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).
15 Severity : enhancement
17 Description: using special macro for print time and cleanup in includes.
21 Description: mds_mfd_close() ASSERTION(rc == 0)
22 Details : In mds_mfd_close(), we need protect inode's writecount change
23 within its orphan write semaphore to prevent possible races.
27 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
31 Frequency : rare, on shutdown ost
32 Description: don't hit live lock with umount ost.
33 Details : shrink_dcache_parent can be in long loop with destroy dentries,
34 use shrink_dcache_sb instead.
38 Description: don't panic with use echo client
39 Details : echo client pass NULL as client nid pointer and this produce null
44 Description: fix build on ppc32
45 Details : compile code with -m64 flag produce wrong object file for ppc32.
49 Description: add message levels for liblustreapi
53 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
54 Details : Move the 'good_osts' check before the 'total_bavail' check. This
55 will result in an -EAGAIN and in the exit call path we call
56 alloc_rr() which will with increasing aggressiveness attempt to
57 aquire precreated objects on the minimum number of required OSCs.
61 Description: Use old size assignment to avoid deadlock
62 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
63 the scheduling while holding a spinlock. We do not need locking
64 for size in ll_update_inode() because size is only updated from
65 the MDS for directories or files without objects, so there is no
66 other place to do the update, and concurrent access to such inodes
67 are protected by the inode lock.
71 Description: resolve "_IOWR redefined" build error on SLES10
75 Description: dump the memory debugging after all modules are unloaded to
76 suppress false negative in conf_sanity test 39
78 Severity : enhancement
80 Description: build kernel-ib packages for OFED 1.3 in our release cycle
85 Description: fix SLES kernel versioning
86 Details : the kernel version for our SLES 10 kernel did not include a "-"
87 before the "smp" at the end. while this was not a problem in
88 general, it did mean that software trying to use the kernel
89 version to try to detect a vendor specific kernel would fail.
90 this was most evident by the OFED build scripts.
94 Description: Don't update lov_desc members until making sure they are valid
95 Details : When updating lov_desc members via proc fs, need fix their
96 validities before doing the real update.
100 Description: don't put request into delay list while invalidate in flight.
101 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
102 import in flight. this produce timeout for invalidate and sometimes
103 can cause stale data.
107 Frequency : on ppc only
108 Description: not convert ost objects for directory because it's not exist.
109 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
113 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
115 Severity : enhancement
117 Description: Update to support 2.6.22.14 vanilla kernel.
121 Frequency : rare, on recovery
122 Description: read procfs can produce deadlock in some situation
123 Details : Holding lprocfs lock which send rpc can produce block for destroy
124 obd objects and this also block reconnect with -EALREADY. This isn't
125 fix all lprocfs bugs - but make it rare.
127 Severity : enhancement
129 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
132 Frequency : frequent on X2 node
134 Description: mdc_set_open_replay_data LBUG
135 Details : Set replay data for requests that are eligible for replay.
139 Description: lustre_mgs: operation 101 on unconnected MGS
140 Details : When MGC is disconnected from MGS long enough, MGS will evict the
141 MGC, and late on MGC cannot successfully connect to MGS and a lot
142 of the error messages complaining that MGS is not connected.
145 Frequency : on start mds
147 Description: Implement get_info(last_id) in obdfilter.
150 Frequency : occasional
152 Description: Correctly check stale fid, not start epoch if ost not support SOM
153 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
154 overwrited with new generated fid, but mds can anwer with one of these
155 two fids and both is not stale. setattr incorectly start epoch and
156 assume will be called done_writeting, but without SOM done_writing
160 Frequency : rare, depends on device drivers and load
162 Description: MDS or OSS nodes crash due to stack overflow
163 Details : Code changes in 1.8.0 increased the stack usage of some functions.
164 In some cases, in conjunction with device drivers that use a lot
165 of stack the MDS (or possibly OSS) service threads could overflow
166 the stack. One change which was identified to consume additional
167 stack has been reworked to avoid the extra stack usage.
170 Frequency : occasional
172 Description: Do not fail import if osc_interpret_create gets -EAGAIN
173 Details : If osc_interpret_create got -EAGAIN it immediately exits and
174 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
175 oscc_has_objects and see OSC has no objests and call
176 oscc_internal_create to resend create request.
178 Severity : enhancement
180 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
182 Severity : enhancement
184 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
187 Frequency : very rare
189 Description: Fix replay if there is an un-replied request and open
190 Details : In some cases, older replay request will revert the
191 mcd->mcd_last_xid on MDS which is used to record the client's
194 Severity : enhancement
196 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
198 Severity : enhancement
200 Description: Add rhel5 support to HEAD.
202 Serverity : enhancement
204 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
209 Description: Don't allow skipping OSTs if index has been specified.
210 Details : Don't allow skipping OSTs if index has been specified, make locking
211 in internal create lots better.
215 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
216 Details : access to bitfield in structure is always rounded to long
217 and this produce problem with not atomic change any bit.
221 Description: Lustre make rpms failed.
222 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
223 builds Lustre rpms from tarball.
226 Frequency : rare on shutdown ost
228 Description: If llog cancel was not send before clean_exports phase, this can
229 produce deadlock in llog code.
230 Details : If llog thread has last reference to obd and call class_import_put
231 this produce deadlock because llog_cleanup_commit_master wait when
232 last llog_commit_thread exited, but this never success because was
233 called from llog_commit_thread.
237 Description: allow userland application know is lost one of stripes.
238 Details : fill lvb_blocks with error code on ost and return it to
239 application if error flag found.
243 Description: NULL lov_tgts causing MDS oops
244 Details : more safe checks for NULL lov_tgts for avoid oops.
246 Severity : enhancement
248 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
252 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
253 Details : If an OST is down the MDS will hang indefinitely in
254 obd_statfs_rqset() waiting for the statfs data. While for
255 MDS QOS usage of statfs, it should not stuck in waiting.
257 Severity : enhancement
259 Description: remote_acl support
260 Details : Support ACL-based permission check for remote user.
261 Support setfacl/getfacl for remote user with the utils
262 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
263 the system "{s,g}etfacl" utils.
265 Severity : enhancement
267 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
269 Severity : enhancement
271 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
275 Description: make mgs_setparam() handle fsname containing dash
276 Details : fsname containing a dash does not work with lctl conf_param
278 Severity : enhancement
280 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
282 Severity : enhancement
284 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
288 Frequency : rare, at shutdown
289 Description: access already free / zero obd_namespace.
290 Details : if client_disconnect_export was called without force flag set,
291 and exist connect request in flight, this can produce access to
292 NULL pointer (or already free pointer) when connect_interpret
293 store ocd flags in obd_namespace.
297 Frequency : only at startup
298 Description: not alloc memory with spinlock held.
299 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
300 if any spinlock held.
302 Severity : enhancement
304 Description: make lustre randomly fail allocating memory
305 Details : Make lustre randomly failed allocating memory for testing purpose.
307 Severity : enhancement
309 Description: lost problems with lov objid file
310 Details : Fixes some scability and access to not inited memory problems
311 in work with lov objdid file.
316 Description: lfs find does not continue on file error
317 Details : Continue other files processing when a file/dir is absent.
321 Description: Inconsistent usage of lustre_pack_reply()
322 Details : Standardize the usage of lustre_pack_reply() such that it
323 always generate a CERROR on failure.
326 Frequency : occasional
328 Description: MDS hang or stay in waiting lock
329 Details : If client receive lock with CBPENDING flag ldlm need send lock
330 cancel as separate rpc, to avoid situation when cancel request
331 can't processed due all i/o threads stay in wait lock.
335 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
339 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
343 Description: Update to SLES9 kernel-2.6.5-7.287.3.
347 Description: Update to RHEL4 latest kernel.
349 Severity : enhancement
351 Description: Build SLES10 patchless client fails
352 Details : The configure was broken by run ./configure with
353 --with-linux-obj=.... argument for patchless client. When the
354 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
355 can't search header adequately. Use absolute path such as
356 -I($LINUX)/include instead.
360 Description: interrupt oig_wait produce painc on resend.
361 Details : brw_redo_request can be used for resend requests from ptlrpcd and
362 private set, and this produce situation when rq_ptlrpcd_data not
363 copyed to new allocated request and triggered LBUG on assert
364 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
365 ptlrpcd set if request is changed and can be safety changed to use
370 Description: LASSERT_{REQ,REP}SWAB macros are buggy
371 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
372 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
373 these macros and replace them with their difinitions instead.
377 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
378 Details : Update which_patch & target file for SLES10 latest kernel.
381 Description: Remove client patches from SLES 10 kernel.
382 Details : This causes SLES 10 clients to behave as patchless clients
383 even on a Lustre-patched (server) kernel.
385 Severity : enhancement
387 Description: self-adjustable client's lru lists
388 Details : use adaptive algorithm for managing client cached locks lru
389 lists according to current server load, other client's work
390 pattern, memory activities, etc. Both, server and client
391 side namespaces provide number of proc tunables for controlling
394 Severity : enhancement
396 Description: light-weight GSS support
397 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
398 and reduce performance overhead.
400 Severity : enhancement
402 Description: Linux keyring support
403 Details : Support using service of Linux keyring for Lustre GSS internal
404 context refresh/cache mechanism.
408 Description: Fix errors in lfs documentation
409 Details : Fixes man pages
413 Description: don't use GFP_* in generic Lustre code.
414 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
418 Description: obdclass is limited by single OBD_ALLOC(idarray)
419 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
423 Description: warnings with build patchless client with vanila 2.6.19 and up
424 Details : change old ctl_table style and replace ctl_table/ctl_table_header
425 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
429 Frequency : Only for SLES
430 Description: sanity_quota fail test_1
431 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
432 file, which makes TSTID[2] inunique.
436 Frequency : Only for SLES9
437 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
438 Details : Update target/ChangeLog/which_patch .
442 Frequency : Only for test_10 in sanity-quota.sh
443 Description: a bug in quota test code
444 Details : checking a test flag in a wrong place
448 Frequency : Only for test_13 in sanity-quota.sh
449 Description: a bug in quota test script
450 Details : shouldn't check overfull allocated quota
454 Description: jbd statistics
455 Details : Port older jbd statistics patch for sles10
457 Severity : enhancement
459 Description: Add printing inode info into message about error in writepage.
464 Description: reply_lock_interpret crash due to race with it and lock cancel.
465 Details : Do not replay locks that are being cancelled. Do not reference
466 locks by their address during replay, just by their handle.
470 Frequency : When flocks are used.
471 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
472 Details : Flock locks might destroy just granted lock if it could be merged
473 with another existing flock, this is done in completion handler,
474 so teach ldlm_cli_enquque_fini that this is a valid case for
480 Description: Oops in read and write path when failing to allocate lock.
481 Details : Check if lock allocation failed and return error back.
485 Description: lstripe command fails for valid OST index
486 Details : The stripe offset is compared to lov->desc.ld_tgt_count
487 instead of lov->desc.ld_active_tgt_count.
491 Description: sanity.sh failed test 103
492 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
497 Description: df doesn't work properly if diskfs blocksize != 4K
498 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
502 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
503 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
507 Description: obd_health_check_timeout too short
508 Details : set obd_health_check_timeout as 1.5x of obd_timeout
512 Description: llapi_file_create() does not allow some changes
513 Details : add llapi_file_open() that allows specifying the mode and
514 open flags, and also returns an open file handle.
518 Description: merge and cleanup kernel patches.
519 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
523 Description: Add journal checksum support.(Kernel part)
524 Details : The journal checksum feature adds two new flags i.e
525 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
526 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
527 indicates that the commit block contains the checksum for
528 the blocks described by the descriptor blocks. Now commit
529 record can be sent to disk without waiting for descriptor
530 blocks to be written to disk. This behavior is controlled
531 using JBD2_FEATURE_ASYNC_COMMIT flag.
535 Description: OSS needs mutliple precreate threads
536 Details : Add ability to start more than one create thread per OSS.
540 Description: Sanity.sh test_65k failed: setstripe should have succeeded
541 Details : execute lfs setstripe on client
545 Description: mds_obd_create error creating tmp object
546 Details : When the user sets quota on root, llog will be affected and can't
547 create files and write files.
550 Frequency : Always on ia64 patchless client, and possibly others.
552 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
553 Details : This allows the patchless client to be loaded on architectures
558 Description: RedHat Update kernel for RHEL5
559 Details : Add the kernel config file for RHEL5.
563 Description: RedHat Update kernel for RHEL5
564 Details : Modify the kernel config file more closer RHEL5.
568 Description: Build failure against Centos5 (RHEL5)
569 Details : Define PAGE_SIZE when it isn't present.
573 Description: test_7 on sanity-quota.sh can't be used on separate machines
574 Details : add facet to handle it
578 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
579 Details : with NFS, the anon dentry's parent was set to itself in
580 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
581 corresponding dentry other than use rec->ur_name.
583 Severity : enhancement
585 Description: lfs setstripe enhancement
586 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
590 Description: enable data checksumming by default
591 Details : enable checksum by default, allow --disable-checksum
592 configure option and "-o nochecksum" mount option. Checksums
593 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
597 Description: lustre support for RHEL5
598 Details : Add support for RHEL5.
602 Description: Client eviction due to failover config
603 Details : after a connection loss, the lustre client should attempt to
604 reconnect to the last active server first before trying the
605 other potential connections.
609 Description: when mds and osts use different quota unit(32bit and 64bit),
610 quota will be released repeatly.
611 Details : void sending multiple quota reqs to mds, which will keep the
612 status between the reqs.
616 Description: osts not allocated evenly to files
617 Details : change the condition to increase offset_idx
620 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
622 Description: Data corruption for OSTs that are formatted larger than 2TB
624 Details : When generating the bio request for lustre file writes the
625 sector number would overflow a temporary variable before being
626 used for the IO. The data reads correctly from Lustre (which
627 will overflow in a similar manner) but other file data or
628 filesystem metadata may be corrupted in some cases.
632 Description: Tune the kernel for good SCSI performance.
633 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
634 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
639 Description: rewrite ext2-derived code in obdclass/uuid.c
640 Details : rewrite inherited code (uuid parsing code from ext2 utils)
641 from scratch preserving functionality.
645 Description: "lfs find -obd UUID" prints directories
646 Details : "lfs find -obd UUID" will return all directory names instead
647 of just file names. It is incorrect because the directories
648 do not reside on the OSTs.
652 Description: use i_size_read and i_size_write in 2.6 port
653 Details : replace inode->i_size access with i_size_read/write()
656 Frequency : when removing large files
658 Description: scheduling issue during removal of large Lustre files
659 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
660 It causes scheduling issues when removing large files (17TB in the
664 Frequency : only with liblustre clients on XT3
666 Description: evictions taking too long
667 Details : allow llrd to evict clients directly on OSTs
670 Frequency : only on ppc
672 Description: /proc/fs/lustre/devices broken on ppc
673 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
674 We should make sure the type of 'index' is loff_t and then cast
675 explicitly as needed below. Do not assign an explicitly cast
679 Frequency : only for rhel5
681 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
682 Details : Modify the target file & which_kernel.
684 Severity : enhancement
686 Description: omit set fsid for export NFS
687 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
688 as FS_REQUIRES_DEV for avoid problems with generate fsid.
692 Frequency : Always, for kernels after 2.6.16
693 Description: Fix warning idr_remove called for id=.. which is not allocated.
694 Details : Last kernels save old s_dev before kill super and not allow
695 to restore from callback - restore it before call kill_anon_super.
699 Description: buffer overruns could theoretically occur
700 Details : llapi_semantic_traverse() modifies the "path" argument by
701 appending values to the end of the origin string, and a
702 overrun may occur. Adding buffer overrun check in liblustreapi.
706 Description: Fix error on 'ls .' at the top of the Lustre mount.
707 Details : Don't revalidate dentry if it is a root dentry.
711 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
712 Details : Modify vm-tunables-rhel4.patch.
716 Description: Kernel config for 2.6.18-vanilla.
717 Details : Modify targets/2.6-vanilla.target.in.
718 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
719 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
720 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
721 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
725 Description: improve handling recoverable errors
726 Details : if request processig with error which can be recoverable on server
727 request should be resend, otherwise page released from cache and
732 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
733 Details : Modify target file & which_patch.
734 A flaw was found in the IA32 system call emulation provided
735 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
736 value could be stored in the %RAX register, which could trigger an
737 out-of-bounds system call table access. An untrusted local user
738 could exploit this flaw to run code in the kernel
739 (ie a root privilege escalation). (CVE-2007-4573).
743 Description: O_DIRECT bypasses client statistics.
744 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
745 read_ahead_stats were updated. copied the stats section from
746 osc_send_oap_rpc() into async_internal().
750 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
754 Description: change order of libsysio includes
755 Details : '#include sysio.h' should always come before '#include xtio.h'
759 Description: handle "serious error: objid * already exists" more gracefully
760 Details : If LAST_ID value on disk is smaller than the objects existing in
761 the O/0/d* directories, it indicates disk corruption and causes an
762 LBUG(). If the object is 0-length, then we should use the existing
763 object. This will help to avoid a full fsck in most cases.
765 Severity : enhancement
767 Description: adapt the lustre_config script to support the upgrade case
768 Details : Add "-u" option for lustre_config script to support upgrading 1.4
769 server targets to 1.6 in parallel.
773 Description: To avoid grant space > avaible space when the disk is almost
774 full. Without this patch you might see the error "grant XXXX >
775 available" or some LBUG about grant, when the disk is almost
777 Details : In filter_check_grant, for non_grant cache write, we should
778 check the left space by if (*left > ungranted + bytes), instead
779 of (*left > ungranted), because only we are sure the left space
780 is enough for another "bytes", then the ungrant space should be
781 increase. In client, we should update cl_avail_grant only there
782 is OBD_MD_FLGRANT in the reply.
786 Description: Update RHEL 4 kernel to fix local root privilege escalation.
787 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
788 described in CVE-2007-4573. This problem could allow untrusted
789 local users to gain root access.
792 Frequency : when using O_DIRECT and quotas
794 Description: Incorrect file ownership on O_DIRECT output files
795 Details : block usage reported by 'lfs quota' does not take into account
796 files that have been written with O_DIRECT.
801 Description: touch file failed when fs is not full
802 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
803 otherwise we can get ENOSP while fs is not full.
807 Description: parallel lock callbacks
808 Details : Instead of sending blocking and completion callbacks as separated
809 requests, adding them to a set and sending in parallel.
812 Frequency : only for Cray XT3
813 Bugzilla : 12829/13455
814 Description: Changing primary group doesn't change the group lustre assigns to
816 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
817 overridden with the primary group provided by the group upcall,
818 whereas the supplied fsgid can be trusted if it is in the list of
819 supplementary groups returned by the group upcall.
821 Severity : enhancement
823 Description: Allow masking D_WARNING, D_ERROR messages from console
824 Details : Console messages can now be disabled via lnet.printk.
828 Description: User code with malformed file open parameter crashes client node
829 Details : Before packing join_file req, all the related reference should be
830 checked carefully in case some malformed flags cause fake
831 join_file req on client.
835 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
836 Details : ldlm_completion_ast() assumes that a lock is granted when the req
837 mode is equal to the granted mode. However, it should also check
838 that LDLM_FL_CP_REQD is not set.
842 Description: Heavy nfs access might result in deadlocks
843 Details : After ELC code landed, it is now improper to enqueue any mds
844 locks under och_sem, because enqueue might want to decide to
845 cancel open locks for same inode we are holding och_sem for.
849 Description: Client eviction while running blogbench
850 Details : A lot of unlink operations with concurrent I/O can lead to a
851 deadlock causing evictions. To address the problem, the number of
852 oustanding OST_DESTROY requests is now throttled to
853 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
854 callbacks are processed in priority.
858 Description: enable ACLs on MDS by default
859 Details : ACLs must be enabled on MDS by default.
862 Frequency : PPC/PPC64 only
864 Description: conflicts between asm-ppc64/types.h and lustre_types.h
865 Details : fix duplicated definitions between asm-ppc64/types.h and
866 lustre_types.h on PPC.
869 Frequency : PPC/PPC64 only
871 Description: asm-ppc/segment.h does not exist
872 Details : fix compile issue on PPC.
876 Description: better handle error messages in extents code
879 Frequency : RHEL4 only
881 Description: mkfs is very slow on IA64/RHEL4
882 Details : A performance regression has been discovered in the MPT Fusion
883 driver between versions 3.02.73rh and 3.02.99.00rh. As a
884 consequence, we have downgraded the MPT Fusion driver in the RHEL4
885 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
887 Severity : enhancement
889 Description: SNMP support enhancement
890 Details : Adding total number of sampled request for an MDS node in snmp
893 Severity : enhancement
895 Description: Optimize ldlm waiting list processing for PR extent locks
896 Details : When processing waiting list for read extent lock and meeting read
897 lock that is same or wider to it that is not contended, skip
898 processing rest of the list and immediatelly return current
899 status of conflictness, since we are guaranteed there are no
900 conflicting locks in the rest of the list.
904 Description: Time out and refuse to reconnect
905 Details : When the failover node is the primary node, it is possible
906 to have two identical connections in imp_conn_list. We must
907 compare not conn's pointers but NIDs, otherwise we can defeat
908 connection throttling.
912 Description: port llog fixes from b1_6 into HEAD
913 Details : Port llog reference couting and some llog cleanups from b1_6
914 (bug 10800) into HEAD, for protect from panic and access to already
915 free llog structures.
919 Description: Detect stride IO mode in read-ahead
920 Details : When a client does stride read, read-ahead should detect that and
921 read-ahead pages according to the detected stride pattern.
925 Description: data checksumming impacts single node performance
926 Details : add support for several checksum algorithm. Currently, only CRC32
927 and Adler-32 are supported. The checksum type can be changed on
928 the fly via /proc/fs/lustre/osc/*/checksum_type.
932 Description: use adler32 for page checksums
933 Details : when available, use the Adler-32 algorithm instead of CRC32 for
938 Description: build for x2 fails
939 Details : fix compile issue on Cray systems.
943 Description: Properly match for duplicate locks
944 Details : Due to different lock order from skiplists code, we need to
945 traverse entire list for now
948 Frequency : only on PPC/SLES10
950 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
951 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
952 be defined. Add a hack in mkfs_lustre.c to work around this compile
957 Description: LASSERT on MDS when client holding flock lock dies
958 Details : ldlm pool logic depends on number of granted locks equal to
959 number of released locks which is not true for flock locks, so
960 just exclude such locks from consideration.
964 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
965 Details : Use fsfilt_sync() for both the whole filesystem sync and
966 individual file sync to eliminate dangerous inode locking
967 with I_LOCK that can lead to a deadlock.
971 Description: performance in 1.6.3
972 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
973 sure that 1MB requests are not fragmented by the block layer.
975 Severity : enhancement
977 Description: organize the server-side client stats on per-nid basis
978 Details : Change the structure of stats under obdfilter and mds to
986 The "uuid"s file would list the uuids of _active_ exports.
987 And the clear entry is to clear all stats and stale nids.
989 Severity : enhancement
991 Description: eliminate client locks in face of contention
992 Details : file contention detection and lockless i/o implementation
997 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
999 --------------------------------------------------------------------------------
1001 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1003 * Support for kernels:
1004 2.6.5-7.283 (SLES 9),
1005 2.6.9-55.EL (RHEL 4),
1006 2.6.16.46-0.14 (SLES 10),
1007 2.6.18.8 vanilla (kernel.org)
1008 * Client support for unpatched kernels:
1009 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1010 2.6.16 - 2.6.22 vanilla (kernel.org)
1011 * Due to recently discovered recovery problems, we do not recommend
1012 using patchless RHEL 4 clients with this or any earlier release.
1013 * Recommended e2fsprogs version: 1.39.cfs8
1014 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1015 * Starting with this release, the ldiskfs backing filesystem required
1016 by Lustre is now in its own package, lustre-ldiskfs. This package
1017 should be installed. It is versioned separately from Lustre and
1018 may be released separately in future.
1022 Description: block reactivating mgc import until all deactivates complete
1023 Details : Fix race when failing back MDT/MGS to itself (testing)
1026 Frequency : at statup only
1028 Description: mds_lov_synchronize race leads to various problems
1029 Details : simultaneous MDT->OST connections at startup can cause the
1030 sync to abort, leaving the OSC in a bad state.
1032 Severity : enhancement
1034 Description: add optional extra BUILD_VERSION info
1035 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1036 to override the lustre version.
1039 Frequency : 2.6.18 servers only
1041 Description: ll_kern_mount() doesn't release the module reference
1042 Details : The ldiskfs module reference count never drops down to 0
1043 because ll_kern_mount() doesn't release the module reference.
1048 Description: server LBUG when using old ost_num_threads parameter
1049 Details : Accept the old ost_num_threads parameter but warn that it
1050 is deprecated, and fix an off-by-one error that caused an LBUG.
1055 Description: Transient SCSI error results in persistent IO issue
1056 Details : iobuf->dr_error is not reinitialized to 0 between two
1060 Frequency : sometimes when underlying device returns I/O errors
1062 Description: OSTs not going read-only during write failures
1063 Details : OSTs are not remounted read-only when the journal commit threads
1064 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1065 instead of the ext3 wrappers.
1069 Description: poor connect scaling with increasing client count
1070 Details : Don't run filter_grant_sanity_check for more than 100 exports
1071 to improve scaling for large numbers of clients.
1074 Frequency : SLES10 only
1076 Description: sanity-quota.sh quotacheck failed: rc = -22
1077 Details : Quotas cannot be enabled on SLES10.
1080 Frequency : liblustre clients only
1082 Description: getdirentries does not give error when run on compute nodes
1083 Details : getdirentries does not fail when the size specified as an argument
1084 is too small to contain at least one entry
1086 Severity : enhancement
1088 Description: Add LNET router traceability for debug purposes
1089 Details : If a checksum failure occurs with a router as part of the
1090 IO path, the NID of the last router that forwarded the bulk data
1091 is printed so it can be identified.
1096 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1097 Details : Due to a race condition, liblustre clients were occasionally
1098 evicted incorrectly.
1100 Severity : enhancement
1102 Description: lfs setstripe use optional parameters instead of postional
1105 Severity : enhancement
1107 Description: Nanosecond timestamp support for ldiskfs
1108 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1109 resolution timestamps. There is not yet support for this at
1110 the Lustre filesystem level.
1113 Frequency : during server recovery
1115 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1116 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1120 Frequency : nfs export on patchless client
1122 Description: connectathon hang when test nfs export over patchless client
1123 Details : Disconnected dentry cannot be found with lookup, so we do not need
1124 to unhash it or make it invalid
1127 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1128 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1129 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1134 Description: open req refcounting wrong on reconnect
1135 Details : If reconnect happened between getting open reply from server and
1136 call to mdc_set_replay_data in ll_file_open, we will schedule
1137 replay for unreferenced request that we are about to free.
1138 Subsequent close will crash in variety of ways.
1139 Check that request is still eligible for replay in
1140 mdc_set_replay_data().
1145 Description: disable writes to filesystem when reading health_check file
1146 Details : the default for reading the health_check proc file has changed
1147 to NOT do a journal transaction and write to disk, because this
1148 can cause reads of the /proc file to hang and block HA state
1149 checking on a healthy but otherwise heavily loaded system. It
1150 is possible to return to the previous behaviour during configure
1151 with --enable-health-write.
1153 Severity : enhancement
1155 Description: 64-bit inode version
1156 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1157 to the inode. This will be required for version-based recovery.
1162 Description: MDS fails to start if a duplicate client export is detected
1163 Details : in some rare cases it was possible for a client to connect to
1164 an MDS multiple times. Upon recovery the MDS would detect this
1165 and fail during startup. Handle this more gracefully.
1167 Severity : enhancement
1169 Description: Add -o localflock option to simulate old noflock
1171 Details : This will achieve local-only flock/fcntl locks
1177 Description: log_commit_thread vs filter_destroy race leads to crash
1178 Details : Take import reference before releasing llog record semaphore
1183 Description: Wrong request locking in request set processing
1184 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1185 imp_delayed_list, in this place should be used imp_lock.
1188 Frequency : when reconnection
1190 Description: Grant Leak when osc reconnect to OST
1191 Details : When osc reconnect ost, OST(filter) should check whether it
1192 should grant more space to client by comparing fed_grant and
1193 cl_avail_grant, and return the granted space to client instead
1194 of "new granted" space, because client will call osc_init_grant
1195 to update the client grant space info.
1198 Frequency : when client reconnect to OST
1200 Description: Grant Leak when osc do resend and replay bulk write
1201 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1202 bulk write request, because the grant info while be sync between
1203 OSC and OST when reconnect, and we should ignore the grant info
1204 these of resend/replay write req.
1209 Description: Grant space more than avaiable left space sometimes.
1210 Details : When then OST is about to be full, if two bulk writing from
1211 different clients came to OST. Accord the avaliable space of the
1212 OST, the first req should be permitted, and the second one
1213 should be denied by ENOSPC. But if the seconde arrived before
1214 the first one is commited. The OST might wrongly permit second
1215 writing, which will cause grant space > avaiable space.
1218 Frequency : when client is evicted
1220 Description: Grant might be wrongly erased when osc is evicted by OST
1221 Details : when the import is evicted by server, it will fork another
1222 thread ptlrpc_invalidate_import_thread to invalidate the
1223 import, where the grant will be set to 0. While the original
1224 thread will update the grant it got when connecting. So if
1225 the former happened latter, the grant will be wrongly errased
1226 because of this race.
1231 Description: Checking Stale with correct fid
1232 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1233 to check whether it is stale, because sometimes, we want the
1234 enqueue happened anyway, and op_data.fid2 will not be initialized.
1236 Severity : enhancement
1238 Description: update patchless client
1239 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1242 Frequency : only with 2.4 kernel
1244 Description: random memory corruption
1245 Details : size of struct ll_inode_info is to big for union inode.u and this
1246 can be cause of random memory corruption.
1251 Description: Memory leak in recovery
1252 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1253 It should also check obd_fail before fsfilt_start, otherwise if
1254 fsfilt_start return -EROFS,(failover mds during mds recovery).
1255 then the req will return with repmsg->transno = 0 and rc = EROFS.
1256 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1257 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1258 be freed no matter whether fsfilt_commit success or not.
1261 Frequency : only with huge count clients
1263 Description: Prevents from taking the superblock lock in llap_from_page for
1265 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1266 for llap_from_page call in ll_removepage prevents from taking the
1267 superblock lock for a soon died page.
1272 Description: Not check open intent error before release open handle
1273 Details : in some rare cases, the open intent error is not checked before
1274 release open handle, which may cause
1275 ASSERTION(open_req->rq_transno != 0), because it tries to release
1276 the failed open handle.
1281 Description: Set cat log bitmap only after create log success.
1282 Details : in some rare cases, the cat log bitmap is set too early. and it
1283 should be set only after create log success.
1287 Description: Accessing a block bevice can re-enable I/O when Lustre is
1288 tearing down a device.
1289 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1294 Description: service threads may hog cpus when there are a lot of requests
1295 Details : Insert cond_resched to give other threads a chance to use some CPU
1300 Description: the cat log was not initialized in recovery
1301 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1302 unlink log on mds will not be initialized until mds post
1303 recovery. And also in mds post recovery, the unlink log will
1304 initialization will be done asynchronausly, so there will be race
1305 between add unlink log and unlink log initialization.
1309 Description: brw_stats were being printed incorrectly
1310 Details : brw_stats were being printed as log2 but all of them were not
1311 recorded as log2. Also remove some code duplication arising from
1312 filter_tally_{read,write}.
1316 Frequency : rare, only in recovery.
1317 Description: ASSERTION(req->rq_type != LI_POISON) failed
1318 Details : imp_lock should be held while iterating over imp_sending_list for
1319 prevent destroy request after get timeout in ptlrpc_queue_wait.
1323 Description: replay-single.sh test 52 fails
1324 Details : A lock's skiplist need to be cleanup when it being unlinked
1325 from its resource list.
1329 Description: Short directio read returns full requested size rather than
1331 Details : Direct I/O operations should return actual amount of bytes
1332 transferred rather than requested size.
1334 Severity : enhancement
1336 Description: metadata RPC reduction (e.g. for rm performance)
1337 Details : decrease the amount of synchronous RPC between clients and servers
1338 by canceling conflicing lock before the operation on the client
1339 and packing thier handles into the main operation RPC to server.
1341 Severity : enhancement
1343 Description: Async OSC create to avoid the blocking unnecessarily.
1344 Details : If a OST has no remain object, system will block on the creating
1345 when need to create a new object on this OST. Now, ways use
1346 pre-created objects when available, instead of blocking on an
1347 empty osc while others are not empty. If we must block, we block
1348 for the shortest possible period of time.
1350 Severity : enhancement
1352 Description: refine locking for avoid write wrong info into lov_objid file
1353 Details : fix possible races with add new target and write/update data in
1356 --------------------------------------------------------------------------------
1358 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1363 Frequency : on some architectures
1365 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1366 Details : Uninitialized flags sometimes cause configuration commands to
1370 Frequency : patchless clients only
1372 Description: missing __iget() symbol export
1373 Details : The __iget() symbol export is missing. To avoid the need for
1374 this on patchless clients the deathrow inode reaper is turned
1375 off, and we depend on the VM to clean up old inodes. This
1376 dependency was during via the fix for bug 12181.
1378 --------------------------------------------------------------------------------
1380 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1382 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1383 INTEROPERATE with older versions automatically. In many cases a
1384 special upgrade step is needed. Please read the
1385 user documentation before upgrading any part of a live system.
1386 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1387 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1388 * WARNING: Lustre configuration and startup changes are required with
1389 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1395 Severity : enhancement
1397 Description: MountConf
1398 Details : Lustre configuration is now managed via mkfs and mount
1399 commands instead of lmc and lconf. New obd types (MGS, MGC)
1400 are added for dynamic configuration management. See
1401 https://mail.clusterfs.com/wikis/lustre/MountConf for
1404 Severity : enhancement
1406 Description: dynamic OST addition
1407 Details : OSTs can now be added to a live filesystem
1409 Severity : enhancement
1411 Description: startup order invariance
1412 Details : MDTs and OSTs can be started in any order. Clients only
1413 require the MDT to complete startup.
1415 Severity : enhancement
1417 Description: parallel, asynchronous orphan cleanup
1418 Details : orphan cleanup is now performed in separate threads for each
1419 OST, allowing parallel non-blocking operation.
1421 Severity : enhancement
1423 Description: optimized stripe assignment
1424 Details : stripe assignments are now made based on ost space available,
1425 ost previous usage, and OSS previous usage, in order to try
1426 to optimize storage space and networking resources.
1428 Severity : enhancement
1430 Description: Permanently set tunables
1431 Details : All writable /proc/fs/lustre tunables can now be permanently
1432 set on a per-server basis, at mkfs time or on a live system.
1434 Severity : enhancement
1436 Description: Lustre message v2
1437 Details : Add lustre message format v2.
1439 Severity : enhancement
1441 Description: client OST exclusion list
1442 Details : Clients can be started with a list of OSTs that should be
1443 declared "inactive" for known non-responsive OSTs.
1447 Description: SPEC SFS validation failure on NFS v2 over lustre.
1448 Details : Changes the blocksize for regular files to be 2x RPC size,
1449 and not depend on stripe size.
1451 Severity : enhancement
1453 Description: Multiple MD RPCs in flight.
1454 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1455 To avoid overly-overloading MDS, introduce a limit on number of
1456 MDS RPCs in flight for a single client and add /proc controls
1457 to adjust this limit.
1459 Severity : enhancement
1461 Description: client read/write statistics
1462 Details : Add client read/write call usage stats for performance
1463 analysis of user processes.
1464 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1465 file access. extents_stats shows chunk size distribution.
1466 extents_stats_per_process show chunk size distribution per
1469 Severity : enhancement
1471 Description: mds statistics
1472 Details : Add detailed mds operations statistics in
1473 /proc/fs/lustre/mds/*/stats.
1477 Description: Failure of copying files with lustre special EAs.
1478 Details : Client side always return success for setxattr call for lustre
1479 special xattr (currently only "trusted.lov").
1484 Description: Refcount LNET uuids
1485 Details : The global LNET uuid list grew linearly with every startup;
1486 refcount repeated list entries instead of always adding to
1489 Severity : enhancement
1491 Description: Dynamic service threads
1492 Details : Within a small range, start extra service threads
1493 automatically when the request queue builds up.
1496 Frequency : mixed-endian client/server environments
1498 Description: mixed-endian crashes
1499 Details : The new msg_v2 system had some failures in mixed-endian
1502 Severity : enhancement
1504 Description: Easy OST removal
1505 Details : OSTs can be permanently deactivated with e.g. 'lctl
1506 conf_param lustre-OST0001.osc.active=0'
1508 Severity : enhancement
1510 Description: MGS proc entries
1511 Details : Added basic proc entries for the MGS showing what filesystems
1514 Severity : enhancement
1516 Description: provide MGS failover
1517 Details : Added config lock reacquisition after MGS server failover.
1519 Severity : enhancement
1521 Description: add Linux 2.4 support
1522 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1526 Description: a large application tries to do I/O to the same resource and dies
1527 in the middle of it.
1528 Details : Check the req->rq_arrival time after the call to
1529 ost_brw_lock_get(), but before we do anything about
1530 processing it & sending the BULK transfer request. This
1531 should help move old stale pending locks off the queue as
1532 quickly as obd_timeout.
1535 Frequency : when an incorrect nid is specified during startup
1537 Description: ptlrpc connect to non-existant node causes kernel crash
1538 Details : LNET can't be re-entered from an event callback, which
1539 happened when we expire a message after the export has been
1540 cleaned up. Instead, hand the zombie cleanup off to another
1543 Severity : enhancement
1545 Description: plain/inodebits lock performance improvement
1546 Details : Grouping plain/inodebits in granted list by their request modes
1547 and bits policy, thus improving the performance of search through
1551 Frequency : only if OST filesystem is corrupted
1553 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1554 Details : for a short time RPCs with bulk IO are in the replay list,
1555 but replay of bulk IOs is unimplemented. If the OST filesystem
1556 is corrupted due to disk cache incoherency and then replay is
1557 started it is possible to trip an assertion. Avoid putting
1558 committed RPCs into the replay list at all to avoid this issue.
1561 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1563 Description: System hang on startup
1564 Details : This bug allowed the liblustre (e.g. catamount) client to
1565 return to the app before handling all startup RPCs. This
1566 could leave the node unresponsive to lustre network traffic
1567 and manifested as a server ptllnd timeout.
1569 Severity : enhancement
1571 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1572 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1573 boolean to control whether to print peer debug info when a
1574 client's RPC times out.
1577 Frequency : only for kernels with patches from Lustre below 1.4.3
1579 Description: Remove old rdonly API
1580 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1583 Frequency : only for devices with external journals
1585 Description: Set external device read-only also
1586 Details : During a commanded failover stop, we set the disk device
1587 read-only while the server shuts down. We now also set any
1588 external journal device read-only at the same time.
1591 Frequency : when upgrading from 1.4 while trying to change parameters
1593 Description: The wrong (new) MDC name was used when setting parameters for
1594 upgraded MDT's. Also allows changing of OSC (and MDC)
1595 parameters if --writeconf is specified at tunefs upgrade time.
1598 Frequency : when setting specific ost indicies
1600 Description: QOS code breaks on skipped indicies
1601 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1602 created with --index need not be sequential.
1604 Severity : enhancement
1606 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1607 Details : The uninit_groups feature works in conjunction with the kernel
1608 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1609 up the pass1 processing of e2fsck. This is a read-only feature
1610 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1611 mount filesystems that have had this feature enabled.
1613 Severity : enhancement
1615 Description: Improve multi-block allocation algorithm to avoid fragmentation
1616 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1617 allocation locality and avoid filesystem fragmentation.
1619 ------------------------------------------------------------------------------
1621 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1623 * Support for kernels:
1624 2.4.21-47.0.1.EL (RHEL 3)
1625 2.6.5-7.283 (SLES 9)
1626 2.6.9-42.0.10.EL (RHEL 4)
1627 2.6.12.6 vanilla (kernel.org)
1628 2.6.16.27-0.9 (SLES 10)
1629 * Recommended e2fsprogs version: 1.39.cfs5
1631 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1635 Frequency : occasional, depends on client load and configuration
1636 Bugzilla : 12181, 12203
1637 Description: data loss for recently-modified files
1639 Details : In some cases it is possible that recently written or created
1640 files may not be written to disk in a timely manner (this should
1641 normally be within 30s unless client IO load is very high).
1642 The problem appears as zero-length files or files that are a
1643 multiple of 1MB in size after a client crash or client eviction
1644 that are missing data at the end of the file.
1646 This problem is more likely to be hit on clients where files are
1647 repeatedly created and unlinked in the same directory, clients
1648 have a large amount of RAM, have many CPUs, the filesystem has
1649 many OSTs, the clients are rebooted frequently, and/or the files
1650 are not accessed by other nodes after being written.
1652 The presence of the problem can be detected by looking at
1653 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1654 smaller than the second (nr_unused) then dirty files will not
1655 be flushed automatically to disk. "sync; sleep 10" should be
1656 run several times on the node before unmounting it to update
1657 Lustre (this is also safe to run on nodes without this problem).
1659 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1660 that can cause this same problem, so customers using that kernel
1661 also need to update the kernel in addition to Lustre. In order
1662 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1664 It is normal that files written just before a client crash (less
1665 than 30s) may not yet have been flushed to disk, even for local
1669 Frequency : frequent on thin XT3 nodes
1671 Description: UUID collision on thin XT3 Linux nodes
1672 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1673 randomly, since we relied on an insufficiently-seeded PRNG.
1678 Description: OSS hangs after "All ost request buffers busy"
1679 Details : A deadlock between quota and journal operations caused OSS
1680 hangs after printing "All ost request buffers busy."
1683 Frequency : always on liblustre builds
1685 Description: Cleanup compiler warnings on liblustre
1688 Frequency : always on liblustre builds on XT3
1690 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1694 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1695 > 32000 subdirectories
1696 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1697 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1698 subdirectory count crosses 32000. This will aid e2fsck to
1699 correctly handle more than 32000 subdirectories.
1702 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1704 Description: System hang on startup
1705 Details : This bug allowed the liblustre (e.g. catamount) client to
1706 return to the app before handling all startup RPCs. This
1707 could leave the node unresponsive to lustre network traffic
1708 and manifested as a server ptllnd timeout.
1710 Severity : enhancement
1712 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1713 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1714 boolean to control whether to print peer debug info when a
1715 client's RPC times out.
1720 Description: make O_SYNC working on 2.6 kernels
1721 Details : 2.6 kernels use different method for mark pages for write,
1722 so need add a code to lustre for O_SYNC work.
1727 Description: Failure to close file and release space on NFS
1728 Details : Put inode details into lock acquired in ll_intent_file_open.
1729 Use mdc_intent_lock in ll_intent_open to properly
1730 detect all kind of errors unhandled by mdc_enqueue.
1735 Description: proc file read during shutdown sometimes raced obd removal,
1737 Details : Add lock to prevent obd access after proc file removal.
1740 Frequency : Only for files larger than 4GB on 32-bit clients.
1742 Description: improperly doing page alignment of locks
1743 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1744 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1749 Description: RPCs being resent when they shouldn't be
1750 Details : Some RPCs that should not be resent are being resent. This
1751 can cause inconsistencies in the RPC state machine. Do not
1752 resend such requests.
1755 Frequency : rare, only with NFS export
1757 Description: Crash on NFS re-export node
1758 Details : under very unusual load conditions an assertion is hit in
1759 ll_intent_file_open()
1762 Frequency : only if OST filesystem is corrupted
1764 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1765 Details : for a short time RPCs with bulk IO are in the replay list,
1766 but replay of bulk IOs is unimplemented. If the OST filesystem
1767 is corrupted due to disk cache incoherency and then replay is
1768 started it is possible to trip an assertion. Avoid putting
1769 committed RPCs into the replay list at all to avoid this issue.
1774 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1775 Details : Large single O_DIRECT read and write calls can fail to allocate
1776 a sufficiently large buffer to process the request. In case of
1777 allocation failure the allocation is retried with a smaller
1778 buffer and broken into smaller requests.
1780 Severity : enhancement
1782 Description: Add -o localflock option to simulate old noflock behaviour.
1783 Details : This will achieve local-only flock/fcntl locks coherentness.
1788 Description: versioning check is incomplete
1789 Details : Checking the version difference of client vs. server, report
1790 error if the gap is too big.
1795 Description: add support PG_writeback bit
1796 Details : add support for PG_writeback bit for Lustre, for more carefull
1797 work with page cache in 2.6 kernel. This also fix some deadlocks
1798 and remove hack for work O_SYNC with 2.6 kernel.
1800 Severity : enhancement
1802 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1803 Details : The uninit_groups feature works in conjunction with the kernel
1804 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1805 up the pass1 processing of e2fsck. This is a read-only feature
1806 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1807 mount filesystems that have had this feature enabled.
1809 Severity : enhancement
1811 Description: Improve multi-block allocation algorithm to avoid fragmentation
1812 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1813 allocation locality and avoid filesystem fragmentation.
1815 ------------------------------------------------------------------------------
1817 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1819 * Support for kernels:
1820 2.6.9-42.0.3.EL (RHEL 4)
1821 2.6.5-7.276 (SLES 9)
1822 2.4.21-47.0.1.EL (RHEL 3)
1823 2.6.12.6 vanilla (kernel.org)
1824 2.6.16.21-0.8 (SLES10)
1825 * Recommended e2fsprogs version: 1.39.cfs2-0
1827 * The backwards-compatible /proc/sys/portals symlink has been removed
1828 in this release. Before upgrading, please ensure that you change
1829 any configuration scripts or /etc/sysctl.conf files that access
1830 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1831 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1832 in advance of the upgrade on any system running Lustre 1.4.6 or
1833 newer, since /proc/sys/lnet was added in that version.
1834 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1838 Frequency : only when quota is used
1840 Description: avoid scanning export list for quota master
1841 Details : Change the algorithms to avoid scanning export list in order
1842 to improve the efficiency.
1845 Frequency : MDS failover only, very rarely
1847 Description: "went back in time" messages on mds failover
1848 Details : The greatest transno may be lost when the current operation
1849 finishes with an error (transno==0) and the client's last_rcvd
1850 record is over-written. Save the greatest transno in the
1851 mds_last_transno for this case.
1854 Frequency : always for specific kernels and striping counts
1856 Description: client may get "Matching packet too big" without ACL support
1857 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1858 when trying to access files in certain configurations. The
1859 clients should in fact be denied when mounting because they do
1860 not understand ACLs.
1863 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1865 Description: many clients connecting with IO in progress causes connect timeouts
1866 Details : Avoid synchronous journal commits to avoid delays caused by many
1867 clients connecting/disconnecting when bulk IO is in progress.
1868 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1869 OST_IO_PORTAL to avoid delays behind potentially many pending
1873 Frequency : occasionally with multiple writers to a single file
1875 Description: shared writes to file may result in wrong size reported by stat()
1876 Details : Allow growing of kms when extent lock is cancelled
1879 Frequency : always with random mmap IO to multi-striped file
1881 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1882 Details : Only if the hole is at the end of OST object so that kms is too
1883 small. Fix is to increase kms accordingly in ll_nopage.
1886 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1888 Description: writes to a missing object would leak memory on the OST
1889 Details : If there is an inconsistency between the MDS and OST filesystems,
1890 such that the MDS references an object that doesn't exist, writes
1891 to that object will leak memory due to incorrect cleanup in the
1892 error handling path, eventually running out of memory on the OST.
1897 Description: Creating too long symlink causes lustre errors
1898 Details : Check symlink and name lengths before sending requests to MDS.
1901 Frequency : only if flock is enabled (not on by default)
1903 Description: posix locks not released on fd closure on 2.6.9+
1904 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1905 kernels, this caused such locks not to be released on fd close and
1906 then assertions on fs unmount about still used locks.
1909 Frequency : MDS failover only, very rarely
1911 Description: clients may get ASSERTION(granted_lock != NULL)
1912 Details : When request was taking a long time, and a client was resending
1913 a getattr by name lock request. The were multiple lock requests
1914 with the same client lock handle and
1915 mds_getattr_name->fixup_handle_for_resent_request found one of the
1916 lock handles but later failed with ASSERTION(granted_lock != NULL).
1921 Description: handle->h_buffer_credits > 0, assertion failure
1922 Details : h_buffer_credits is zero after truncate, causing assertion
1923 failure. This patch extends the transaction or creates a new
1927 Frequency : NFS re-export or patchless client
1928 Bugzilla : 11179, 10796
1929 Description: Crash on NFS re-export node (__d_move)
1930 Details : We do not want to hash the dentry if we don't have a lock.
1931 But if this dentry is later used in d_move, we'd hit uninitialised
1932 list head d_hash, so we just do this to init d_hash field but
1933 leave dentry unhashed.
1936 Frequency : NFS re-export or patchless client
1938 Description: NFS exports has problem with symbolic link
1939 Details : lustre client didn't properly install dentry when re-exported
1940 to NFS or running patchless client.
1943 Frequency : NFS re-export or patchless client
1945 Description: Various nfs/patchless fixes.
1946 Details : fixes reuse disconected alias for lookup process - this fixes
1947 warning "find_exported_dentry: npd != pd",
1948 fix permission error with open files at nfs.
1949 fix apply umask when do revalidate.
1952 Frequency : occasional
1954 Description: Crash on NFS re-export node
1955 Details : calling clear_page() on the wrong pointer triggered oops in
1956 generic_mapping_read().
1959 Frequency : rarely, using O_DIRECT IO
1961 Description: unaligned directio crashes client with LASSERT
1962 Details : check for unaligned buffers before trying any requests.
1965 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1967 Description: stale data returned from RAID cache
1968 Details : If only a small amount of IO is done to the RAID device before
1969 reading it again it is possible to get stale data from the RAID
1970 cache instead of reading it from disk.
1973 Frequency : always for sles10 kernel
1975 Description: sles10 support
1976 Details : ll_follow_link: compile fixes and using of nd_set_link
1977 under newer kernels.
1980 Frequency : depends on arch, kernel and compiler version, always on sles10
1983 Description: recursive or deep enough symlinks cause stack overflow
1984 Details : getting rid of large stack-allocated variable in
1988 Frequency : depends on hardware
1990 Description: lustre write performance loss in the SLES10 kernel
1991 Details : the performance loss is caused by using of write barriers in the
1992 ext3 code. The SLES10 kernel turns barrier support on by
1993 default. The fix is to undo that change for ldiskfs.
1995 ------------------------------------------------------------------------------
1997 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1999 * Support for kernels:
2000 2.6.9-42.0.3EL (RHEL 4)
2001 2.6.5-7.276 (SLES 9)
2002 2.4.21-47.0.1.EL (RHEL 3)
2003 2.6.12.6 vanilla (kernel.org)
2007 Frequency : quota enabled and large files being deleted
2009 Description: releasing more than 4GB of quota at once hangs OST
2010 Details : If a user deletes more than 4GB of files on a single OST it
2011 will cause the OST to spin in an infinite loop. Release
2012 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2017 Description: statfs data retrieved from /proc may be stale or zero
2018 Details : When reading per-device statfs data from /proc, in the
2019 {kbytes,files}_{total,free,avail} files, it may appear
2020 as zero or be out of date.
2023 Frequency : systems with MD RAID1 external journal devices
2025 Description: lconf's call to blkid is confused by RAID1 journal devices
2026 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2027 of returning all block devices that match the journal UUID.
2030 Frequency : always, for aggregate stripe size over 4GB
2032 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2033 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2034 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2037 Frequency : always if "lfs find" used on a local file/directory
2039 Description: "lfs find" segfaults if used on a local file/directory
2040 Details : The case where a directory component was not specified wasn't
2041 handled correctly. Handle this properly.
2044 Frequency : always on ppc64
2046 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2047 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2048 perform endianness conversion.
2051 Frequency : rarely (truncate to non-zero file size after write under load)
2052 Bugzilla : 10730, 10687
2053 Description: Files padded with zeros to next 4K multiple
2054 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2055 it is possible that files that are truncated to a non-zero size
2056 immediately after being written are filled with zero bytes beyond
2057 the truncated size. No file data is lost.
2059 Severity : enhancement
2061 Description: Allow recovery/failover for liblustre clients.
2062 Details : liblustre clients were unaware of failover configurations until
2065 Severity : enhancement
2067 Description: user file locks should fail when not mounting with flock option
2068 Details : Set up an error-returning stub in ll_file_operations.lock field
2069 to prevent incorrect behaviour when client is mounted without
2070 flock option. Also, set up properly f_op->flock field for
2074 Frequency : always on ia64
2076 Description: "lfs df" loops on printing out MDS statfs information
2077 Details : The obd_ioctl_data was not initialized and in some systems
2078 this caused a failure during the ioctl that did not return
2079 an error. Initialize the struct and return an error on failure.
2082 Frequency : SLES 9 only
2084 Description: Error of copying files with lustre special EAs as root
2085 Details : Client side always return success for setxattr call for lustre
2086 special xattr (currently only "trusted.lov").
2089 Frequency : rarely on clusters with both ia64+i386 clients
2091 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2092 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2093 clients are doing concurrent writes to the same file it is
2094 possible that the ia64 clients may LASSERT because the OST
2095 extent locks are not PAGE_SIZE aligned. Ensure that grown
2096 locks are always aligned on the request boundary.
2099 Frequency : specific use, occasional
2101 Description: Overwriting in use executable truncates on-disk binary image
2102 Details : If one node attempts to overwrite an executable in use by
2103 another node, we now correctly return ETXTBSY instead of
2104 truncating the file.
2106 Severity : enhancement
2108 Description: Async OSC create to avoid the blocking unnecessarily.
2109 Details : If a OST has no remain object, system will block on the creating
2110 when need to create a new object on this OST. Now, ways use
2111 pre-created objects when available, instead of blocking on an
2112 empty osc while others are not empty. If we must block, we block
2113 for the shortest possible period of time.
2118 Description: chmod on Lustre root is propagated to other clients
2119 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2120 invalid by the follow_mount time.
2123 Frequency : liblustre clients only
2125 Description: Race in 'instant cancel' lock handling could lead to such locks
2126 never to be granted in case of SMP MDS
2127 Details : Do not destroy not yet granted but cbpending locks in
2131 Frequency : replay/resend of open
2133 Description: non null lock assetion failure in mds_intent_policy
2134 Details : Trying to replay/resend lockless open requests resulted in
2135 mds_open() returning 0 with no lock. Now it sets a flag if
2136 a lock is going to be returned.
2138 Severity : enhancement
2140 Description: Checksum enhancements
2141 Details : New checksum enhancements allow for resending RPCs that failed
2144 Severity : enhancement
2146 Description: Tunables on number of dirty pages in cacche
2147 Details : Allow to set limit on number of dirty pages cached.
2152 Description: client crash on unmount - lock still has references
2153 Details : In some error handling cases it was possible to leak a lock
2154 reference on a client while accessing a file. This was not
2155 harmful to the client during operation, but would cause the
2156 client to crash when the filesystem is unmounted.
2159 Frequency : specific case, rare
2161 Description: ETXTBSY on mds though file not in use by client
2162 Details : ETXTBSY is no longer incorrectly returned when attempting to
2163 chmod or chown a directory that the user previously tried to
2164 execute or a currently-executing binary.
2167 Frequency : extremely rare except on liblustre-based clients
2169 Description: Lustre space not freed when files are deleted
2170 Details : Clean up open-unlinked files after client eviction. Previously
2171 the unlink was skipped and the files remained as orphans.
2176 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2177 Details : In some cases it was possible to send a blocking callback to
2178 a client doing a glimpse, even though that client didn't get
2179 a lock granted. When the glimpse lock is cancelled on the OST
2180 the freed lock is left on the waiting list and corrupted the list.
2183 Frequency : all core dumps
2185 Description: Broke core dumps to lustre
2186 Details : Negative dentry may be unhashed if parent does not have UPDATE
2187 lock, but some callers, e.g. do_coredump, expect dentry to be
2188 hashed after successful create, hash it in ll_create_it.
2190 ------------------------------------------------------------------------------
2192 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2194 * Support for kernels:
2195 2.6.9-42.0.2.EL (RHEL 4)
2196 2.6.5-7.276 (SLES 9)
2197 2.4.21-40.EL (RHEL 3)
2198 2.6.12.6 vanilla (kernel.org)
2202 Frequency : always on RHEL 3
2204 Description: Number of open files grows over time
2205 Details : The number of open files grows over time, whether or not
2206 Lustre is started. This was due to a filp leak introduced
2207 by one of our kernel patches.
2209 ------------------------------------------------------------------------------
2211 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2213 * Support for kernels:
2214 2.6.9-42.EL (RHEL 4)
2215 2.6.5-7.276 (SLES 9)
2216 2.4.21-40.EL (RHEL 3)
2217 2.6.12.6 vanilla (kernel.org)
2222 Bugzilla : 5719, 9635, 9792, 9684
2223 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2224 Details : If a server is under heavy load and cannot reply to new
2225 connection requests before the client resends the (re)connect,
2226 the connection handling code can behave badly if two service
2227 threads are concurrently handing separate (re)connections from
2228 the same client. Add better locking to the connection handling
2229 code, and ensure that only a single connection will be processed
2230 for a given client UUID, even if the lock is dropped.
2232 Severity : enhancement
2234 Description: add TCP zero-copy support to kernel
2235 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2236 sends if the hardware supports scatter-gather and checksumming.
2237 This allows socklnd to do client-write and server-read more
2238 efficiently and reduce CPU utilization from skbuf copying.
2241 Frequency : only if NFS exporting from client
2243 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2244 Details : If setting attributes on a file created under NFS that had
2245 never been opened it would be possible to oops the client
2246 if the file had no objects.
2249 Frequency : always for liblustre
2251 Description: liblustre client does MDS+OSTs setattr RPC for each write
2252 Details : When doing a write from a liblustre client, the client
2253 incorrectly issued an RPC to the MDS and each OST the file was
2254 striped over in order to update the timestamps. When writing
2255 with small chunks and many clients this could overwhelm the MDS
2256 with RPCs. In all cases it would slow down the write because
2257 these RPCs are unnecessary.
2259 Severity : enhancement
2261 Description: allow number of MDS service threads to be changed at module load
2262 Details : It is now possible to change the number of MDS service threads
2263 running. Adding "options mds mds_num_threads={N}" to the MDS's
2264 /etc/modprobe.conf will set the number of threads for the next
2265 time Lustre is restarted (assuming the "mds" module is also
2266 reloaded at that time). The default number of threads will
2267 stay the same, 32 for most systems.
2272 Description: OST crash if filesystem is unformatted or corrupt
2273 Details : If an OST is started on a device that has never been formatted
2274 or if the filesystem is corrupt and cannot even mount then the
2275 error handling cleanup routines would dereference a NULL pointer.
2280 Description: NULL pointer deref in llap_from_page.
2281 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2282 incorrect) mapping if the page was truncated/reclaimed while it was
2283 searched for. Check for this condition and skip such pages when
2284 doing readahead. Introduce extra check to llap_from_page() to
2285 verify page->mapping->host is non-NULL (so page is not anonymous).
2288 Frequency : Sometimes when using sys_sendfile
2290 Description: "page not covered by a lock" warnings from ll_readpage
2291 Details : sendfile called ll_readpage without right page locks present.
2292 Now we introduced ll_file_sendfile that does necessary locking
2293 around call to generic_file_sendfile() much like we do in
2297 Frequency : with certain MDS communication failures at client mount time
2299 Description: NULL pointer deref after failed client mount
2300 Details : a client connection request may delayed by the network layer
2301 and not be sent until after the PTLRPC layer has timed out the
2302 request. If the client fails the mount immediately it will try
2303 to clean up before the network times out the request. Add a
2304 reference from the request import to the obd device and delay
2305 the cleanup until the network drops the request.
2308 Frequency : occasionally during client (re)connect
2310 Description: assertion failure during client (re)connect
2311 Details : processing a client connection request may be delayed by the
2312 client or server longer than the client connect timeout. This
2313 causes the client to resend the connection request. If the
2314 original connection request is replied in this interval, the
2315 client may trip an assertion failure in ptlrpc_connect_interpret()
2316 which thought it would be the only running connect process.
2319 Frequency : only with obd_echo servers and clients that are rebooted
2321 Description: kernel BUG accessing uninitialized data structure
2322 Details : When running an obd_echo server it did not start the ping_evictor
2323 thread, and when a client was evicted an uninitialized data
2324 structure was accessed. Start the ping_evictor in the RPC
2325 service startup instead of the OBD startup.
2327 Severity : enhancement
2328 Bugzilla : 10193 (patchless)
2329 Description: Remove dependency on various unexported kernel interfaces.
2330 Details : No longer need reparent_to_init, exit_mm, exit_files,
2331 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2334 Frequency : rare (only users of deprecated and unsupported LDAP config)
2336 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2337 Details : LDAP apparently contains 'lustreName' attributes instead of
2338 'name'. A simple remapping of the name is sufficient.
2341 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2343 Description: waiting_locks_callback trips kernel BUG if client is evicted
2344 Details : Running with the dump_on_timeout debug flag turned on makes
2345 it possible that the waiting_locks_callback() can try to dump
2346 the Lustre kernel debug logs from an interrupt handler. Defer
2347 this log dumping to the expired_lock_main() thread.
2349 Severity : enhancement
2351 Description: Support NFS exporting on 2.6 kernels.
2352 Details : Implement non-rawops metadata methods for NFS server to use without
2353 changing NFS server code.
2356 Frequency : very rare (synthetic metadata workload only)
2358 Description: two racing renames might cause an MDS thread to deadlock
2359 Details : Running the "racer" program may cause one MDS thread to rename
2360 a file from being the source of a rename to being the target of
2361 a rename at exactly the same time that another thread is doing
2362 so, and the second thread has already enqueued these locks after
2363 doing a lookup of the target and is trying to relock them in
2364 order. Ensure that we don't try to re-lock the same resource.
2367 Frequency : only very large systems with liblustre clients
2369 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2370 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2371 all OSTs in parallel. This allows the eviction of stale liblustre
2372 clients to proceed much faster than if they were done in series,
2373 and also offers similar improvements for other set_info RPCs.
2378 Description: excessive CPU usage during initial read phase on client
2379 Details : During the initial read phase on a client, it would agressively
2380 retry readahead on the file, consuming too much CPU and impacting
2381 performance (since 1.4.5.8). Improve the readahead algorithm
2382 to avoid this, and also improve some other common cases (read
2383 of small files in particular, where "small" is files smaller than
2384 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2389 Description: MDS crash when receiving packet with unknown intent.
2390 Details : Do not LBUG in unknown intent case, just return -EFAULT
2392 Severity : enhancement
2393 Bugzilla : 9293, 9385
2394 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2395 Details : Do not serialize getattr (non-intent version) and statfs.
2398 Frequency : occasional, when OST network is overloaded/intermittent
2400 Description: client evicted by OST after bulk IO timeout
2401 Details : If a client sends a bulk IO request (read or write) the OST
2402 may evict the client if it is unresposive to its data GET/PUT
2403 request. This is incorrect if the network is overloaded (takes
2404 too long to transfer the RPC data) or dropped the OST GET/PUT
2405 request. There is no need to evict the client at all, since
2406 the pinger and/or lock callbacks will handle this, and the
2407 client can restart the bulk request.
2410 Frequency : Always when mmapping file with no objects
2412 Description: client crashes when mmapping file with no objects
2413 Details : Check that we actually have objects in a file before doing any
2414 operations on objects in ll_vm_open, ll_vm_close and
2420 Description: Request leak when working with deleted CWD
2421 Details : Introduce advanced request refcount tracking for requests
2422 referenced from lustre intent.
2424 Severity : Enhancement
2426 Description: Cache open file handles on client.
2427 Details : MDS now will return special lock along with openhandle, if
2428 requested and client is allowed to hold openhandle, even if unused,
2429 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2430 closing files for every read/write openration.
2432 Severity : Enhancement
2434 Description: Cache open negative dentries on client when possible.
2435 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2436 negative dentries on lock revocation.
2441 Description: Remounting a client read-only wasn't possible with a zconf mount
2442 Details : It wasn't possible to remount a client read-only with llmount.
2444 Severity : enhancement
2445 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2446 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2447 MPICH is included to add Lustre-specific ADIO interfaces.
2448 This is based closely on the UFS ADIO layer and only differs
2449 in file creation, in order to allow the OST striping to be set.
2450 This is user-contributed code and not supported by CFS.
2455 Description: extended inode attributes (immutable, append-only) work improperly
2456 when 2.4 and 2.6 kernels are used on client/server or vice versa
2457 Details : Introduce kernel-independent values for these flags.
2459 Severity : enhancement
2462 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2463 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2464 granularity. Now, Fractional MB granularity is be supported,
2465 this is very useful for low memory system.
2467 Severity : enhancement
2469 Description: Getattr by fid
2470 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2471 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2472 have locks and avoids getattr rpc requests for it on every stat.
2475 Frequency : Always, for filesystems larger than 2TB
2477 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2478 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2479 filesystems larger than 2TB to be mounted with mballoc enabled.
2482 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2484 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2485 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2486 filesystems larger than 2TB is mounted then the kernel will
2487 silently corrupt the start of the filesystem. CONFIG_LBD is
2488 enabled for all CFS-supported kernels, but the possibility of
2489 this happening with a modified kernel config exists.
2491 Severity : enhancement
2493 Description: add client O_DIRECT support for 2.6 kernels
2494 Details : It is now possible to do O_DIRECT reads and writes to files
2495 in the Lustre client mountpoint on 2.6 kernel clients.
2497 Severity : enhancement
2499 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2500 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2501 parallel, not waiting for response from every OST before sending
2502 a rpc to the next OST.
2507 Description: setattr vs write race when updating file timestamps
2508 Details : Client processes that update a file timestamp into the past
2509 right after writing to the file (e.g. tar) it is possible that
2510 the updated file modification time can be reset to the current
2511 time due to a race between processing the setattr and write RPC.
2513 Severity : enhancement
2515 Description: Bring 'lfs find' closer in line with regular Linux find.
2516 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2517 -print0 options and obtains all the needed info through the lustre
2520 Severity : enhancement
2522 Description: support up to 1024 configured devices on one node
2523 Details : change obd_dev array from statically allocated to dynamically
2524 allocated structs as they are first used to reduce memory usage
2529 Description: Flush dirty partially truncated pages during truncate
2530 Details : Immediatelly flush partially truncated pages in filter_setattr,
2531 this way we completely avoid having any pages in page cache on OST
2532 and can retire ugly workarounds during writes to flush such pages.
2537 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2538 Details : protect inode from truncation within vfs_unlink() context
2539 just take a reference before calling vfs_unlink() and release it
2540 when parent's i_sem is free.
2545 Description: last_id value checked outside lock on OST caused LASSERT failure
2546 Details : If there were multiple MDS->OST object precreate requests in
2547 flight, it was possible that the OST's last object id was checked
2548 outside a lock and incorrectly tripped an assertion. Move checks
2549 inside locks, and discard old precreate requests.
2552 Frequency : always, if extents are used on OSTs
2554 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2555 Details : OSTs using the extents format would not zero the high 16 bits of
2556 the index physical block number. This is not a problem for any
2557 OST filesystems smaller than 16TB, and no kernels support ext3
2558 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2559 new/modified files) and can be fixed for existing filesystems
2560 with e2fsprogs-1.39-cfs1.
2565 Description: import connection selection may be incorrect if timer wraps
2566 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2567 import connections to be ignored if the 32-bit jiffies counter
2568 wraps. Use a 64-bit jiffies counter.
2571 Frequency : very large clusters immediately after boot
2573 Description: LNET request buffers exhausted under heavy short-term load
2574 Details : If a large number of client requests are generated on a service
2575 that has previously never seen so many requests it is possible
2576 that the request buffer growth cannot keep up with the spike in
2577 demand. Instead of dropping incoming requests, they are held in
2578 the LND until the RPC service can accept more requests.
2581 Frequency : Sometimes during replay
2583 Description: Assertion failure in ll_local_open after replay.
2584 Details : If replay happened on an open request reply before we were able
2585 to set replay handler, reply will become not swabbed tripping the
2586 assertion in ll_local_open. Now we set the handler right after
2587 recognising of open request
2590 Frequency : very rare
2592 Description: kernel reports "badness in vsnprintf"
2593 Details : Reading from the "recovery_status" /proc file in small chunks
2594 may cause a negative length in lprocfs_obd_rd_recovery_status()
2595 call to vsnprintf() (which is otherwise harmless). Exit early
2596 if there is no more space in the output buffer.
2598 Severity : enhancement
2600 Description: clear OBD RPC statistics by writing to them
2601 Details : It is now possible to clear the OBD RPC statistics by writing
2602 to the "stats" file.
2607 Description: Client mtime is not the same on different clients after utimes
2608 Details : In some cases, the client was using the utimes() syscall on
2609 a file cached on another node. The clients now validate the
2610 ctime from the MDS + OSTs to determine which one is right.
2615 Description: Inability to activate failout mode
2616 Details : lconf script incorrectly assumed that in pythong string's numeric
2617 value is used in comparisons.
2620 Frequency : always with multiple stripes per file
2622 Description: Inefficient object allocation for mutli-stripe files
2623 Details : When selecting which OSTs to stripe files over, for files with
2624 a stripe count that divides evenly into the number of OSTs,
2625 the MDS is always picking the same starting OST for each file.
2626 Return the OST selection heuristic to the original design.
2631 Description: mount failures may take full timeout to return an error
2632 Details : Under some heavy load conditions it is possible that a
2633 failed mount can wait for the full obd_timeout interval,
2634 possibly several minutes, before reporting an error.
2635 Instead return an error as soon as the status is known.
2637 Frequency : quota enabled and large files being deleted
2639 Description: releasing more than 4GB of quota at once hangs OST
2640 Details : If a user deletes more than 4GB of files on a single OST it
2641 will cause the OST to spin in an infinite loop. Release
2642 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2647 Description: statfs data retrieved from /proc may be stale or zero
2648 Details : When reading per-device statfs data from /proc, in the
2649 {kbytes,files}_{total,free,avail} files, it may appear
2650 as zero or be out of date.
2653 Frequency : systems with MD RAID1 external journal devices
2655 Description: lconf's call to blkid is confused by RAID1 journal devices
2656 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2657 of returning all block devices that match the journal UUID.
2660 Frequency : always, for aggregate stripe size over 4GB
2662 Description: assertion fails when trying to use 4GB stripe size
2663 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2664 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2667 Frequency : always on ppc64
2669 Description: the first write on an ext3 filesystem with mballoc got stuck
2670 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2671 perform endianness conversion.
2673 ------------------------------------------------------------------------------
2675 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2677 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2678 INTEROPERATE with older versions automatically. Please read the
2679 user documentation before upgrading any part of a live system.
2680 * WARNING: Lustre networking configuration changes are required with
2681 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2684 * Support for kernels:
2685 2.6.9-22.0.2.EL (RHEL 4)
2686 2.6.5-7.244 (SLES 9)
2687 2.6.12.6 vanilla (kernel.org)
2690 Severity : enhancement
2691 Bugzilla : 7981/8208
2692 Description: Introduced Lustre Networking (LNET)
2693 Details : LNET is new networking infrastructure for Lustre, it includes
2694 a reorganized network configuration mode (see the user
2695 documentation for full details) as well as support for routing
2696 between different network fabrics. Lustre Networking Devices
2697 (LNDs) for the supported network fabrics have also been
2698 created for this new infrastructure.
2700 Severity : enhancement
2701 Description: Introduced Access control lists
2702 Details : clients can set ACLs on files and directories in order to have
2703 more fine-grained permissions than the standard Unix UGO+RWX.
2704 The MDS must be started with the "-o acl" mount option.
2706 Severity : enhancement
2707 Description: Introduced filesystem quotas
2708 Details : Administrators may now establish per-user quotas on the
2711 Severity : enhancement
2713 Description: Configuration change for the XT3
2714 The PTLLND is now used to run Lustre over Portals on the XT3
2715 The configure option(s) --with-cray-portals are no longer used.
2716 Rather --with-portals=<path-to-portals-includes> is used to
2717 enable building on the XT3. In addition to enable XT3 specific
2718 features the option --enable-cray-xt3 must be used.
2723 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2724 Details : A race between allocating a new llcd and re-getting the llcd_lock
2725 allowed another thread to grab newly-allocated llcd.
2727 Severity : enhancement
2729 Description: 2.6 OST async journal commit and locking fix to improve performance
2730 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2731 2.6 kernels are now async as they already were in 2.4 kernels so
2732 that they can commit concurrently with the network bulk transfer.
2733 For block-allocated files the filter allocation semaphore is held
2734 to avoid filesystem fragmentation during allocation. BKL lock
2735 removed for 2.6 xattr operations where it is no longer needed.
2740 Description: lconf incorrectly determined whether two IP networks could talk
2741 Details : In some more complicated routing and multiple-network
2742 configurations, lconf will avoid trying to make a network
2743 connection to a disjoint part of the IP space. It was doing the
2744 math incorrectly for one set of cases.
2749 Description: Fix for potential infinite loop processing records in an llog.
2750 Details : If an llog record is corrupted/zeroed, it is possible to loop
2751 forever in llog_process(). Validate the llog record length
2752 and skip the remainder of the block on error.
2755 Frequency : occasional (liblustre only)
2757 Description: liblustre could not open files whose last component is a symlink
2758 Details : sysio_path_walk() would incorrectly pass the open intent to
2759 intermediate path components.
2762 Frequency : rare (liblustre only with non-standard tuning)
2763 Bugzilla : 7201 (7350)
2764 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2765 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2766 after it has been released, only do this for referenced locks
2768 Severity : enhancement
2770 Description: specifying an (invalid) directory default stripe_size of -1
2771 would reset the directory default striping
2772 Details : stripe_size -1 was used internally to signal directory stripe
2773 removal, now use "all default" to signal dir stripe removal
2774 as a directory striping of "all default" is not useful
2777 Frequency : common for large clusters running liblustre clients
2779 Description: doing an ls when liblustre clients are running is slow
2780 Details : sending a glimpse AST to a liblustre client waits for every AST
2781 to time out, as liblustre clients will not respond. Since they
2782 cannot cache data we refresh the OST lock LVB from disk instead.
2784 Severity : enhancement
2786 Description: doing an ls at the same time as file IO can be slow
2787 Details : enqueue and other "small" requests can be blocked behind many
2788 large IO requests. Create a new OST IO portal for non-IO
2789 requests so they can be processed faster.
2792 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2794 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2795 created file with mode 000
2796 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2797 allow an ftruncate() as a non-root user to a file with mode 000.
2798 Lustre now respects this flag to disable mode checks when
2799 truncating a file owned by the user
2802 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2805 Description: Revoking locks from clients that went dead or catatonic might take
2807 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2808 cancellation of such locks instant on servers without waiting for
2809 any reply from clients. Clients drops these locks when cancel
2810 notification from server is received without replying.
2813 Frequency : liblustre-only, when liblustre client dies or becomes busy
2815 Description: Doing ls on Linux clients can take a long time with active
2817 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2818 granting such locks to it in the first place if possible. Locks
2819 are taken by proxy on the OST during the read or write and
2820 dropped immediately afterward. Add connect flags handling, do
2821 not grant locks to liblustre clients for glimpse ASTs.
2823 Severity : enhancement
2825 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2826 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2827 sequential reads and behaves badly with applications doing random
2828 reads. Improve it by reading ahead at least read region, and
2829 avoiding excessive large RPC for small reads.
2831 Severity : enhancement
2833 Description: Creating more than 1000 files for a single job may cause a load
2834 imbalance on the OSTs if there are also a large number of OSTs.
2835 Details : qos_prep_create() uses an OST index reseed value that is an
2836 even multiple of the number of available OSTs so that if the
2837 reseed happens in the middle of the object allocation it will
2838 still utilize the OSTs as uniformly as possible.
2843 Description: OST or MDS may oops in ping_evictor_main()
2844 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2845 but doesn't restart at beginning of obd_exports_timed list
2848 Severity : enhancement
2850 Description: improve by-nid export eviction on the MDS and OST
2851 Details : allow multiple exports with the same NID to be evicted at one
2852 time without re-searching the exports list.
2855 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2857 Description: MDS may oops in groups_free()
2858 Details : in rare race conditions a newly allocated group_info struct is
2859 freed again, and this can be NULL. The 2.4 compatibility code
2860 for groups_free() checked for a NULL pointer, but 2.6 did not.
2863 Frequency : common for liblustre clients doing little filesystem IO
2864 Bugzilla : 9352, 7313
2865 Description: server may evict liblustre clients accessing contended locks
2866 Details : if a client is granted a lock or receives a completion AST
2867 with a blocking AST already set it would not reply to the AST
2868 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2871 Frequency : lfs setstripe, only systems with more than 160 OSTs
2873 Description: unable to set striping with a starting offset beyond OST 160
2874 Details : llapi_create_file() incorrectly limited the starting stripe
2875 index to the maximum single-file stripe count.
2878 Frequency : LDAP users only
2880 Description: lconf did not handle in-kernel recovery with LDAP properly
2881 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2883 Severity : enhancement
2885 Description: bind OST threads to NUMA nodes to improve performance
2886 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2887 node and do their allocations there to localize memory access
2889 Severity : enhancement
2891 Description: llmount can determine client NID directly from Myrinet (GM)
2892 Details : the client NID code from gmnalnid was moved directly into
2893 llmount, removing the need to use this or specifying the
2894 client NID explicitly when mounting GM clients with zeroconf
2897 Frequency : if client is started with down MDS
2899 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2900 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2901 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2902 ptlrpc_import_delayed_req() because we didn't check if the
2903 request was interrupted, and we also didn't break out of the
2904 event loop if there was no timeout
2909 Description: data loss during non-page-aligned writes to a single file from
2910 both multiple nodes and multiple threads on one node at same time
2911 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2912 inconsistency led to false short-reads, that were cached and later
2913 used by ->prepare_write() to fill in partially written page,
2914 leading to data loss.
2917 Frequency : always, if lconf --abort_recovery used
2919 Description: lconf --abort_recovery fails with 'Operation not supported'
2920 Details : lconf was attempting to abort recovery on the MDT device and not
2923 Severity : enhancement
2925 Description: remove cleanup logs
2926 Details : replace lconf-generated cleanup logs with lustre internal
2927 cleanup routines. Eliminates the need for client-cleanup and
2930 Severity : enhancement
2932 Description: add support for EAs (user and system) on lustre filesystems
2933 Details : it is now possible to store extended attributes in the Lustre
2934 client filesystem, and with the user_xattr mount option it
2935 is possible to allow users to store EAs on their files also
2937 Severity : enhancement
2939 Description: Add possibility (config option) to show minimal available OST free
2941 Details : When compiled with --enable-mindf configure option, statfs(2)
2942 (and so, df) will return least minimal free space available from
2943 all OSTs as amount of free space on FS, instead of summary of
2944 free spaces of all OSTs.
2946 Severity : enhancement
2948 Description: do not expand extent locks acquired on OST-side
2949 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2950 by server: they are not cached anyway.
2953 Frequency : when mmap is used/binaries executed from Lustre
2955 Description: Unmmap pages before throwing them away from read cache.
2956 Details : llap_shrink cache now attempts to unmap pages before discarding
2957 them (if unmapping failed - do not discard). SLES9 kernel has
2958 extra checks that trigger if this unmapping is not done first.
2963 Description: lconf didn't resolve symlinks before checking to see whether a
2964 given mountpoint was already in use
2967 Frequency : when migrating failover services
2968 Bugzilla : 6395, 9514
2969 Description: When migrating a subset of services from a node (e.g. failback
2970 from a failover service node) the remaining services would
2971 time out and evict clients.
2972 Details : lconf --force (implied by --failover) sets the global obd_timeout
2973 to 5 seconds in order to quickly disconnect, but this caused
2974 other RPCs to time out too quickly. Do not change the global
2975 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2977 Severity : enhancement
2978 Frequency : if MDS is started with down OST
2979 Bugzilla : 9439,5706
2980 Description: Allow startup/shutdown of an MDS without depending on the
2981 availability of the OSTs.
2982 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2983 Add appropriate locking and lov-osc refcounts for safe
2984 cleaning. Add osc abort_inflight calls in case the
2985 synchronize never started.
2988 Frequency : occasional (Cray XT3 only)
2990 Description: root not authorized to access files in CRAY_PORTALS environment
2991 Details : The client process capabilities were not honoured on the MDS in
2992 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2993 been accessed by an authorized user then root was able to access
2994 the file on the local client also. The root user capabilities
2995 are now allowed on the MDS, as this environment has secure UID.
2998 Frequency : occasional
3000 Description: ldiskfs "too long searching" message happens too often
3001 Details : A debugging message (otherwise harmless) prints too often on
3002 the OST console. This has been reduced to only happen when
3003 there are fragmentation problems on the filesystem.
3008 Description: Division by zero in statfs when all OSCs are inactive
3009 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3010 lov_statfs divides by value returned by lov_get_stripecnt().
3014 Bugzilla : 9489, 3273
3015 Description: First write from each client to each OST was only 4kB in size,
3016 to initialize client writeback cache, which caused sub-optimal
3017 RPCs and poor layout on disk for the first writen file.
3018 Details : Clients now request an initial cache grant at (re)connect time
3019 and so that they can start streaming writes to the cache right
3020 away and always do full-sized RPCs if there is enough data.
3021 If the OST is rebooted the client also re-establishes its grant
3022 so that client cached writes will be honoured under the grant.
3027 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3028 Details : Now I/O RPCs go to different portal number and (presumably) fast
3029 lock requests (and glimses) and other RPCs get their own service
3030 threads pool that should be able to service those RPCs
3033 Severity : enhancement
3035 Description: Ability to exchange lustre version between client and servers and
3036 issue warnings at client side if client is too old. Also for
3037 liblustre clients there is ability to refuse connection of too old
3039 Details : New 'version' field is added to connect data structure that is
3040 filled with version info. That info is later checked by server and
3044 Frequency : rare, liblustre only.
3045 Bugzilla : 9296, 9581
3046 Description: Two simultaneous writes from liblustre at offset within same page
3047 might proceed at the same time overwriting eachother with stale
3049 Details : I/O lock withing llu_file_prwv was released too early, before data
3050 actually was hitting the wire. Extended lock-holding time until
3051 server acknowledges receiving data.
3054 Frequency : extremely rare. Never observed in practice.
3056 Description: avoid generating lustre_handle cookie of 0.
3057 Details : class_handle_hash() generates handle cookies by incrementing
3058 global counter, and can hit 0 occasionaly (this is unlikely, but
3059 not impossible, because initial value of cookie counter is
3060 selected randonly). Value of 0 is used as a sentinel meaning
3061 "unassigned handle" --- avoid it. Also coalesce two critical
3062 sections in this function into one.
3064 Severity : enhancement
3066 Description: allow liblustre clients to delegate truncate locking to OST
3067 Details : To avoid overhead of locking, liblustre client instructs OST to
3068 take extent lock in ost_punch() on client's behalf. New connection
3069 flag is added to handle backward compatibility.
3071 Severity : enhancement
3072 Bugzilla : 4928, 7341, 9758
3073 Description: allow number of OST service threads to be specified
3074 Details : a module parameter allows the number of OST service threads
3075 to be specified via "options ost ost_num_threads={N}" in the
3076 OSS's /etc/modules.conf or /etc/modprobe.conf.
3080 Bugzilla : 6146, 9635, 9895
3081 Description: servers crash with bad pointer in target_handle_connect()
3082 Details : In rare cases when a client is reconnecting it was possible that
3083 the connection request was the last reference for that export.
3084 We would temporarily drop the export reference and get a new
3085 one, but this may have been the last reference and the export
3086 was just destroyed. Get new reference before dropping old one.
3088 Severity : enhancement
3089 Frequency : if client is started with failover MDS
3091 Description: Allow multiple MDS hostnames in the mount command
3092 Details : Try to read the configuration from all specified MDS
3093 hostnames during a client mount in case the "primary"
3096 Severity : enhancement
3098 Description: Stop sending data to evicted clients as soon as possible.
3099 Details : Check if the client we are about to send or are sending data to
3100 was evicted already. (Check is done every second of waiting,
3101 for which l_wait_event interface was extended to allow checking
3102 of exit condition at specified intervals).
3105 Frequency : rare, normally only when NFS exporting is done from client
3107 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3109 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3112 Frequency : always, if srand() is called before liblustre initialization
3114 Description: Liblustre uses system PRNG disturbing its usage by user application
3115 Details : Introduce internal to lustre fast and high-quality PRNG for
3116 lustre usage and make liblustre and some other places in generic
3117 lustre code to use it.
3119 Severity : enhancement
3120 Bugzilla : 9477, 9557, 9870
3121 Description: Verify that the MDS configuration logs are updated when xml is
3122 Details : Check if the .xml configuration logs are newer than the config
3123 logs stored on the MDS and report an error if this is the case.
3124 Request --write-conf, or allow starting with --old_conf.
3126 Severity : enhancement
3128 Description: Handle symlinks in the path when checking if Lustre is mounted.
3129 Details : Resolve intermediate symlinks when checking if a client has
3130 mounted a filesystem to avoid duplicate client mounts.
3135 Description: lconf can hit an error exception but still return success.
3136 Details : The lconf command catches the Command error exception at the top
3137 level script context and will exit with the associated exit
3138 status, but doesn't ensure that this exit status is non-zero.
3143 Description: failure of ptlrpc thread startup can cause oops
3144 Details : Starting a ptlrpc service thread can fail if there are a large
3145 number of threads or the server memory is very fragmented.
3146 Handle this without oopsing.
3149 Frequency : always, only if liblustre and non-default acceptor port was used
3151 Description: liblustre cannot connect to servers with non-default acceptor port
3152 Details : tcpnal_set_default_params() was not called and was therefore
3153 ignoring the environment varaible TCPNAL_PORT, as well as other
3154 TCPNAL_ environment variables
3159 Description: two objects could be created on the same OST for a single file
3160 Details : If an OST is down, in some cases it was possible to create two
3161 objects on a single OST for a single file. No problems other
3162 than potential performance impact and spurious error messages.
3166 Bugzilla : 5681, 9562
3167 Description: Client may oops in ll_unhash_aliases
3168 Details : Client dcache may become inconsistent in race condition.
3169 In some cases "getcwd" can fail if the current directory is
3175 Description: Inode refcounting problems in NFS export code
3176 Details : link_raw functions used to call d_instantiate without obtaining
3177 extra inode reference first.
3181 Bugzilla : 9942, 9903
3182 Description: Referencing freed requests leading to crash, memleaks with NFS.
3183 Details : We used to require that call to ll_revalidate_it was always
3184 followed by ll_lookup_it. Also with revalidate_special() it is
3185 possible to call ll_revalidate_it() twice for the same dentry
3186 even if first occurence returned success. This fix changes semantic
3187 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3188 reference on a request referred from the intent.
3189 ll_intent_release() then releases such a request.
3192 Frequency : rare, normally benchmark loads only
3194 Description: unlinked inodes were kept in memory on the client
3195 Details : If a client is repeatedly creating and unlinking files it
3196 can accumulate a lot of stale inodes in the inode slab cache.
3197 If there is no other client load running this can cause the
3198 client node to run out of memory. Instead flush old inodes
3199 from client cache that have the same inode number as a new inode.
3202 Frequency : SLES9 2.6.5 kernel and long filenames only
3203 Bugzilla : 9969, 10379
3204 Description: utime reports stale NFS file handle
3205 Details : SLES9 uses out-of-dentry names in some cases, which confused
3206 the lustre dentry revalidation. Change it to always use the
3210 Frequency : rare, unless heavy write-truncate concurrency is continuous
3211 Bugzilla : 4180, 6984, 7171, 9963, 9331
3212 Description: OST becomes very slow and/or deadlocked during object unlink
3213 Details : filter_destroy() was holding onto the parent directory lock
3214 while truncating+unlinking objects. For very large objects this
3215 may block other threads for a long time and slow overall OST
3216 responsiveness. It may also be possible to get a lock ordering
3217 deadlock in this case, or run out of journal credits because of
3218 the combined truncate+unlink. Solution is to do object truncate
3219 first in one transaction without parent lock, and then do the
3220 final unlink in a new transaction with the parent lock. This
3221 reduces the lock hold time dramatically.
3224 Frequency : rare, 2.4 kernels only
3226 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3227 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3228 before unmounting their filesystem in order to avoid deadlock.
3229 The kernel_locked() function in 2.4 kernels only checks whether
3230 the kernel lock is held, not whether it is this process that is
3231 holding it as 2.6 kernels do.
3236 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3237 Details : The client ptlrpc code may be trying to reconnect to a down
3238 server before a previous connection attempt has timed out.
3239 Increase the reconnect interval to be longer than the connection
3240 timeout interval to avoid sending duplicate connections to
3244 Frequency : echo_client brw_test command
3246 Description: fix echo_client to work with OST preallocated code
3247 Details : OST preallocation code (5137) didn't take echo_client IO path
3248 into account: echo_client calls filter methods outside of any
3249 OST thread and, hence, there is no per-thread preallocated
3250 pages and buffers to use. Solution: hijack pga pages for IO. As
3251 a byproduct, this avoids unnecessary data copying.
3255 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3256 Description: Client can oops in mdc_commit_close() after open replay
3257 Details : It was possible for the MDS to return an open request with no
3258 transaction number in mds_finish_transno() if the client was
3259 evicted, but without actually returning an error. Clients
3260 would later try to replay that open and may trip an assertion
3261 Simplify the client close codepath, and always return an error
3262 from the MDS in case the open is not successful.
3265 Frequency : rare, 2.6 OSTs only
3267 Description: OST may deadlock under high load on fragmented files
3268 Details : If there was a heavy load and highly-fragmented OST filesystems
3269 it was possible to have all the OST threads deadlock waiting on
3270 allocation of biovecs, because the biovecs were not released
3271 until the entire RPC IO was completed. Instead, release biovecs
3272 as soon as they are complete to ensure forward IO progress.
3274 Severity : enhancement
3276 Description: Support for specifying external journal device at mount
3277 Details : If an OST or MDS device is formatted with an external journal
3278 device, this device major/minor is stored in the ext3 superblock
3279 and may not be valid for failover. Allow detecting and
3280 specifying the external journal at mount time.
3285 Description: Mounting an MDS with pending unlinked files may cause oops
3286 Details : target_finish_recovery() calls mds_postrecov() which returned
3287 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3288 considers this an error and immediately begins cleaning up the
3289 lov, just after starting the mds_lov process
3291 Severity : enhancement
3293 Description: Implement 'lfs df' to report actual free space on per-OST basis
3294 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3295 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3296 usage of objects; '-h' to report in human readable format.
3298 ------------------------------------------------------------------------------
3300 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3307 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3308 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3309 at startup during orphan unlinking. Instead, skip buddy update
3310 before setup, as it will be regenerated after recovery is complete.
3315 Description: If an OST is inactive, its locks might reference stale inodes.
3316 Details : lov_change_cbdata() must iterate over all namespaces, even if
3317 they are inactive to clear inode references from the lock.
3319 Severity : enhancement
3320 Frequency : occasional, if non-standard max_dirty_mb used
3322 Description: Client will block write RPCs if not enough grant
3323 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3324 then the client will block writes while waiting for another RPC
3325 to complete instead of consuming its dirty limit. With change
3326 we get improved performance when max_dirty_mb is small.
3328 Severity : enhancement
3329 Bugzilla : 3389, 6253
3330 Description: Add support for supplementary groups on the MDS.
3331 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3332 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3333 lookups for user supplementary groups into a cache.
3337 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3338 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3339 for liblustre O_EXCL is enforced by MDS.
3343 Description: readdir never returns NULL in liblustre.
3344 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3345 dentry in struct dirent.
3349 Frequency : liblustre only, depends on application IO pattern
3350 Description: liblustre clients evicted if not contacting servers
3351 Details : Don't put liblustre clients into the ping_evictor list, so
3352 they will not be evicted by the pinger ever.
3354 Severity : enhancement
3356 Description: Add ability to evict clients by NID from MDS.
3357 Details : By echoing "nid:$NID" string into
3358 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3359 $NID would be instantly evicted from this MDS and from all active
3360 OSTs connected to it.
3364 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3365 Details : lookup_it_finish() used to query file size from OSTs that was not
3370 Description: service threads change working directory to that of init
3371 Details : Starting lustre service threads may pin the working directory
3372 of the parent thread, making that filesystem busy. Threads
3373 now change to the working directory of init to avoid this.
3377 Frequency : during shutdown only
3378 Description: shutdown with a failed MDS or OST can cause unmount to hang
3379 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3384 Frequency : 2.6 only
3385 Description: chmod/chown may include an extra supplementary group
3386 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3387 supplementary group and if none is specified this is used.
3390 Bugzilla : 5479 (6816)
3392 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3393 Details : If lookup is in progress on a file that is unlinked we might try
3394 to revalidate the inode and fail in revalidate after lookup is
3395 complete and ll_file_open() enqueues the open again but
3396 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3399 Frequency : always, if lconf --abort_recovery used
3401 Description: lconf --abort_recovery fails with 'Operation not supported'
3402 Details : lconf was attempting to abort recovery on the MDT device and not
3405 ------------------------------------------------------------------------------
3407 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3412 Frequency : rare (only unsupported configurations with a node running as an
3414 Bugzilla : 6514, 5137
3415 Description: Mounting a Lustre file system on a node running as an OST could
3417 Details : OSTs now preallocates memory needed to write out data at
3418 startup, instead of when needed, to avoid having to
3419 allocate memory in possibly low memory situations.
3420 Specifically, if the file system is mounted on on OST,
3421 memory pressure could force it to try to write out data,
3422 which it needed to allocate memory to do. Due to the low
3423 memory, it would be unable to do so and the node would
3424 become unresponsive.
3426 Severity : enhancement
3428 Description: Addition of lconf --service command line option
3429 Details : lconf now accepts a '--service <arg>' option, which is
3430 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3432 Severity : enhancement
3434 Description: Failover mode is now the default for OSTs.
3435 Details : By default, OSTs will now run in failover mode. To return to
3436 the old behaviour, add '--failout' to the lmc line for OSTs.
3438 Severity : enhancement
3440 Description: Health checks are now provided for MDS and OSTs
3441 Details : Additional detailed health check information on MSD and OSTs
3442 is now provided through the procfs health_check value.
3445 Frequency : occasional, depends on IO load
3447 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3448 after numerous create/delete cycles
3449 Details : The ext3 inode allocation policy would not allocate new inodes
3450 very well on the OSTs because there are no new directories
3451 being created. Instead we look for groups with free space if
3452 the parent directories are nearly full.
3457 Description: Network or server problems during mount may cause partially
3458 mounted clients instead of returning an error.
3459 Details : The config llog parsing code may overwrite the error return
3460 code during mount error handling, returning success instead
3466 Description: MDS can fail to allocate large reply buffers
3467 Details : After long uptimes the MDS can fail to allocate large reply
3468 buffers (e.g. zconf client mount config records) due to memory
3469 fragmentation or consumption by the buffer cache. Preallocate
3470 some large reply buffers so that these replies can be sent even
3471 under memory pressure.
3475 Frequency : rare (liblustre)
3476 Description: fsx running with liblustre complained that using truncate() to
3477 extend the file doesn't work. This patch corrects that issue.
3478 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3479 ATTR_SIZE and lsm use in llu_setattr_raw.
3483 Frequency : rare, only 2.6 kernels
3484 Description: Unusual file access patterns on the MDS may result in inode
3485 data being lost in very rare circumstances.
3486 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3487 code in the 2.6 kernel caused the inode and/or EA data not to
3488 be read from disk, causing single-file corruption.
3492 Frequency : rare, only 2.6 filesystems using extents
3493 Description: Heavy concurrent write and delete load may cause data corruption.
3494 Details : It was possible under high-load situations to have an extent
3495 metadata block in the block device cache from a just-unlinked
3496 file overwrite a newly-allocated data block. We now unmap any
3497 metadata buffers that alias just-allocated data blocks.
3501 Frequency : filesystems with default stripe_count larger than 77
3502 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3503 Details : lconf specifies an inode size of 4096 bytes when the default
3504 stripe_count is larger than 77. This conflicts with the default
3505 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3506 this case to avoid pinning too much memory for large EAs.
3508 ------------------------------------------------------------------------------
3510 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3515 Frequency : rare (extremely heavy IO load with hundreds of clients)
3517 Description: Client is evicted, gets IO error writing to file
3518 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3519 caused clients to be evicted for AST timeouts. The fixes in
3520 bug 5192 mean we no longer need to have such short AST timeouts
3521 so ldlm_timeout has been increased.
3524 Frequency : occasional during --force or --failover shutdown under load
3525 Bugzilla : 5949, 4834
3526 Description: Server oops/LBUG if stopped with --force or --failover under load
3527 Details : a collection of import/export refcount and cleanup ordering
3528 issues fixed for safer force cleanup
3531 Frequency : only filesystems larger than 120 OSTs
3532 Bugzilla : 5990, 6223
3533 Description: lfs getstripe would oops on a very large filesystem
3534 Details : lov_getconfig used kfree on vmalloc'd memory
3537 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3538 Bugzilla : 6242, 6243
3539 Description: reading from files that had been truncated to a non-zero size
3540 but never opened returned no data
3541 Details : ll_file_read() reads zeros from no-object files to EOF
3546 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3547 Details : There's little harm in aborting MDS/OSS recovery and letting it
3548 try again, so I removed the LASSERT and return an error instead.
3550 Severity : enhancement
3552 Description: New debugging infrastructure for tracking down data corruption
3553 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3554 (b) cover more of the client-side code path, and (c) try to narrow
3555 down where problems occurred
3559 Bugzilla : 3819, 4364, 4397, 6313
3560 Description: Racing close and eviction MDS could cause assertion in mds_close
3561 Details : It was possible to get multiple mfd references during close and
3562 client eviction, leading to one thread referencing a freed mfd.
3564 Severity: : enhancement
3565 Bugzilla : 3262, 6359
3566 Description: Attempts to reconnect to servers are now more aggressive.
3567 Details : This builds on the enhanced upcall-less recovery that was added
3568 in 1.4.2. When trying to reconnect to servers, clients will
3569 now try each server in the failover group every 10 seconds. By
3570 default, clients would previously try one server every 25 seconds.
3575 Description: After recovery, certain operations trigger a failed
3576 assertion on a client.
3577 Details : Failing over an mds, using lconf -d --failover, while a
3578 client was doing a readdir() call would cause the client to
3579 LBUG after recovery completed and the readdir() was resent.
3581 Severity : enhancement
3583 Description: Default groups are now added by lconf
3584 Details : You can now run lconf --group <servicename> without having to
3585 manually add groups with lmc.
3588 Frequency : occasional
3590 Description: Nodes with an elan id of 0 trigger a failed assertion
3593 Frequency : always when accessing e.g. tty/console device nodes
3595 Description: tty and some other devices nodes cannot be used on lustre
3596 Details : file's private_data field is used by device data and lustre
3597 values in there got lost. New field was added to struct file to
3598 store fs-specific private data.
3601 Frequency : when exporting Lustre via NFS
3603 Description: NFSD failed occasionally when looking up a path component
3604 Details : NFSD is looking up ".." which was broken in ext3 directories
3605 that had grown large enough to become hashed.
3608 Frequency : Clusters with multiple interfaces not on the same subnet
3610 Description: Nodes will repeatedly try to reconnect to an interface which it
3611 cannot reach and report an error to the log.
3612 Details : Extra peer list entries will be created by lconf with some peers
3613 unreachable. lconf now validates the peer before adding it.
3616 Frequency : Only if a default stripe is set on the filesystem root.
3618 Description: Setting a default stripe on the filesystem root prevented the
3619 filesystem from being remounted.
3620 Details : The client was sending extra request flags in the root getattr
3621 request and did not allocate a reply buffer for the dir EA.
3624 Frequency : occasional, higher if lots of files are accessed by one client
3625 Bugzilla : 6159, 6097
3626 Description: Client trips assertion regarding lsm mismatch/magic
3627 Details : While revalidating inodes the VFS looks up inodes with ifind()
3628 and in rare cases can find an inode that is being freed.
3629 The ll_test_inode() code will free the lsm during ifind()
3630 when it finds an existing inode and then the VFS later attaches
3631 this free lsm to a new inode.
3635 Bugzilla : 6422, 7030
3636 Description: MDS deadlock between mkdir and client eviction
3637 Details : Creating a new file via mkdir or mknod (starting a transaction
3638 and getting the ns lock) can deadlock with client eviction
3639 (gets ns lock and trying to finish a synchronous transaction).
3642 Frequency : occasional
3643 Description: While starting a server, the fsfilt_ext3 module could not be
3645 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3646 kernels. Previously, lconf would still use the ext3 name
3647 when trying to load modules. Now, it will correctly use
3648 ext3 on 2.4 and ldiskfs on 2.6.
3650 Severity : enhancement
3651 Description: The default stripe count has been changed to 1
3652 Details : The interpretation of the default stripe count (0, to lfs
3653 or lmc) has been changed to mean striping across a single
3654 OST, rather than all available. For general usage we have
3655 found a stripe count of 1 or 2 works best.
3657 Severity : enhancement
3658 Description: Add support for compiling against Cray portals.
3659 Details : Conditional compiling for some areas that are different
3663 Frequency : occasional
3664 Bugzilla : 6409, 6834
3665 Description: Creating files with an explicit stripe count may lead to
3666 a failed assertion on the MDS
3667 Details : If some OSTs are full or unavailable, creating files may
3668 trigger a failed assertion on the MDS. Now, Lustre will
3669 try to use other servers or return an error to the
3673 Frequency : occasional
3675 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3676 caused lock timeout during readahead (since 1.4.2).
3677 Details : Processes doing readahead might match a lock that hasn't been
3678 granted yet if there are overlapping and conflicting lock
3679 requests. The readahead process waits on ungranted lock
3680 (original lock is CBPENDING), while OST waits for that process
3681 to cancel CBPENDING read lock and eventually evicts client.
3683 Severity : enhancement
3685 Description: Initial enabling of flock support for clients
3686 Details : Implements fcntl advisory locking and file status functions.
3687 This feature is provided as an optional mount flag (default
3688 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3689 locking are implemented yet, and those that are are not guaranteed
3690 to be completely correct in production environments.
3691 mount -t lustre -o [flock|noflock] ...
3694 Frequency : occasional
3696 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3697 assertion in the ext3 JBD (journaling) layer.
3698 Details : The b_committed_data struct is protected by the big kernel lock
3699 in 2.4 kernels, serializing journal_commit_transaction() and
3700 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3701 access to this struct. In 2.6 kernels there is finer grained
3702 locking to improve SMP performance of the JBD layer.
3706 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3711 Description: merge and cleanup kernel patches.
3714 -----------------------------------------------------------------------------
3716 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3718 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3719 versions of Lustre. Please update all servers and clients to
3720 version 1.4.2 or later at the same time. You must also run
3721 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3722 to update the configuration logs.
3724 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3725 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3726 - swab llog records of type '0' so we get proper header size/idx (5861)
3727 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3728 - fix rename of one directory over another leaking an inode (5953)
3729 - avoid SetPageDirty on 2.6 (5981)
3730 - don't re-add just-being-destroyed locks to the waiting list (5653)
3731 - when creating new directories, inherit the parent's custom
3732 striping settings if present parent (3048)
3733 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3734 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3735 - don't hold client locks on temporary worklist from l_lru (5666)
3736 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3737 - automatically evict dead clients (5921)
3738 - Update file size properly in create+truncate+fstat case (6196)
3739 - Do not unhash mountpoint dentries, do not allow removal of
3741 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3742 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3743 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3744 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3745 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3746 - Correctly handle reads of files with no objects (6243)
3747 - lctl recover will also mark a device active if deactivate used (5933)
3749 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3750 - allow --write-conf on an MDS with different nettype than client (5619)
3751 - don't write config llogs to MDS for mounts not from that MDS (5617)
3752 - lconf should create multiple TCP connections from a client (5201)
3753 - init scripts are now turned off by default; run chkconfig --on
3754 lustre and chkconfig --on lustrefs to use them
3755 - upcalls are no longer needed for clients to recover to failover
3757 - add --abort-recovery option to lconf to abort recovery on device
3759 - add support for an arbitrary number of OSTs (3026)
3760 - Quota support protocol changes.
3761 - forward compatibility changes to wire structs (6007)
3762 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3763 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3764 - improved "lustre status" script
3765 - initialize blocksize for non-regular files (6062)
3766 - added --disable-server and --disable-client configure options (5782)
3767 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3768 - Vanilla 2.4.29 support
3769 - increase maximum number of obd devices to 520 (6242)
3770 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3771 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3772 SLES 9 2.6.5 kernels
3773 - limit stripes per file to 160 (the maximum EA size) (6093)
3775 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3778 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3779 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3780 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3781 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3782 - small ext3 extents cleanups and fixes (5733)
3783 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3784 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3785 - avoid needless client->OST connect, fix handle mismatch (5317)
3786 - fix DLM error path that led to out-of-sync client, long delays (5779)
3787 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3788 - fix several locking issues related to i_size (5492,5624,5654,5672)
3789 - don't move pending lock onto export if it is already evicted (5683)
3790 - fix kernel oops when creating .foo in unlinked directory (5548)
3791 - fix deadlock in obdfilter statistics vs. object create (5811)
3792 - use time_{before,after} to avoid timer jiffies wrap (5882)
3793 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3794 - Do not leak request if server was not able to process it (5154)
3795 - If mds_open unable to find parent dir, make that negative lookup(5154)
3796 - don't create new directories with extent-mapping (5909, 5936)
3798 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3799 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3800 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3801 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3802 - support for RHEL4 kernel on i686 (5773)
3803 - provide error messages when incompatible logs are encountered (5898)
3805 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3806 * version 1.4.0.10 (1.4.1 release candidate 1)
3808 - don't keep a lock reference when lock is not granted (4238)
3809 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3810 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3811 - drop import inflight refcount on signal_completed_replay error (5255)
3812 - unlock page after async write error during send (3677)
3813 - handle missing objects in filter_preprw_read properly (5265)
3814 - no transno return for symlink open, don't save no-trasno open (3440)
3815 - don't try to complete elan receive that already failed (4012)
3816 - free RPC server reply state on error (5406)
3817 - clean up thread from ptlrpc_start_thread() on error (5160)
3818 - readahead could read extra page into cache that wasn't ejected (5388)
3819 - prevent races in class_attach/setup/cleanup/detach (5260)
3820 - don't dereference de->d_inode after l_dput of de (5458)
3821 - use "int" for stripe value returned from lock_to_stripe (5544)
3822 - mballoc allocation and error-checking fixes in 2.6 (5504)
3823 - block device patches to fix I/O request sizes in 2.6 (5482)
3824 - look up hostnames for IB nals (5602)
3825 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3826 - don't start multiple acceptors for the same port (5277)
3827 - fix incorrect LASSERT in mds_getattr_name (5635)
3828 - export a proc file for general "ping" checking (5628)
3829 - fix "lfs check" to not block when the MDS is down (5628)
3831 - service request history (4965)
3832 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3833 - create an "evict_client" /proc entry on OSTs, like the MDS has
3834 - fix mount usage message, return errors per mount(8) (5168)
3835 - change grep [] to grep "[]" in tests so they work in more UMLs
3836 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3837 - remove ancient LOV_MAGIC_V0 EA support (5047)
3838 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3839 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3840 - allow readahead allocations to fail when low on memory (5383)
3841 - mmap locking landed again, after considerable improvement (2828)
3842 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3844 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3847 - send OST transaction number in read/write reply to free req (4966)
3848 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3849 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3850 - the watchdog thread now runs as interruptible (5246)
3851 - flock/lockf fixes (but it's still disabled, pending 5135)
3852 - don't use EXT3 constants in llite code (5094)
3853 - memory shortage at startup could cause assertion (5176)
3855 - reorganization of lov code
3856 - single portals codebase
3858 - add extents/mballoc support (5025)
3859 - direct I/O reads in the obdfilter (4048)
3860 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3862 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3865 - send OST transaction number in read/write reply to free req (4966)
3866 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3867 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3868 - the watchdog thread now runs as interruptible (5246)
3869 - handle missing objects in filter_preprw_read properly (5265)
3870 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3871 - drop import inflight refcount on signal_completed_replay error (5255)
3872 - unlock page after async write error during send (3677)
3873 - return original error code on reconstructed replies (3761)
3874 - no transno return for symlink open, don't save no-trasno open (3440)
3876 - add pid to ldlm debugging output (4922)
3877 - bump the watchdog timeouts -- we can't handle 30sec yet
3878 - extra debugging for orphan dentry/inode bug (5259)
3880 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3883 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3884 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3885 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3886 - fix NULL dereference in /proc/sys/portals/routes (4827)
3887 - allow failed mdc_close() operations to be interrupted (4561)
3888 - stop precreate on OST before MDS would time out on it (4778)
3889 - don't send partial-page writes before EOF from client (4410)
3890 - discard client grant for sub-page writes on large-page clients (4520)
3891 - don't free dentries not owned by NFS code, check generation (4806)
3892 - fix lsm leak if mds_create_objects() fails (4801)
3893 - limit debug_daemon file size, always print CERROR messages (4789)
3894 - use transno after validating reply (3892)
3895 - process timed out requests if import state changes (3754)
3896 - update mtime on OST during writes, return in glimpse (4829)
3897 - add mkfsoptions to LDAP (4679)
3898 - use ->max_readahead method instead of zapping global ra (5039)
3899 - don't interrupt __l_wait_event() during strace
3901 - add software watchdogs to catch hung threads quickly (4941)
3902 - make lustrefs init script start after nfs is mounted
3903 - fix CWARN/ERROR duplication (4930)
3904 - return async write errors to application if possible (2248)
3905 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3906 - print NAL number in %x format (4645)
3907 - update barely-supported suse-2.4.21-171 series (4842)
3908 - support for sles 9 %post scripts
3909 - support for building 2.6 kernel-source packages
3910 - support for sles km_* packages
3912 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3915 - ignore -ENOENT errors in osc_destroy (3639)
3916 - notify osc create thread that OSC is being cleaned up (4600)
3917 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3918 - reconstruct ost_handle() like mds_handle() (4657)
3919 - create a new thread to do import eviction to avoid deadlock (3969)
3920 - let lconf resolve symlinked-to devices (4629)
3921 - don't unlink "objects" from directory with default EA (4554)
3922 - hold socknal file ref over connect in case target is down (4394)
3923 - allow more than 32000 subdirectories in a single directory (3244)
3924 - fix blocks count for O_DIRECT writes (3751)
3925 - OST returns ENOSPC from object create when no space left (4539)
3926 - don't send truncate RPC if file size isn't changing (4410)
3927 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3928 - bind to privileged port in socknal and tcpnal (3689)
3930 - rate limit CERROR/CWARN console message to avoid overload (4519)
3931 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3932 - basic mmap support (3918)
3933 - kernel patch series update from b1_4 (4711)
3935 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3938 - avoid crash during MDS cleanup with OST shut down (2775)
3939 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3940 - don't use bad inodes on the MDS (3744)
3941 - dynamic object preallocation to improve recovery speed (4236)
3942 - don't hold spinlock over lock dumping or change debug flags (4401)
3943 - don't zero obd_dev when it is force cleaned (3651)
3944 - print grants to console if they go negative (4431)
3945 - "lctl deactivate" will stop automatic recovery attempts (3406)
3946 - look for existing locks in ldlm_handle_enqueue() (3764)
3947 - don't resolve lock handle twice in recovery avoiding race (4401)
3948 - revalidate should check working dir is a directory (4134)
3950 - don't always mark "slow" obdfilter messages as errors (4418)
3952 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3955 - don't close LustreDB during write_conf until it is done (3860)
3956 - fix typo in lconf for_each_profile (3821)
3957 - allow dumping logs from multiple threads at one time (3820)
3958 - don't allow multiple threads in OSC recovery (3812)
3959 - fix debug_size parameters (3864)
3960 - fix mds_postrecov to initialize import for llog ctxt (3121)
3961 - replace config semaphore with spinlock (3306)
3962 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3963 - don't allow enqueue to complete on a destroyed export (3822)
3964 - down write_lock before checking llog header bitmap (3825)
3965 - recover from lock replay timeout (3764)
3966 - up llog sem before sending rpc (3652)
3967 - reduce ns lock hold times when setting kms (3267)
3968 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3969 - fix NULL deref and obd_dev leak on setup error (3312)
3970 - replace some LBUG about llog ops with error handling (3841)
3971 - don't match INVALID dentries from d_lookup and spin (3784)
3972 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3973 - fix invalid assertion in ptlrpc_set_wait (3880)
3975 - add libwrap support for the TCP acceptor (3996)
3976 - add /proc/sys/portals/routes for non-root route listing (3994)
3977 - allow setting MDS UUID in .xml (2580)
3978 - print the stack of a process that LBUGs (4228)
3980 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3983 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3984 - make sure to unset replay flag from failed open requests (3440)
3985 - if default stripe count is 0, use OST count for inode size (3636)
3986 - update parent mtime/ctime on client for create/unlink (2611)
3987 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3988 - free recovery state on server during a forced cleanup (3571)
3989 - unregister_reply for resent reqs (3063)
3990 - loop back devices mounting and status check on 2.6 (3563)
3991 - fix resource-creation race that can provoke i_size == 0 (3513)
3992 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3993 - more debugging for page-accounting assertion (3746)
3994 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3995 - avoid deadlock after precreation failure (3758)
3996 - fix race and lock order deadlock in orphan handling (3450, 3750)
3997 - add validity checks when grabbing inodes from l_ast_data (3599)
3999 - add /proc/.../recovery_status to obdfilter (3428)
4000 - lightweight CDEBUG infrastructure, debug daemon (3668)
4001 - change default OSC RPC parameters to be better on small clusters
4002 - turn off OST read cache for files smaller than 32MB
4003 - install man pages and include them in rpms (3100)
4004 - add new init script for (un)mounting lustre filesystems (2593)
4005 - run chkconfig in %post for init scripts (3701)
4006 - drop scimac NAL (unmaintained)
4008 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4011 - clean kiobufs before and after use (3485)
4012 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4013 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4014 - change init script to fail more clearly if not run as root (1528)
4015 - allow clients to reconnect during replay (1742)
4016 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4017 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4018 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4019 - fix cond_resched (3554)
4020 - only evict unfinished clients after recovery (3515)
4021 - allow bulk resend, prevent data loss (3570)
4022 - dynamic ptlrpc request buffer allocation (2102)
4023 - don't allow unlinking open directory if it isn't empty (2904)
4024 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4025 - remove extraneous obd dereference causing LASSERT failure (3334)
4026 - don't use get_cycles() when creating temp. files on the mds (3156)
4027 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4028 - handle EEXIST for set-stripe, set proper directory name (3336)
4030 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4031 - fix ksocknal_fmb_callback() error messages (2918)
4033 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4036 - don't copy lvb into (possibly NULL) reply on error (2983)
4037 - don't deref dentry after dput, don't free lvb on error (2922)
4038 - use the kms to determine writeback rpc length (2947)
4039 - increment oti_logcookies when osc is inactive (2948)
4040 - update client's i_blocks count via lvb messages (2543)
4041 - handle intent open/close of special files properly (1557)
4042 - mount MDS with errors=remount-ro, like obdfilter (2009)
4043 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4044 - don't use cancelling-locks' kms values (2947)
4045 - use highest lock extent for kms, not last one (2925)
4046 - don't dereference ERR_PTR() dentry in error handling path (3107)
4047 - fix thread race in portals_debug_dumplog() (3122)
4048 - create lprocfs device entries at setup instead of at attach (1519)
4049 - common AST error handler, don't evict client on completion race (3145)
4050 - zero nameidata in detach_mnt in 2.6 (3118)
4051 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4052 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4053 - initialize RPC timeout timer earlier for 2.6 (3219)
4054 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4055 - print nal/nid for unknown nid (3258)
4056 - additional checks for oscc recovery before doing precreate (3284)
4057 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4058 - don't crash in mdc_close for bad permissions on open (3285)
4059 - zero i_rdev for non-device files (3147)
4060 - clear page->private before handing to FS, better assertion (3119)
4061 - tune the read pipeline (3236)
4062 - fix incorrect decref of invalidated dentry (2350)
4063 - provide read-ahead stats and refine rpc in flight stats (3328)
4064 - don't hold journal transaction open across create RPC (3313)
4065 - update atime on MDS at close time (3265)
4066 - close LDAP connection when recovering to avoid server load (3315)
4067 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4068 - don't leak open file on MDS after open resend (3325)
4069 - serialize filter_precreate and filter_destroy_precreated (3329)
4070 - loop device shouldn't call sync_dev() for nul device (3092)
4071 - clear page cache after eviction (2766)
4072 - resynchronize MDS->OST in background (2824)
4073 - refuse to mount the same filesystem twice on same mountpoint (3394)
4074 - allow llmount to create routes for mounting behind routers (3320)
4075 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4076 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4077 - fix rare problem with rename on htree directories (3417)
4079 - allow default OST striping configuration per directory (1414)
4080 - fix compilation for qswnal for 2.6 kernels (3125)
4081 - increase maximum number of MDS request buffers for large systems
4082 - change liblustreapi to be useful for external progs like lfsck (3098)
4083 - increase local configuration timeout for slow disks (3353)
4084 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4086 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4089 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4090 - don't overwrite extent policy data in reply if lock was blocked (2901)
4091 - drop filter export grants atomically with removal from device (2663)
4092 - del obd_self_export from work_list in class_disconnect_exports (2908)
4093 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4094 - swab reply message in mdc_close, other PPC fixes (2464)
4095 - fix destroying of named logs (2325)
4096 - overwrite old logs when running lconf --write_conf (2264)
4097 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4098 - fix race in target_handle_connect (2898)
4099 - mds_reint_create() should take same inode create lock (2926)
4100 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4101 - don't close files for self_export to avoid uninitialized obd (2936)
4102 - allow MDS with the same name as client node (2939)
4103 - hold dentry reference for closed log files for unlink (2325)
4104 - reserve space for all logs during transactions (2059)
4105 - don't evict page beyond end of stripe extent (2925)
4106 - don't oops on a deleted current working directory (2399)
4107 - handle hard links to targets without a parent properly (2517)
4108 - don't dereference NULL lock when racing during eviction (2867)
4109 - don't grow lock extents when lots of conflicting locks (2919)
4111 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4114 - account for cache space usage on clients to avoid data loss (974)
4115 - lfsck support in lustre kernel code (2349)
4116 - reduce journal credits needed for BRW writes (2370)
4117 - orphan handling to avoid losing space on client/server crashes
4118 - ptlrpcd can be blocked, stopping ALL progress (2477)
4119 - use lock value blocks to assist in proper KMS, faster stat (1021)
4120 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4121 - recovery for initial connections (2355)
4122 - fixes for mds_cleanup_orphans (1934)
4123 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4124 - block all file creations until orphan recovery completes (1901)
4125 - client remove rq_connection from request struct (2423)
4126 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4127 - recovery timer race (2670)
4128 - mdc_close recovey bug (2532)
4129 - ptlrpc cleanup bug (2710)
4130 - mds timeout on local locks (2588)
4131 - namespace lock held during RPCs (2431)
4132 - handle interrupted sync write properly (2503)
4133 - don't try to handle a message that hasn't been replied to (2699)
4134 - client assert failure during cleanup after abort recovery (2701)
4135 - leak mdc device after failed mount (2712)
4136 - ptlrpc_check_set allows timedout requests to complete (2714)
4137 - wait for inflight reqs when ptlrpcd finishes (2710)
4138 - make sure unregistered services are removed from the srv_list
4139 - reset bulk XID's when resending them (caught by 1138 test)
4140 - unregister_bulk after timeout
4141 - fix lconf error (2694)
4142 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4143 - readahead locks pages, leaves pending causing memory pressure (2673)
4144 - increase OST request buffers to 4096 on large machines (2729)
4145 - fix up permission of existing directories in simple_mkdir (2661)
4146 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4147 - don't assign transno to errored transactions (2742)
4148 - don't delete objects on OST if given a bogus objid from MDS (2751)
4149 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4150 - if rq_no_resend, then timeout request after recovery (2432)
4151 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4152 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4153 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4154 - MDS recovery shouldn't delete objects that it has given out (2730)
4155 - if enqueue arrives after completion, don't clobber LVB (2819)
4156 - don't unlock pages twice when trigger_group_io returns error (2814)
4157 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4158 - don't write pages to disk if there was an error (1450)
4159 - don't ping imports that have recovery disabled (2676)
4160 - take buffered bytes into account when balancing socknal conn (2817)
4161 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4162 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4163 - remove little-endian swabbing of llog records (1987)
4164 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4165 - retry reposting request buffers if they fail (1191)
4166 - grow extent at grant time to avoid granting a revoked lock (2809)
4167 - lock revoke doesn't evict page if covered by a second lock (2765)
4168 - disable VM readahead to avoid reading outside lock extents (2805)
4170 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4171 - updated kernel patches for hp-2.4.20 kernel (2681)
4173 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4176 - fix truncated write corruption (2366)
4177 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4179 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4180 - ensure len(uuid) < 37 in lmc (1171)
4181 - fix ia64 OOPS in llog_test (2255)
4182 - zero end of page at obdfilter for partial page writes (2648)
4183 - don't leave stale dentries around after renames (bug 2428)
4184 - fix timeouts when evicting a client with a single lock held (2642)
4185 - set deadline for the initial HELLO message to drain (2634)
4186 - print out dotted-quad IP addresses in the socknal (2302)
4188 - additional debugging for MDS client eviction problem (2443)
4189 - fix mkfsoptions support for osts (2603, 2604)
4191 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4194 - add series for the vanilla 2.6.0 kernel
4195 - add series for the vanilla 2.4.24 kernel
4196 - add series for a cray x86/64 UL kernel drop
4197 - fix xattr patches for the vanilla 2.4.19 series
4199 - generate true UUIDs in lmc (1171)
4200 - have portals stack dumping break in UML (2466)
4201 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4202 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4203 - wake callback waiting threads on client eviction (2460)
4204 - Add --ptldebug and --subsystem to lmc (1719)
4205 - update assertion to allow safe interrupt allocation
4206 - set rq_no_resend for cancel requests (2432)
4207 - recalculate ptlrpcd timeout after resend (2494)
4208 - call vfs_rmdir when removing pending directories (2368)
4209 - fix renaming a file to itself (2429)
4210 - lmc creates a default one-stripe lov (2454)
4211 - expand procfs space to handle large clusters (2326)
4212 - increase UML stack to avoid overflow
4213 - update lconf's list of debug and subsystem masks
4214 - fix lfs find --obd (2510)
4215 - /proc tunable for disabling filter read caching (2591)
4216 - stop rpm packages from altering slapd.conf (2301)
4217 - disable nagle in the socknal under 0conf (2578)
4218 - choose mds inode size based on stripe count (2572)
4219 - fix kernel-source rpm problems (2516)
4221 - add --disable-doc to avoid pdf generation (2421)
4222 - update documentation, tests, type-os, comments
4223 - avoid format warnings on ia64
4224 - remove the TOE NAL
4225 - tiny code cleanups by removing unused fields
4227 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4230 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4231 - remove the most blatant lies from BUILDING file (2371)
4232 - change default debug level to reasonable production setting
4233 - reduce client side cache size to reduce cache flush time
4234 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4235 - make TCP zerocopy and pinger support enabled by default (2476)
4236 - sync writes completed after process exits caused crashes (2319)
4237 - maintain correct mount count on the MDS (2356)
4238 - backout 1557, because 2316 wasn't really fixed
4239 - better file I/O statistics gathering in /proc
4240 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4241 - another kernel patch to fix zero-copy TCP function export
4242 - don't take duplicate lock when processing re-sent getattr (2420)
4243 - lctl uses obd_self_export instead of creating new conn (2353)
4244 - MDS/OST recovery case which requires object creation asserted (2425)
4245 - move lfs from /usr/sbin to /usr/bin in packages
4246 - fix race between mds_client_add and mds_client_free (2417)
4247 - use kmalloc instead of slabs in portals (2430)
4248 - don't create duplicate records when a failover MDS is present (2442)
4249 - remove unnecessary mount age check (2332)
4250 - don't remove directory inodes from locks prematurely (2451)
4251 - don't break if MDS service name is the same as hostname (2103)
4252 - fix races in client write RPC generation when cache full (2482)
4254 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4257 - remove now-unused request->rq_obd (278)
4258 - if an allocation fails, print out how much memory we've used (1933)
4259 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4260 - add the "configurable stack size" patch to most series files (1256)
4261 - ability to write large log records, for 100+ OST configs (2306)
4262 - fix NULL deref when filter_prep fails (2314)
4263 - fix operator precedence error in filter_sync
4264 - dynamic allocation of socknal TX descriptors (2315)
4265 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4266 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4267 - ensure that CWARN messages in Portals always get to the syslog
4268 - __init/__exit are not for prototype decls (ldlm_init/exit)
4269 - x86-64 compile warning fixes
4270 - fix gateway LMC keyword conflict (2318)
4271 - fix MDS lock inversions in getattr/reint paths (1844)
4272 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4273 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4274 - DLM race condition prevented some lock evictions (2328)
4275 - ENOMEM detection and retry on socknal sends (2230)
4276 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4277 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4278 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4279 - better error messages when a client is rejected during recovery (1505)
4280 - avoid cancelling locks which were never granted, after failure (2330)
4281 - fix i_sem/journal inversion in mds_client_add (2333)
4282 - fix truncate/getattr lock cycle deadlock (2334)
4283 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4284 - fix two rare exit paths which could leak an l_lock() ref (2321)
4285 - fencepost error in MDS/OST orphan recovery (2226)
4286 - make log record alignment 8 bytes (1988)
4287 - lstripe now fails when requested offset > ost_count (2237)
4288 - ensure that all kernel series have a complete list.h (1607)
4289 - fix crashes in special-file operations (2316)
4290 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4292 - allow configurable automake binary, for testing new versions
4293 - small update to the lfs documentation
4295 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4297 * fix negative export reference count in fsfilt_sync (2312)
4299 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4300 * release candidate 0.9.1
4302 - orphans are moved into the PENDING directory for possible recovery
4303 - replayed opens now open by fid for orphan/rename safety (1042)
4304 - last close of an orphan inode generates a transno (683)
4305 - chdir() and mount() now pin the directory entry (1020)
4306 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4307 - discard very old requests without processing them (1502)
4308 - remove some common, well-understood CERRORs (1505)
4309 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4310 - clear "grant" flags in OST replies until OST grant code lands (1644)
4311 - fix read performance by not clobbering i_blksize on client (1598)
4312 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4313 - make LRU size a /proc tunable, clears locks when reduced (707)
4314 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4315 - server locks take references on exports now (1558)
4316 - build fixes for 2.4.20-rh trees (1663)
4317 - return an error from lov_create if all OSCs are inactive (1751)
4318 - fix import levels when a reconnect happens without a timeout (1597)
4319 - exit early from mds_open if we get a lookup error (1749)
4320 - partial page read at EOF wouldn't wait for disk before sending (1642)
4321 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4322 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4323 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4324 - write/truncate lock inversion (1639)
4325 - Don't auto-load obdclass, portals modules during cleanup (1495)
4326 - fix timestamps from jumping to "now" (1763)
4327 - extra journal assertions (1648)
4328 - add an extra multiunlink test (1771)
4329 - fix read_record/write_record API (1776)
4330 - fix leak of offset_extent, possible incorrect i_size later (1772)
4331 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4332 - Debugging for the kqswnal_get_idle_tx problems (1820)
4333 - Allow recovery to be attempted multiple times (1536)
4334 - Write out MDS last_rcvd file after it is first created (1600)
4335 - Fix tx_descriptor leak in failed transmit situations (1827)
4336 - ext3 journaling fixes for assertion failure after IO error (1871)
4337 - class_export_put() on freed export after completion AST error (1896)
4338 - Fix revalidate looping in VFS (1322)
4339 - Don't access a freed export during MDS_REINT timeout (1521)
4340 - Add open-unlink recovery support on the MDS (1673,1764)
4341 - Return an error if no MDS data was read from last_rcvd (1946)
4342 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4343 - Don't setup a disk device that doesn't match exported UUID (317)
4344 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4345 - avoid committing NULL handle in force close
4346 - local.sh is now a one-stripe LOV configuration
4347 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4348 - fix ext3 htree duplicate directory entry corruption (1516)
4349 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4350 - update to new LOV EA format (2097)
4351 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4353 2003-06-15 Phil Schwan <phil@clusterfs.com>
4356 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4357 - per-import/export recovery handling (958, 931, 959)
4358 - multiple last-rcvd slots, for serving multiple FSes (949)
4359 - connections are again shared between multiple imp/exports (963, 964)
4360 - "umount -f" would hang if any requests needed to be sent (393, 978)
4361 - avoid pinning large req buffer by copying for queued messages (989)
4362 - add "uuid" to "lctl device" command to help upcalls (991)
4363 - "open" RPCs with transnos would confuse recovery counters (1037)
4364 - do proper endian conversion of all wire messages (288, 340, 891)
4365 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4366 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4367 - fix ptlrpc_connection leak in target_handle_connect (1174)
4368 - fix import refcounting bug in OST and MDS cleanup (1134)
4369 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4370 - fix very unlikely obd_types race condition (501)
4371 - remove osc_open hack for echo_client (1187)
4372 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4373 - a failure in read_inode2 leads to deadlock (1139)
4374 - cancel ack-locks as soon as transaction is committed (1072)
4375 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4376 - make sure to commitrw after any preprw to avoid deadlock (1162)
4377 - failing to execute a file in a lustre FS would lock inode (1203)
4378 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4379 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4380 - obd_uuid2tgt would walk off the end of the list (1255)
4381 - on IA64 the osc would give portals incorrect bulk size (1258)
4382 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4383 - fix lock inversion caused by new llite matching code (1282)
4384 - limit the number of dirty pages on a client to 10MB (1286)
4385 - timed out locks were not being corrected cancelled (1289)
4386 - fix O_DIRECT above 4GB on IA-32 (1292)
4387 * major user-visible changes
4388 - fail out/fail over policy now controlled by the upcall (993)
4390 - add OBD_PING to check server availability and failure (954)
4391 - lustre messages are now sent in sending host order (288, 340, 891)
4392 - add eadatalen to MDS getattr reply (340)
4393 - OST read replies may contain second buffer, with per-page status (593)
4395 2003-03-11 Phil Schwan <phil@clusterfs.com>
4398 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4399 - failed open()s could cause deadlock; fixed (867, 869)
4400 - stop cancelling OST locks when files are closed (481)
4401 - overlapping XID spaces caused network corruption (851, 853)
4402 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4403 - setattr_raw added, to do single-RPC, server-side setattrs
4404 - lmc/lconf syntax change for OST UUIDs
4405 - fix crashy race condition between ptlrpc_free_req and osc_close
4406 - don't use request in mdc_enqueue if we hit a timeout (889)
4407 - don't set the inode i_size for regular files from the MDS (896)
4408 - handle out of order completion AST (842)
4409 - don't LBUG if a lock request times out after receiving AST (913)
4410 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4411 - if a bad lock AST arrives, send an error instead of dropping entirely
4412 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4413 - fix leak in bulk IO when only partially completed (899, 900, 926)
4414 - fix O_DIRECT for ia64 (55)
4415 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4416 - C-z after timeout could hang a process forever; fixed (977)
4418 - client-side I/O cache (678, 924, 929, 941, 970)
4420 - READPAGE and SETATTRs which don't take server-side locks get
4423 2003-02-11 Phil Schwan <phil@clusterfs.com>
4426 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4427 - Fix fsfilt_extN_readpage() to read a full page of directory
4428 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4429 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4430 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4431 - Most intent operations were removed, in favour of a new RPC mode
4432 that does a single RPC to the server and bypasses most of the VFS
4433 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4434 - Aggressively cancel local locks on DLM servers
4435 - mds_reint_unlink sends EA to the client if it's the last nlink.
4436 client uses that EA to unlink OST objects.
4437 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4438 - recursive symlinks were fixed (439)
4439 - fixed NULL deref in DEBUG_REQ
4440 - filter_update_lastobjid no longer calls sync, which annoyed extN
4441 - fixed multi-client small-writes to a single file problem (445)
4442 - fixed mtime updates during file writes (607)
4443 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4444 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4445 - fixed Linux OST size reporting problem (444, 656)
4446 - OST now updates object mtime with writes or setattr (607, 619)
4447 - client verifies file size before zeroing page past EOF (445)
4448 - OST now writes last allocated objid to disk with allocation (108)
4449 - LOV on echo now works (409)
4451 - mds_reint_unlink sends a new buffer, with the EA included. this
4452 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4453 set if a regular file was being unlinked, and it was the last link
4454 - use PtlGet from the target for bulk writes (315)
4455 - OST now updates object mtime with writes or setattr (607, 619)
4456 - LDLM now has a grant-time callback to revalidate locked items, if
4458 - Many MDS operations were reorganized to combat race conditions
4460 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4461 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4463 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4466 - Fully reactivate OST imports after reconnection (512, others)
4467 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4468 - More graceful error handling for truncating on dead OST (515)
4469 - Don't error out unless we're actually accessing dead stripes (474)
4470 - Fix garbage sizes when stripes are missing (410)
4471 - LRU counters were broken, causing constant lock purge (433, 432)
4472 - garbage on read from stripes with failed OSTs (441)
4473 - mark OSCs as active before reconnecting during recovery (438)
4474 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4475 - lfind did not preserve OST order in output (443)
4476 - symlinks cause hung clients, incorrect data (439)
4477 - stop dereferencing request after dropping refcount (457)
4478 - don't LASSERT(spin_is_locked) on non-SMP (455)
4479 - fixes for many rename() bugs
4480 - fstat didn't correctly synchronize attributes (399)
4481 - server must handle lock cancellation during blocking AST prep (487)
4482 - bulk descriptors were free()d too soon (511)
4483 - fix paths in lconf, which would load incorrect modules (451, 507)
4484 - fix confusing lconf 'host not found' error message (386)
4485 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4486 - fix race condition in mdc_blocking_ast() for inode access (526)
4487 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4488 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4489 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4490 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4491 - fix lstripe in several ways (kernel side) (527)
4492 - fix request leak in ldlm_cli_enqueue (262)
4493 - incorrect OSC was marked inactive after OST failure
4494 - call mds_fs_cleanup before unmounting filesystem (524)
4495 - fix races between taking ns_lock and ldlm_lock_change_resource
4496 - fix races updating LOV export open file list
4497 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4498 - fix recovery NULL deref in ldlm_cli_cancel_unused
4499 - fix some DLM races by using new hash table for lock handles (419)
4500 - permit the client to specify desired inodes, at replay
4501 - duplicate requests when we queue them for replay reintegration
4502 - fix last_rcvd offset calculation
4503 - sync after each recovered transaction, so we always make progress
4504 - never, not always, ERESTART requests without transnos
4505 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4507 - skip replay if the MDS says that the client is already connected
4508 - don't check for a recovery-enabled export to match lctl's UUID
4509 - don't INC_USE_COUNT for phantom exports
4510 - don't crash when cleaning up phantom exports (567)
4511 - don't double-finish or set replay data for errored mdc_open requests
4512 - abort requests when they time out, so we don't get old replies
4513 - send/receive replies for AST messages again
4514 - if the client says that it doesn't have the lock, cancel it on the
4516 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4517 mark it as destroyed, it will all work out when decref is called
4518 - fix module use counts (22, 581)
4520 - ASTs now expect a reply (server cancels lock on error reply)
4522 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4525 - fix many simultaneous client startup (392)
4526 - fix dentry->d_it clobbering
4527 - credentials weren't being shipped for readdir/getattr operations
4528 - remove invalid assertions triggered during some concurrent MD
4530 - proper Lustre versions added (336, 389)
4531 - fix memory leak for create error case (398)
4532 - fix LOV locking bug that would get cli/srv out of sync
4533 - fix echo client over LOV (409)
4534 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4535 - fix double-O_EXCL intent crash (424)
4536 - avoid sending multiple lock CANCELs (352)
4538 - MDS can do multi-client recovery (modulo bugs in new code)
4540 - many updates, edits, cleanups
4542 2002-11-18 Phil Schwan <phil@clusterfs.com>
4545 - fix null d_it dereference (346)
4546 - fix full OST/dbench hang (333)
4547 - fix permission problem with file removal (286)
4548 - fix removal of OSCs from LOV when they fail
4549 - fix NULL deref during bulk timeout (214)
4550 - fix problems related to multiple filesystems on one MDS (241)
4551 - fixed serious subtle metadata locking bugs
4552 - free locks on clients when inodes are removed due to memory
4554 - fix inode pointer in lock data (285)
4555 - partial support for multiple MDS on a single host (241)
4556 - data locks weren't cancelled at clear_inode time (290, 311)
4557 - intent locks could lead to unbounded lock growth (205)
4558 - added a maximum lock count, an LRU list, and a flusher
4559 - fix multiple rename (365)
4560 - properly abstracted the echo client
4561 - OSC locked 1 byte too many; fixed
4562 - rewrote brw callback code:
4563 - fixed recovery bugs related to LOVs (306)
4564 - fixed too-many-pages-in-one-write crash (191)
4565 - fixed (again) crash in sync_io_timeout (214)
4566 - probably fixed callback-related race (385)
4568 - Add capability to MDS protocol
4569 - LDLM cancellations and callbacks on different portals
4571 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4574 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4575 - defer open object destruction to close time (601981, 138)
4576 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4577 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4578 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4579 - don't call obd_setattr in ll_file_release for destroyed objects
4580 * protocol change to lustre_msg: move |version| and add |flags|
4581 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4582 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4583 * added replay of create, unlink, link and rename operations during
4584 MDS failover; recovery should be much more robust now
4585 * remove failed OSCs from LOVs (only lov_create uses this so far)
4586 * the lustre-HOWTO was brought (more) up to date (582544)
4588 2002-10-23 Phil Schwan <phil@clusterfs.com>
4591 - in-use dentries weren't being reused properly (617851)
4592 - prevent multiple LDLM setup (599178)
4593 - fix LOV size calculations for truncate (617853)
4594 - fix client handling of MDS intent errors (POSIX)
4595 - fix permission bug in lovstripe.c test (624321)
4596 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4597 - truncate past end of file could corrupt data
4598 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4599 - a race in recovery could return ETIMEDOUT to apps (623947)
4600 - building outside the source directory was fixed
4601 * the lustre-HOWTO was brought (more) up to date (582544)
4602 * major progress was made on recovery functionality
4604 2002-10-10 Phil Schwan <phil@clusterfs.com>
4607 - recovery deadlock fix
4608 - rm -rf causes LBUG fix (617817)
4609 - file open by multiple tasks fix (618962)
4610 - directory permissions bugs (602707 and 620007)
4611 - journal_stop fixed with locking (611313)
4612 - O_APPEND failures resolved (618273, perhaps 614459)
4613 - lconf PATH fix (619770)
4614 - IA64 build fix (621450)
4615 - RPC buffer sizes scale with amount of memory
4617 2002-10-01 Phil Schwan <phil@clusterfs.com>
4620 - locks would be cancelled without throwing away data pages,
4621 resulting in inconsistent data (605627)
4622 - inode attributes were not always being refreshed (605627, 612449)
4623 - lconf now continues to cleanup after lctl reports an error
4624 - MDS now enforces user permissions (602707)
4625 - lprocfs cleanup fixed, but not yet enabled (614157)
4626 - fixed infinite server hang, should a client not respond to an AST
4627 - avoid going into recovery if user calls readlink() with a buffer
4628 that's too small (613941)
4629 - AST RPCs no longer require replies (614867) -- this may be changed
4630 - don't crash server if client sends an IOV that's too big (611336)
4631 - fixed lock conversion deadlock (611892)
4632 - fixed the following of symlinks (614622)
4633 * recovery: the server can remove locks from a client that dies, other
4634 clients can make progress
4635 * more extN patch fixes
4636 * compile-time configurable ptlrpc buffer allocations
4638 - collaborative read cache document
4639 - Lustre Lite Performance CDR document-in-progress
4641 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4644 - fix typo in patch-2.4.18
4646 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4649 - clear ptlrpc request each time in handle_incoming_request()
4650 - unlink of files now destroys the object on the OST
4652 2002-09-19 Peter Braam <braam@clusterfs.com>
4654 * add hard link support
4655 * change obdfile creation method
4656 * kernel patch changed
4658 2002-09-19 Peter Braam <braam@clusterfs.com>
4661 - stack overflow bug in extN fixed
4663 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4665 * documentation updates
4666 - add man pages for config tools
4667 - update tests/README to describe testing with new config tools
4668 - finish metadata API descriptions
4669 * bug fixes and cleanups
4670 - statfs workaround for 16TB limit
4671 - LOV stripe allocation improved, can stripe on subset of OSTs
4672 - LOV file size/IO offset was wrong for files > 4GB in size
4673 - object EA data was being dropped, caused files to be unreadable
4674 - memory overflow with non-LOV OST caused memory corruption
4675 - fixed regression tests to work with new config tools, obdfilter
4676 - fixed bug when directory size became larger than 1 block
4677 - fixed bug (for single client case) when PWD was deleted
4678 - invalidate local directory pages when doing intent-based ops
4679 - avoid LDLM oops when lock callback contained bad data
4681 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4683 * documentation updates
4684 * bug fixes and cleanups
4685 - configuration tools
4688 - 64-bit compile warnings
4689 - 64-bit internal statfs data
4691 * test_brw on persistent OST devices
4693 * lprocfs (disabled)
4695 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4697 * documentation updates
4698 * bug fixes and cleanups
4699 * configuration tools
4701 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4704 * many small fixes to 0_5_4
4705 * io/network handling
4706 * thinkos in MDS operations
4708 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4711 * crucial basic fixes to 0.5.3
4712 * IOR, Iozone work over Elan
4715 2002-08-07 Phil Schwan <phil@clusterfs.com>
4716 * version 0_5_3, our first alpha
4717 * we use the new Portals iovs
4718 * documentation updates
4719 * bug fixes and cleanups
4720 * small changes in the DLM wire protocol
4722 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4723 * version 0_5_1 with some initial stability,
4724 * locking on MD and file I/O.
4725 * documentation updates
4726 * several bug fixes since 0.5.0
4727 * small changes in wire protocol
4729 2002-07-18 Phil Schwan <phil@clusterfs.com>
4731 * delivered as Lustre Light Alpha
4732 * fixed a crash after handling invalid MDS requests
4733 * fixed directory pages for architectures with non-4k pages sizes
4735 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4736 * release version v0_4_4
4737 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4739 * Fixes a number of interruption problems with OST operations.
4740 * Update documentation for portals header changes
4741 * Move all wire protocol structs/defines to lustre_idl.h
4742 * Fixes symlink length bug.
4743 * Add tcpdump to repository.
4745 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4746 * release version v0_4_3
4747 * Fixes statfs for inodes on extN.
4748 * Fixes bug in runtests which would delete /etc/hosts.
4749 * Use 64-bit object IDs wherever possible (not into VFS though)
4750 Remove ost_get_info, which is unused by lustre, and out of date.
4752 2002-07-03 Peter Braam <braam@clusterfs.com>
4753 * release version v0_4_2 Fixes a lookup error (type not passed)
4754 * move forward to head of Portals
4755 * move forward to latest Lustre kernel
4757 2002-06-25 Peter Braam <braam@clusterfs.com>
4758 * release version v0_4_1. Hopefully stable on single node use.