911 - revalidate should check working dir is a directory (4134)
3913 - don't always mark "slow" obdfilter messages as errors (4418)
3915 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3918 - don't close LustreDB during write_conf until it is done (3860)
3919 - fix typo in lconf for_each_profile (3821)
3920 - allow dumping logs from multiple threads at one time (3820)
3921 - don't allow multiple threads in OSC recovery (3812)
3922 - fix debug_size parameters (3864)
3923 - fix mds_postrecov to initialize import for llog ctxt (3121)
3924 - replace config semaphore with spinlock (3306)
3925 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3926 - don't allow enqueue to complete on a destroyed export (3822)
3927 - down write_lock before checking llog header bitmap (3825)
3928 - recover from lock replay timeout (3764)
3929 - up llog sem before sending rpc (3652)
3930 - reduce ns lock hold times when setting kms (3267)
3931 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3932 - fix NULL deref and obd_dev leak on setup error (3312)
3933 - replace some LBUG about llog ops with error handling (3841)
3934 - don't match INVALID dentries from d_lookup and spin (3784)
3935 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3936 - fix invalid assertion in ptlrpc_set_wait (3880)
3938 - add libwrap support for the TCP acceptor (3996)
3939 - add /proc/sys/portals/routes for non-root route listing (3994)
3940 - allow setting MDS UUID in .xml (2580)
3941 - print the stack of a process that LBUGs (4228)
3943 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3946 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3947 - make sure to unset replay flag from failed open requests (3440)
3948 - if default stripe count is 0, use OST count for inode size (3636)
3949 - update parent mtime/ctime on client for create/unlink (2611)
3950 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3951 - free recovery state on server during a forced cleanup (3571)
3952 - unregister_reply for resent reqs (3063)
3953 - loop back devices mounting and status check on 2.6 (3563)
3954 - fix resource-creation race that can provoke i_size == 0 (3513)
3955 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3956 - more debugging for page-accounting assertion (3746)
3957 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3958 - avoid deadlock after precreation failure (3758)
3959 - fix race and lock order deadlock in orphan handling (3450, 3750)
3960 - add validity checks when grabbing inodes from l_ast_data (3599)
3962 - add /proc/.../recovery_status to obdfilter (3428)
3963 - lightweight CDEBUG infrastructure, debug daemon (3668)
3964 - change default OSC RPC parameters to be better on small clusters
3965 - turn off OST read cache for files smaller than 32MB
3966 - install man pages and include them in rpms (3100)
3967 - add new init script for (un)mounting lustre filesystems (2593)
3968 - run chkconfig in %post for init scripts (3701)
3969 - drop scimac NAL (unmaintained)
3971 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3974 - clean kiobufs before and after use (3485)
3975 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3976 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3977 - change init script to fail more clearly if not run as root (1528)
3978 - allow clients to reconnect during replay (1742)
3979 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3980 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3981 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3982 - fix cond_resched (3554)
3983 - only evict unfinished clients after recovery (3515)
3984 - allow bulk resend, prevent data loss (3570)
3985 - dynamic ptlrpc request buffer allocation (2102)
3986 - don't allow unlinking open directory if it isn't empty (2904)
3987 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3988 - remove extraneous obd dereference causing LASSERT failure (3334)
3989 - don't use get_cycles() when creating temp. files on the mds (3156)
3990 - hold i_sem when setting i_size in ll_extent_lock() (3564)
3991 - handle EEXIST for set-stripe, set proper directory name (3336)
3993 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
3994 - fix ksocknal_fmb_callback() error messages (2918)
3996 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
3999 - don't copy lvb into (possibly NULL) reply on error (2983)
4000 - don't deref dentry after dput, don't free lvb on error (2922)
4001 - use the kms to determine writeback rpc length (2947)
4002 - increment oti_logcookies when osc is inactive (2948)
4003 - update client's i_blocks count via lvb messages (2543)
4004 - handle intent open/close of special files properly (1557)
4005 - mount MDS with errors=remount-ro, like obdfilter (2009)
4006 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4007 - don't use cancelling-locks' kms values (2947)
4008 - use highest lock extent for kms, not last one (2925)
4009 - don't dereference ERR_PTR() dentry in error handling path (3107)
4010 - fix thread race in portals_debug_dumplog() (3122)
4011 - create lprocfs device entries at setup instead of at attach (1519)
4012 - common AST error handler, don't evict client on completion race (3145)
4013 - zero nameidata in detach_mnt in 2.6 (3118)
4014 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4015 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4016 - initialize RPC timeout timer earlier for 2.6 (3219)
4017 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4018 - print nal/nid for unknown nid (3258)
4019 - additional checks for oscc recovery before doing precreate (3284)
4020 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4021 - don't crash in mdc_close for bad permissions on open (3285)
4022 - zero i_rdev for non-device files (3147)
4023 - clear page->private before handing to FS, better assertion (3119)
4024 - tune the read pipeline (3236)
4025 - fix incorrect decref of invalidated dentry (2350)
4026 - provide read-ahead stats and refine rpc in flight stats (3328)
4027 - don't hold journal transaction open across create RPC (3313)
4028 - update atime on MDS at close time (3265)
4029 - close LDAP connection when recovering to avoid server load (3315)
4030 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4031 - don't leak open file on MDS after open resend (3325)
4032 - serialize filter_precreate and filter_destroy_precreated (3329)
4033 - loop device shouldn't call sync_dev() for nul device (3092)
4034 - clear page cache after eviction (2766)
4035 - resynchronize MDS->OST in background (2824)
4036 - refuse to mount the same filesystem twice on same mountpoint (3394)
4037 - allow llmount to create routes for mounting behind routers (3320)
4038 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4039 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4040 - fix rare problem with rename on htree directories (3417)
4042 - allow default OST striping configuration per directory (1414)
4043 - fix compilation for qswnal for 2.6 kernels (3125)
4044 - increase maximum number of MDS request buffers for large systems
4045 - change liblustreapi to be useful for external progs like lfsck (3098)
4046 - increase local configuration timeout for slow disks (3353)
4047 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4049 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4052 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4053 - don't overwrite extent policy data in reply if lock was blocked (2901)
4054 - drop filter export grants atomically with removal from device (2663)
4055 - del obd_self_export from work_list in class_disconnect_exports (2908)
4056 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4057 - swab reply message in mdc_close, other PPC fixes (2464)
4058 - fix destroying of named logs (2325)
4059 - overwrite old logs when running lconf --write_conf (2264)
4060 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4061 - fix race in target_handle_connect (2898)
4062 - mds_reint_create() should take same inode create lock (2926)
4063 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4064 - don't close files for self_export to avoid uninitialized obd (2936)
4065 - allow MDS with the same name as client node (2939)
4066 - hold dentry reference for closed log files for unlink (2325)
4067 - reserve space for all logs during transactions (2059)
4068 - don't evict page beyond end of stripe extent (2925)
4069 - don't oops on a deleted current working directory (2399)
4070 - handle hard links to targets without a parent properly (2517)
4071 - don't dereference NULL lock when racing during eviction (2867)
4072 - don't grow lock extents when lots of conflicting locks (2919)
4074 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4077 - account for cache space usage on clients to avoid data loss (974)
4078 - lfsck support in lustre kernel code (2349)
4079 - reduce journal credits needed for BRW writes (2370)
4080 - orphan handling to avoid losing space on client/server crashes
4081 - ptlrpcd can be blocked, stopping ALL progress (2477)
4082 - use lock value blocks to assist in proper KMS, faster stat (1021)
4083 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4084 - recovery for initial connections (2355)
4085 - fixes for mds_cleanup_orphans (1934)
4086 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4087 - block all file creations until orphan recovery completes (1901)
4088 - client remove rq_connection from request struct (2423)
4089 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4090 - recovery timer race (2670)
4091 - mdc_close recovey bug (2532)
4092 - ptlrpc cleanup bug (2710)
4093 - mds timeout on local locks (2588)
4094 - namespace lock held during RPCs (2431)
4095 - handle interrupted sync write properly (2503)
4096 - don't try to handle a message that hasn't been replied to (2699)
4097 - client assert failure during cleanup after abort recovery (2701)
4098 - leak mdc device after failed mount (2712)
4099 - ptlrpc_check_set allows timedout requests to complete (2714)
4100 - wait for inflight reqs when ptlrpcd finishes (2710)
4101 - make sure unregistered services are removed from the srv_list
4102 - reset bulk XID's when resending them (caught by 1138 test)
4103 - unregister_bulk after timeout
4104 - fix lconf error (2694)
4105 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4106 - readahead locks pages, leaves pending causing memory pressure (2673)
4107 - increase OST request buffers to 4096 on large machines (2729)
4108 - fix up permission of existing directories in simple_mkdir (2661)
4109 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4110 - don't assign transno to errored transactions (2742)
4111 - don't delete objects on OST if given a bogus objid from MDS (2751)
4112 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4113 - if rq_no_resend, then timeout request after recovery (2432)
4114 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4115 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4116 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4117 - MDS recovery shouldn't delete objects that it has given out (2730)
4118 - if enqueue arrives after completion, don't clobber LVB (2819)
4119 - don't unlock pages twice when trigger_group_io returns error (2814)
4120 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4121 - don't write pages to disk if there was an error (1450)
4122 - don't ping imports that have recovery disabled (2676)
4123 - take buffered bytes into account when balancing socknal conn (2817)
4124 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4125 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4126 - remove little-endian swabbing of llog records (1987)
4127 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4128 - retry reposting request buffers if they fail (1191)
4129 - grow extent at grant time to avoid granting a revoked lock (2809)
4130 - lock revoke doesn't evict page if covered by a second lock (2765)
4131 - disable VM readahead to avoid reading outside lock extents (2805)
4133 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4134 - updated kernel patches for hp-2.4.20 kernel (2681)
4136 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4139 - fix truncated write corruption (2366)
4140 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4142 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4143 - ensure len(uuid) < 37 in lmc (1171)
4144 - fix ia64 OOPS in llog_test (2255)
4145 - zero end of page at obdfilter for partial page writes (2648)
4146 - don't leave stale dentries around after renames (bug 2428)
4147 - fix timeouts when evicting a client with a single lock held (2642)
4148 - set deadline for the initial HELLO message to drain (2634)
4149 - print out dotted-quad IP addresses in the socknal (2302)
4151 - additional debugging for MDS client eviction problem (2443)
4152 - fix mkfsoptions support for osts (2603, 2604)
4154 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4157 - add series for the vanilla 2.6.0 kernel
4158 - add series for the vanilla 2.4.24 kernel
4159 - add series for a cray x86/64 UL kernel drop
4160 - fix xattr patches for the vanilla 2.4.19 series
4162 - generate true UUIDs in lmc (1171)
4163 - have portals stack dumping break in UML (2466)
4164 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4165 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4166 - wake callback waiting threads on client eviction (2460)
4167 - Add --ptldebug and --subsystem to lmc (1719)
4168 - update assertion to allow safe interrupt allocation
4169 - set rq_no_resend for cancel requests (2432)
4170 - recalculate ptlrpcd timeout after resend (2494)
4171 - call vfs_rmdir when removing pending directories (2368)
4172 - fix renaming a file to itself (2429)
4173 - lmc creates a default one-stripe lov (2454)
4174 - expand procfs space to handle large clusters (2326)
4175 - increase UML stack to avoid overflow
4176 - update lconf's list of debug and subsystem masks
4177 - fix lfs find --obd (2510)
4178 - /proc tunable for disabling filter read caching (2591)
4179 - stop rpm packages from altering slapd.conf (2301)
4180 - disable nagle in the socknal under 0conf (2578)
4181 - choose mds inode size based on stripe count (2572)
4182 - fix kernel-source rpm problems (2516)
4184 - add --disable-doc to avoid pdf generation (2421)
4185 - update documentation, tests, type-os, comments
4186 - avoid format warnings on ia64
4187 - remove the TOE NAL
4188 - tiny code cleanups by removing unused fields
4190 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4193 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4194 - remove the most blatant lies from BUILDING file (2371)
4195 - change default debug level to reasonable production setting
4196 - reduce client side cache size to reduce cache flush time
4197 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4198 - make TCP zerocopy and pinger support enabled by default (2476)
4199 - sync writes completed after process exits caused crashes (2319)
4200 - maintain correct mount count on the MDS (2356)
4201 - backout 1557, because 2316 wasn't really fixed
4202 - better file I/O statistics gathering in /proc
4203 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4204 - another kernel patch to fix zero-copy TCP function export
4205 - don't take duplicate lock when processing re-sent getattr (2420)
4206 - lctl uses obd_self_export instead of creating new conn (2353)
4207 - MDS/OST recovery case which requires object creation asserted (2425)
4208 - move lfs from /usr/sbin to /usr/bin in packages
4209 - fix race between mds_client_add and mds_client_free (2417)
4210 - use kmalloc instead of slabs in portals (2430)
4211 - don't create duplicate records when a failover MDS is present (2442)
4212 - remove unnecessary mount age check (2332)
4213 - don't remove directory inodes from locks prematurely (2451)
4214 - don't break if MDS service name is the same as hostname (2103)
4215 - fix races in client write RPC generation when cache full (2482)
4217 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4220 - remove now-unused request->rq_obd (278)
4221 - if an allocation fails, print out how much memory we've used (1933)
4222 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4223 - add the "configurable stack size" patch to most series files (1256)
4224 - ability to write large log records, for 100+ OST configs (2306)
4225 - fix NULL deref when filter_prep fails (2314)
4226 - fix operator precedence error in filter_sync
4227 - dynamic allocation of socknal TX descriptors (2315)
4228 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4229 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4230 - ensure that CWARN messages in Portals always get to the syslog
4231 - __init/__exit are not for prototype decls (ldlm_init/exit)
4232 - x86-64 compile warning fixes
4233 - fix gateway LMC keyword conflict (2318)
4234 - fix MDS lock inversions in getattr/reint paths (1844)
4235 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4236 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4237 - DLM race condition prevented some lock evictions (2328)
4238 - ENOMEM detection and retry on socknal sends (2230)
4239 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4240 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4241 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4242 - better error messages when a client is rejected during recovery (1505)
4243 - avoid cancelling locks which were never granted, after failure (2330)
4244 - fix i_sem/journal inversion in mds_client_add (2333)
4245 - fix truncate/getattr lock cycle deadlock (2334)
4246 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4247 - fix two rare exit paths which could leak an l_lock() ref (2321)
4248 - fencepost error in MDS/OST orphan recovery (2226)
4249 - make log record alignment 8 bytes (1988)
4250 - lstripe now fails when requested offset > ost_count (2237)
4251 - ensure that all kernel series have a complete list.h (1607)
4252 - fix crashes in special-file operations (2316)
4253 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4255 - allow configurable automake binary, for testing new versions
4256 - small update to the lfs documentation
4258 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4260 * fix negative export reference count in fsfilt_sync (2312)
4262 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4263 * release candidate 0.9.1
4265 - orphans are moved into the PENDING directory for possible recovery
4266 - replayed opens now open by fid for orphan/rename safety (1042)
4267 - last close of an orphan inode generates a transno (683)
4268 - chdir() and mount() now pin the directory entry (1020)
4269 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4270 - discard very old requests without processing them (1502)
4271 - remove some common, well-understood CERRORs (1505)
4272 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4273 - clear "grant" flags in OST replies until OST grant code lands (1644)
4274 - fix read performance by not clobbering i_blksize on client (1598)
4275 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4276 - make LRU size a /proc tunable, clears locks when reduced (707)
4277 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4278 - server locks take references on exports now (1558)
4279 - build fixes for 2.4.20-rh trees (1663)
4280 - return an error from lov_create if all OSCs are inactive (1751)
4281 - fix import levels when a reconnect happens without a timeout (1597)
4282 - exit early from mds_open if we get a lookup error (1749)
4283 - partial page read at EOF wouldn't wait for disk before sending (1642)
4284 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4285 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4286 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4287 - write/truncate lock inversion (1639)
4288 - Don't auto-load obdclass, portals modules during cleanup (1495)
4289 - fix timestamps from jumping to "now" (1763)
4290 - extra journal assertions (1648)
4291 - add an extra multiunlink test (1771)
4292 - fix read_record/write_record API (1776)
4293 - fix leak of offset_extent, possible incorrect i_size later (1772)
4294 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4295 - Debugging for the kqswnal_get_idle_tx problems (1820)
4296 - Allow recovery to be attempted multiple times (1536)
4297 - Write out MDS last_rcvd file after it is first created (1600)
4298 - Fix tx_descriptor leak in failed transmit situations (1827)
4299 - ext3 journaling fixes for assertion failure after IO error (1871)
4300 - class_export_put() on freed export after completion AST error (1896)
4301 - Fix revalidate looping in VFS (1322)
4302 - Don't access a freed export during MDS_REINT timeout (1521)
4303 - Add open-unlink recovery support on the MDS (1673,1764)
4304 - Return an error if no MDS data was read from last_rcvd (1946)
4305 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4306 - Don't setup a disk device that doesn't match exported UUID (317)
4307 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4308 - avoid committing NULL handle in force close
4309 - local.sh is now a one-stripe LOV configuration
4310 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4311 - fix ext3 htree duplicate directory entry corruption (1516)
4312 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4313 - update to new LOV EA format (2097)
4314 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4316 2003-06-15 Phil Schwan <phil@clusterfs.com>
4319 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4320 - per-import/export recovery handling (958, 931, 959)
4321 - multiple last-rcvd slots, for serving multiple FSes (949)
4322 - connections are again shared between multiple imp/exports (963, 964)
4323 - "umount -f" would hang if any requests needed to be sent (393, 978)
4324 - avoid pinning large req buffer by copying for queued messages (989)
4325 - add "uuid" to "lctl device" command to help upcalls (991)
4326 - "open" RPCs with transnos would confuse recovery counters (1037)
4327 - do proper endian conversion of all wire messages (288, 340, 891)
4328 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4329 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4330 - fix ptlrpc_connection leak in target_handle_connect (1174)
4331 - fix import refcounting bug in OST and MDS cleanup (1134)
4332 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4333 - fix very unlikely obd_types race condition (501)
4334 - remove osc_open hack for echo_client (1187)
4335 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4336 - a failure in read_inode2 leads to deadlock (1139)
4337 - cancel ack-locks as soon as transaction is committed (1072)
4338 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4339 - make sure to commitrw after any preprw to avoid deadlock (1162)
4340 - failing to execute a file in a lustre FS would lock inode (1203)
4341 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4342 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4343 - obd_uuid2tgt would walk off the end of the list (1255)
4344 - on IA64 the osc would give portals incorrect bulk size (1258)
4345 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4346 - fix lock inversion caused by new llite matching code (1282)
4347 - limit the number of dirty pages on a client to 10MB (1286)
4348 - timed out locks were not being corrected cancelled (1289)
4349 - fix O_DIRECT above 4GB on IA-32 (1292)
4350 * major user-visible changes
4351 - fail out/fail over policy now controlled by the upcall (993)
4353 - add OBD_PING to check server availability and failure (954)
4354 - lustre messages are now sent in sending host order (288, 340, 891)
4355 - add eadatalen to MDS getattr reply (340)
4356 - OST read replies may contain second buffer, with per-page status (593)
4358 2003-03-11 Phil Schwan <phil@clusterfs.com>
4361 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4362 - failed open()s could cause deadlock; fixed (867, 869)
4363 - stop cancelling OST locks when files are closed (481)
4364 - overlapping XID spaces caused network corruption (851, 853)
4365 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4366 - setattr_raw added, to do single-RPC, server-side setattrs
4367 - lmc/lconf syntax change for OST UUIDs
4368 - fix crashy race condition between ptlrpc_free_req and osc_close
4369 - don't use request in mdc_enqueue if we hit a timeout (889)
4370 - don't set the inode i_size for regular files from the MDS (896)
4371 - handle out of order completion AST (842)
4372 - don't LBUG if a lock request times out after receiving AST (913)
4373 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4374 - if a bad lock AST arrives, send an error instead of dropping entirely
4375 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4376 - fix leak in bulk IO when only partially completed (899, 900, 926)
4377 - fix O_DIRECT for ia64 (55)
4378 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4379 - C-z after timeout could hang a process forever; fixed (977)
4381 - client-side I/O cache (678, 924, 929, 941, 970)
4383 - READPAGE and SETATTRs which don't take server-side locks get
4386 2003-02-11 Phil Schwan <phil@clusterfs.com>
4389 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4390 - Fix fsfilt_extN_readpage() to read a full page of directory
4391 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4392 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4393 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4394 - Most intent operations were removed, in favour of a new RPC mode
4395 that does a single RPC to the server and bypasses most of the VFS
4396 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4397 - Aggressively cancel local locks on DLM servers
4398 - mds_reint_unlink sends EA to the client if it's the last nlink.
4399 client uses that EA to unlink OST objects.
4400 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4401 - recursive symlinks were fixed (439)
4402 - fixed NULL deref in DEBUG_REQ
4403 - filter_update_lastobjid no longer calls sync, which annoyed extN
4404 - fixed multi-client small-writes to a single file problem (445)
4405 - fixed mtime updates during file writes (607)
4406 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4407 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4408 - fixed Linux OST size reporting problem (444, 656)
4409 - OST now updates object mtime with writes or setattr (607, 619)
4410 - client verifies file size before zeroing page past EOF (445)
4411 - OST now writes last allocated objid to disk with allocation (108)
4412 - LOV on echo now works (409)
4414 - mds_reint_unlink sends a new buffer, with the EA included. this
4415 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4416 set if a regular file was being unlinked, and it was the last link
4417 - use PtlGet from the target for bulk writes (315)
4418 - OST now updates object mtime with writes or setattr (607, 619)
4419 - LDLM now has a grant-time callback to revalidate locked items, if
4421 - Many MDS operations were reorganized to combat race conditions
4423 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4424 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4426 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4429 - Fully reactivate OST imports after reconnection (512, others)
4430 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4431 - More graceful error handling for truncating on dead OST (515)
4432 - Don't error out unless we're actually accessing dead stripes (474)
4433 - Fix garbage sizes when stripes are missing (410)
4434 - LRU counters were broken, causing constant lock purge (433, 432)
4435 - garbage on read from stripes with failed OSTs (441)
4436 - mark OSCs as active before reconnecting during recovery (438)
4437 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4438 - lfind did not preserve OST order in output (443)
4439 - symlinks cause hung clients, incorrect data (439)
4440 - stop dereferencing request after dropping refcount (457)
4441 - don't LASSERT(spin_is_locked) on non-SMP (455)
4442 - fixes for many rename() bugs
4443 - fstat didn't correctly synchronize attributes (399)
4444 - server must handle lock cancellation during blocking AST prep (487)
4445 - bulk descriptors were free()d too soon (511)
4446 - fix paths in lconf, which would load incorrect modules (451, 507)
4447 - fix confusing lconf 'host not found' error message (386)
4448 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4449 - fix race condition in mdc_blocking_ast() for inode access (526)
4450 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4451 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4452 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4453 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4454 - fix lstripe in several ways (kernel side) (527)
4455 - fix request leak in ldlm_cli_enqueue (262)
4456 - incorrect OSC was marked inactive after OST failure
4457 - call mds_fs_cleanup before unmounting filesystem (524)
4458 - fix races between taking ns_lock and ldlm_lock_change_resource
4459 - fix races updating LOV export open file list
4460 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4461 - fix recovery NULL deref in ldlm_cli_cancel_unused
4462 - fix some DLM races by using new hash table for lock handles (419)
4463 - permit the client to specify desired inodes, at replay
4464 - duplicate requests when we queue them for replay reintegration
4465 - fix last_rcvd offset calculation
4466 - sync after each recovered transaction, so we always make progress
4467 - never, not always, ERESTART requests without transnos
4468 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4470 - skip replay if the MDS says that the client is already connected
4471 - don't check for a recovery-enabled export to match lctl's UUID
4472 - don't INC_USE_COUNT for phantom exports
4473 - don't crash when cleaning up phantom exports (567)
4474 - don't double-finish or set replay data for errored mdc_open requests
4475 - abort requests when they time out, so we don't get old replies
4476 - send/receive replies for AST messages again
4477 - if the client says that it doesn't have the lock, cancel it on the
4479 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4480 mark it as destroyed, it will all work out when decref is called
4481 - fix module use counts (22, 581)
4483 - ASTs now expect a reply (server cancels lock on error reply)
4485 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4488 - fix many simultaneous client startup (392)
4489 - fix dentry->d_it clobbering
4490 - credentials weren't being shipped for readdir/getattr operations
4491 - remove invalid assertions triggered during some concurrent MD
4493 - proper Lustre versions added (336, 389)
4494 - fix memory leak for create error case (398)
4495 - fix LOV locking bug that would get cli/srv out of sync
4496 - fix echo client over LOV (409)
4497 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4498 - fix double-O_EXCL intent crash (424)
4499 - avoid sending multiple lock CANCELs (352)
4501 - MDS can do multi-client recovery (modulo bugs in new code)
4503 - many updates, edits, cleanups
4505 2002-11-18 Phil Schwan <phil@clusterfs.com>
4508 - fix null d_it dereference (346)
4509 - fix full OST/dbench hang (333)
4510 - fix permission problem with file removal (286)
4511 - fix removal of OSCs from LOV when they fail
4512 - fix NULL deref during bulk timeout (214)
4513 - fix problems related to multiple filesystems on one MDS (241)
4514 - fixed serious subtle metadata locking bugs
4515 - free locks on clients when inodes are removed due to memory
4517 - fix inode pointer in lock data (285)
4518 - partial support for multiple MDS on a single host (241)
4519 - data locks weren't cancelled at clear_inode time (290, 311)
4520 - intent locks could lead to unbounded lock growth (205)
4521 - added a maximum lock count, an LRU list, and a flusher
4522 - fix multiple rename (365)
4523 - properly abstracted the echo client
4524 - OSC locked 1 byte too many; fixed
4525 - rewrote brw callback code:
4526 - fixed recovery bugs related to LOVs (306)
4527 - fixed too-many-pages-in-one-write crash (191)
4528 - fixed (again) crash in sync_io_timeout (214)
4529 - probably fixed callback-related race (385)
4531 - Add capability to MDS protocol
4532 - LDLM cancellations and callbacks on different portals
4534 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4537 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4538 - defer open object destruction to close time (601981, 138)
4539 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4540 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4541 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4542 - don't call obd_setattr in ll_file_release for destroyed objects
4543 * protocol change to lustre_msg: move |version| and add |flags|
4544 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4545 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4546 * added replay of create, unlink, link and rename operations during
4547 MDS failover; recovery should be much more robust now
4548 * remove failed OSCs from LOVs (only lov_create uses this so far)
4549 * the lustre-HOWTO was brought (more) up to date (582544)
4551 2002-10-23 Phil Schwan <phil@clusterfs.com>
4554 - in-use dentries weren't being reused properly (617851)
4555 - prevent multiple LDLM setup (599178)
4556 - fix LOV size calculations for truncate (617853)
4557 - fix client handling of MDS intent errors (POSIX)
4558 - fix permission bug in lovstripe.c test (624321)
4559 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4560 - truncate past end of file could corrupt data
4561 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4562 - a race in recovery could return ETIMEDOUT to apps (623947)
4563 - building outside the source directory was fixed
4564 * the lustre-HOWTO was brought (more) up to date (582544)
4565 * major progress was made on recovery functionality
4567 2002-10-10 Phil Schwan <phil@clusterfs.com>
4570 - recovery deadlock fix
4571 - rm -rf causes LBUG fix (617817)
4572 - file open by multiple tasks fix (618962)
4573 - directory permissions bugs (602707 and 620007)
4574 - journal_stop fixed with locking (611313)
4575 - O_APPEND failures resolved (618273, perhaps 614459)
4576 - lconf PATH fix (619770)
4577 - IA64 build fix (621450)
4578 - RPC buffer sizes scale with amount of memory
4580 2002-10-01 Phil Schwan <phil@clusterfs.com>
4583 - locks would be cancelled without throwing away data pages,
4584 resulting in inconsistent data (605627)
4585 - inode attributes were not always being refreshed (605627, 612449)
4586 - lconf now continues to cleanup after lctl reports an error
4587 - MDS now enforces user permissions (602707)
4588 - lprocfs cleanup fixed, but not yet enabled (614157)
4589 - fixed infinite server hang, should a client not respond to an AST
4590 - avoid going into recovery if user calls readlink() with a buffer
4591 that's too small (613941)
4592 - AST RPCs no longer require replies (614867) -- this may be changed
4593 - don't crash server if client sends an IOV that's too big (611336)
4594 - fixed lock conversion deadlock (611892)
4595 - fixed the following of symlinks (614622)
4596 * recovery: the server can remove locks from a client that dies, other
4597 clients can make progress
4598 * more extN patch fixes
4599 * compile-time configurable ptlrpc buffer allocations
4601 - collaborative read cache document
4602 - Lustre Lite Performance CDR document-in-progress
4604 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4607 - fix typo in patch-2.4.18
4609 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4612 - clear ptlrpc request each time in handle_incoming_request()
4613 - unlink of files now destroys the object on the OST
4615 2002-09-19 Peter Braam <braam@clusterfs.com>
4617 * add hard link support
4618 * change obdfile creation method
4619 * kernel patch changed
4621 2002-09-19 Peter Braam <braam@clusterfs.com>
4624 - stack overflow bug in extN fixed
4626 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4628 * documentation updates
4629 - add man pages for config tools
4630 - update tests/README to describe testing with new config tools
4631 - finish metadata API descriptions
4632 * bug fixes and cleanups
4633 - statfs workaround for 16TB limit
4634 - LOV stripe allocation improved, can stripe on subset of OSTs
4635 - LOV file size/IO offset was wrong for files > 4GB in size
4636 - object EA data was being dropped, caused files to be unreadable
4637 - memory overflow with non-LOV OST caused memory corruption
4638 - fixed regression tests to work with new config tools, obdfilter
4639 - fixed bug when directory size became larger than 1 block
4640 - fixed bug (for single client case) when PWD was deleted
4641 - invalidate local directory pages when doing intent-based ops
4642 - avoid LDLM oops when lock callback contained bad data
4644 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4646 * documentation updates
4647 * bug fixes and cleanups
4648 - configuration tools
4651 - 64-bit compile warnings
4652 - 64-bit internal statfs data
4654 * test_brw on persistent OST devices
4656 * lprocfs (disabled)
4658 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4660 * documentation updates
4661 * bug fixes and cleanups
4662 * configuration tools
4664 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4667 * many small fixes to 0_5_4
4668 * io/network handling
4669 * thinkos in MDS operations
4671 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4674 * crucial basic fixes to 0.5.3
4675 * IOR, Iozone work over Elan
4678 2002-08-07 Phil Schwan <phil@clusterfs.com>
4679 * version 0_5_3, our first alpha
4680 * we use the new Portals iovs
4681 * documentation updates
4682 * bug fixes and cleanups
4683 * small changes in the DLM wire protocol
4685 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4686 * version 0_5_1 with some initial stability,
4687 * locking on MD and file I/O.
4688 * documentation updates
4689 * several bug fixes since 0.5.0
4690 * small changes in wire protocol
4692 2002-07-18 Phil Schwan <phil@clusterfs.com>
4694 * delivered as Lustre Light Alpha
4695 * fixed a crash after handling invalid MDS requests
4696 * fixed directory pages for architectures with non-4k pages sizes
4698 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4699 * release version v0_4_4
4700 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4702 * Fixes a number of interruption problems with OST operations.
4703 * Update documentation for portals header changes
4704 * Move all wire protocol structs/defines to lustre_idl.h
4705 * Fixes symlink length bug.
4706 * Add tcpdump to repository.
4708 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4709 * release version v0_4_3
4710 * Fixes statfs for inodes on extN.
4711 * Fixes bug in runtests which would delete /etc/hosts.
4712 * Use 64-bit object IDs wherever possible (not into VFS though)
4713 Remove ost_get_info, which is unused by lustre, and out of date.
4715 2002-07-03 Peter Braam <braam@clusterfs.com>
4716 * release version v0_4_2 Fixes a lookup error (type not passed)
4717 * move forward to head of Portals
4718 * move forward to latest Lustre kernel
4720 2002-06-25 Peter Braam <braam@clusterfs.com>
4721 * release version v0_4_1. Hopefully stable on single node use.