1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
21 Frequency : rare, on shutdown ost
22 Description: don't hit live lock with umount ost.
23 Description: shrink_dcache_parent can be in long loop with destroy dentries,
24 use shrink_dcache_sb instead.
28 Description: don't panic with use echo client
29 Details : echo client pass NULL as client nid pointer and this produce null
34 Description: fix build on ppc32
35 Details : compile code with -m64 flag produce wrong object file for ppc32.
39 Description: add message levels for liblustreapi
43 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
44 Details : Move the 'good_osts' check before the 'total_bavail' check. This
45 will result in an -EAGAIN and in the exit call path we call
46 alloc_rr() which will with increasing aggressiveness attempt to
47 aquire precreated objects on the minimum number of required OSCs.
51 Description: Use old size assignment to avoid deadlock
52 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
53 the scheduling while holding a spinlock. We do not need locking
54 for size in ll_update_inode() because size is only updated from
55 the MDS for directories or files without objects, so there is no
56 other place to do the update, and concurrent access to such inodes
57 are protected by the inode lock.
61 Description: resolve "_IOWR redefined" build error on SLES10
65 Description: dump the memory debugging after all modules are unloaded to
66 suppress false negative in conf_sanity test 39
68 Severity : enhancement
70 Description: build kernel-ib packages for OFED 1.3 in our release cycle
75 Description: fix SLES kernel versioning
76 Details : the kernel version for our SLES 10 kernel did not include a "-"
77 before the "smp" at the end. while this was not a problem in
78 general, it did mean that software trying to use the kernel
79 version to try to detect a vendor specific kernel would fail.
80 this was most evident by the OFED build scripts.
84 Description: Don't update lov_desc members until making sure they are valid
85 Details : When updating lov_desc members via proc fs, need fix their
86 validities before doing the real update.
90 Description: don't put request into delay list while invalidate in flight.
91 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
92 import in flight. this produce timeout for invalidate and sometimes
97 Frequency : on ppc only
98 Description: not convert ost objects for directory because it's not exist.
99 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
103 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
105 Severity : enhancement
107 Description: Update to support 2.6.22.14 vanilla kernel.
111 Frequency : rare, on recovery
112 Description: read procfs can produce deadlock in some situation
113 Details : Holding lprocfs lock which send rpc can produce block for destroy
114 obd objects and this also block reconnect with -EALREADY. This isn't
115 fix all lprocfs bugs - but make it rare.
117 Severity : enhancement
119 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
122 Frequency : frequent on X2 node
124 Description: mdc_set_open_replay_data LBUG
125 Details : Set replay data for requests that are eligible for replay.
129 Description: lustre_mgs: operation 101 on unconnected MGS
130 Details : When MGC is disconnected from MGS long enough, MGS will evict the
131 MGC, and late on MGC cannot successfully connect to MGS and a lot
132 of the error messages complaining that MGS is not connected.
135 Frequency : on start mds
137 Description: Implement get_info(last_id) in obdfilter.
140 Frequency : occasional
142 Description: Correctly check stale fid, not start epoch if ost not support SOM
143 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
144 overwrited with new generated fid, but mds can anwer with one of these
145 two fids and both is not stale. setattr incorectly start epoch and
146 assume will be called done_writeting, but without SOM done_writing
150 Frequency : rare, depends on device drivers and load
152 Description: MDS or OSS nodes crash due to stack overflow
153 Details : Code changes in 1.8.0 increased the stack usage of some functions.
154 In some cases, in conjunction with device drivers that use a lot
155 of stack the MDS (or possibly OSS) service threads could overflow
156 the stack. One change which was identified to consume additional
157 stack has been reworked to avoid the extra stack usage.
160 Frequency : occasional
162 Description: Do not fail import if osc_interpret_create gets -EAGAIN
163 Details : If osc_interpret_create got -EAGAIN it immediately exits and
164 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
165 oscc_has_objects and see OSC has no objests and call
166 oscc_internal_create to resend create request.
168 Severity : enhancement
170 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
172 Severity : enhancement
174 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
177 Frequency : very rare
179 Description: Fix replay if there is an un-replied request and open
180 Details : In some cases, older replay request will revert the
181 mcd->mcd_last_xid on MDS which is used to record the client's
184 Severity : enhancement
186 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
188 Severity : enhancement
190 Description: Add rhel5 support to HEAD.
192 Serverity : enhancement
194 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
199 Description: Don't allow skipping OSTs if index has been specified.
200 Details : Don't allow skipping OSTs if index has been specified, make locking
201 in internal create lots better.
205 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
206 Details : access to bitfield in structure is always rounded to long
207 and this produce problem with not atomic change any bit.
211 Description: Lustre make rpms failed.
212 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
213 builds Lustre rpms from tarball.
216 Frequency : rare on shutdown ost
218 Description: If llog cancel was not send before clean_exports phase, this can
219 produce deadlock in llog code.
220 Details : If llog thread has last reference to obd and call class_import_put
221 this produce deadlock because llog_cleanup_commit_master wait when
222 last llog_commit_thread exited, but this never success because was
223 called from llog_commit_thread.
227 Description: allow userland application know is lost one of stripes.
228 Details : fill lvb_blocks with error code on ost and return it to
229 application if error flag found.
233 Description: NULL lov_tgts causing MDS oops
234 Details : more safe checks for NULL lov_tgts for avoid oops.
236 Severity : enhancement
238 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
242 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
243 Details : If an OST is down the MDS will hang indefinitely in
244 obd_statfs_rqset() waiting for the statfs data. While for
245 MDS QOS usage of statfs, it should not stuck in waiting.
247 Severity : enhancement
249 Description: remote_acl support
250 Details : Support ACL-based permission check for remote user.
251 Support setfacl/getfacl for remote user with the utils
252 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
253 the system "{s,g}etfacl" utils.
255 Severity : enhancement
257 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
259 Severity : enhancement
261 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
265 Description: make mgs_setparam() handle fsname containing dash
266 Details : fsname containing a dash does not work with lctl conf_param
268 Severity : enhancement
270 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
272 Severity : enhancement
274 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
278 Frequency : rare, at shutdown
279 Description: access already free / zero obd_namespace.
280 Details : if client_disconnect_export was called without force flag set,
281 and exist connect request in flight, this can produce access to
282 NULL pointer (or already free pointer) when connect_interpret
283 store ocd flags in obd_namespace.
287 Frequency : only at startup
288 Description: not alloc memory with spinlock held.
289 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
290 if any spinlock held.
292 Severity : enhancement
294 Description: make lustre randomly fail allocating memory
295 Details : Make lustre randomly failed allocating memory for testing purpose.
297 Severity : enhancement
299 Description: lost problems with lov objid file
300 Details : Fixes some scability and access to not inited memory problems
301 in work with lov objdid file.
306 Description: lfs find does not continue on file error
307 Details : Continue other files processing when a file/dir is absent.
311 Description: Inconsistent usage of lustre_pack_reply()
312 Details : Standardize the usage of lustre_pack_reply() such that it
313 always generate a CERROR on failure.
316 Frequency : occasional
318 Description: MDS hang or stay in waiting lock
319 Details : If client receive lock with CBPENDING flag ldlm need send lock
320 cancel as separate rpc, to avoid situation when cancel request
321 can't processed due all i/o threads stay in wait lock.
325 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
329 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
333 Description: Update to SLES9 kernel-2.6.5-7.287.3.
337 Description: Update to RHEL4 latest kernel.
339 Severity : enhancement
341 Description: Build SLES10 patchless client fails
342 Details : The configure was broken by run ./configure with
343 --with-linux-obj=.... argument for patchless client. When the
344 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
345 can't search header adequately. Use absolute path such as
346 -I($LINUX)/include instead.
350 Description: interrupt oig_wait produce painc on resend.
351 Details : brw_redo_request can be used for resend requests from ptlrpcd and
352 private set, and this produce situation when rq_ptlrpcd_data not
353 copyed to new allocated request and triggered LBUG on assert
354 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
355 ptlrpcd set if request is changed and can be safety changed to use
360 Description: LASSERT_{REQ,REP}SWAB macros are buggy
361 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
362 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
363 these macros and replace them with their difinitions instead.
367 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
368 Details : Update which_patch & target file for SLES10 latest kernel.
371 Description: Remove client patches from SLES 10 kernel.
372 Details : This causes SLES 10 clients to behave as patchless clients
373 even on a Lustre-patched (server) kernel.
375 Severity : enhancement
377 Description: self-adjustable client's lru lists
378 Details : use adaptive algorithm for managing client cached locks lru
379 lists according to current server load, other client's work
380 pattern, memory activities, etc. Both, server and client
381 side namespaces provide number of proc tunables for controlling
384 Severity : enhancement
386 Description: light-weight GSS support
387 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
388 and reduce performance overhead.
390 Severity : enhancement
392 Description: Linux keyring support
393 Details : Support using service of Linux keyring for Lustre GSS internal
394 context refresh/cache mechanism.
398 Description: Fix errors in lfs documentation
399 Details : Fixes man pages
403 Description: don't use GFP_* in generic Lustre code.
404 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
408 Description: obdclass is limited by single OBD_ALLOC(idarray)
409 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
413 Description: warnings with build patchless client with vanila 2.6.19 and up
414 Details : change old ctl_table style and replace ctl_table/ctl_table_header
415 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
419 Frequency : Only for SLES
420 Description: sanity_quota fail test_1
421 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
422 file, which makes TSTID[2] inunique.
426 Frequency : Only for SLES9
427 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
428 Details : Update target/ChangeLog/which_patch .
432 Frequency : Only for test_10 in sanity-quota.sh
433 Description: a bug in quota test code
434 Details : checking a test flag in a wrong place
438 Frequency : Only for test_13 in sanity-quota.sh
439 Description: a bug in quota test script
440 Details : shouldn't check overfull allocated quota
444 Description: jbd statistics
445 Details : Port older jbd statistics patch for sles10
447 Severity : enhancement
449 Description: Add printing inode info into message about error in writepage.
454 Description: reply_lock_interpret crash due to race with it and lock cancel.
455 Details : Do not replay locks that are being cancelled. Do not reference
456 locks by their address during replay, just by their handle.
460 Frequency : When flocks are used.
461 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
462 Details : Flock locks might destroy just granted lock if it could be merged
463 with another existing flock, this is done in completion handler,
464 so teach ldlm_cli_enquque_fini that this is a valid case for
470 Description: Oops in read and write path when failing to allocate lock.
471 Details : Check if lock allocation failed and return error back.
475 Description: lstripe command fails for valid OST index
476 Details : The stripe offset is compared to lov->desc.ld_tgt_count
477 instead of lov->desc.ld_active_tgt_count.
481 Description: sanity.sh failed test 103
482 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
487 Description: df doesn't work properly if diskfs blocksize != 4K
488 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
492 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
493 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
497 Description: obd_health_check_timeout too short
498 Details : set obd_health_check_timeout as 1.5x of obd_timeout
502 Description: llapi_file_create() does not allow some changes
503 Details : add llapi_file_open() that allows specifying the mode and
504 open flags, and also returns an open file handle.
508 Description: merge and cleanup kernel patches.
509 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
513 Description: Add journal checksum support.(Kernel part)
514 Details : The journal checksum feature adds two new flags i.e
515 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
516 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
517 indicates that the commit block contains the checksum for
518 the blocks described by the descriptor blocks. Now commit
519 record can be sent to disk without waiting for descriptor
520 blocks to be written to disk. This behavior is controlled
521 using JBD2_FEATURE_ASYNC_COMMIT flag.
525 Description: OSS needs mutliple precreate threads
526 Details : Add ability to start more than one create thread per OSS.
530 Description: Sanity.sh test_65k failed: setstripe should have succeeded
531 Details : execute lfs setstripe on client
535 Description: mds_obd_create error creating tmp object
536 Details : When the user sets quota on root, llog will be affected and can't
537 create files and write files.
540 Frequency : Always on ia64 patchless client, and possibly others.
542 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
543 Details : This allows the patchless client to be loaded on architectures
548 Description: RedHat Update kernel for RHEL5
549 Details : Add the kernel config file for RHEL5.
553 Description: RedHat Update kernel for RHEL5
554 Details : Modify the kernel config file more closer RHEL5.
558 Description: Build failure against Centos5 (RHEL5)
559 Details : Define PAGE_SIZE when it isn't present.
563 Description: test_7 on sanity-quota.sh can't be used on separate machines
564 Details : add facet to handle it
568 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
569 Details : with NFS, the anon dentry's parent was set to itself in
570 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
571 corresponding dentry other than use rec->ur_name.
573 Severity : enhancement
575 Description: lfs setstripe enhancement
576 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
580 Description: enable data checksumming by default
581 Details : enable checksum by default, allow --disable-checksum
582 configure option and "-o nochecksum" mount option. Checksums
583 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
587 Description: lustre support for RHEL5
588 Details : Add support for RHEL5.
592 Description: Client eviction due to failover config
593 Details : after a connection loss, the lustre client should attempt to
594 reconnect to the last active server first before trying the
595 other potential connections.
599 Description: when mds and osts use different quota unit(32bit and 64bit),
600 quota will be released repeatly.
601 Details : void sending multiple quota reqs to mds, which will keep the
602 status between the reqs.
606 Description: osts not allocated evenly to files
607 Details : change the condition to increase offset_idx
610 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
612 Description: Data corruption for OSTs that are formatted larger than 2TB
614 Details : When generating the bio request for lustre file writes the
615 sector number would overflow a temporary variable before being
616 used for the IO. The data reads correctly from Lustre (which
617 will overflow in a similar manner) but other file data or
618 filesystem metadata may be corrupted in some cases.
622 Description: Tune the kernel for good SCSI performance.
623 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
624 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
629 Description: rewrite ext2-derived code in obdclass/uuid.c
630 Details : rewrite inherited code (uuid parsing code from ext2 utils)
631 from scratch preserving functionality.
635 Description: "lfs find -obd UUID" prints directories
636 Details : "lfs find -obd UUID" will return all directory names instead
637 of just file names. It is incorrect because the directories
638 do not reside on the OSTs.
642 Description: use i_size_read and i_size_write in 2.6 port
643 Details : replace inode->i_size access with i_size_read/write()
646 Frequency : when removing large files
648 Description: scheduling issue during removal of large Lustre files
649 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
650 It causes scheduling issues when removing large files (17TB in the
654 Frequency : only with liblustre clients on XT3
656 Description: evictions taking too long
657 Details : allow llrd to evict clients directly on OSTs
660 Frequency : only on ppc
662 Description: /proc/fs/lustre/devices broken on ppc
663 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
664 We should make sure the type of 'index' is loff_t and then cast
665 explicitly as needed below. Do not assign an explicitly cast
669 Frequency : only for rhel5
671 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
672 Details : Modify the target file & which_kernel.
674 Severity : enhancement
676 Description: omit set fsid for export NFS
677 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
678 as FS_REQUIRES_DEV for avoid problems with generate fsid.
682 Frequency : Always, for kernels after 2.6.16
683 Description: Fix warning idr_remove called for id=.. which is not allocated.
684 Details : Last kernels save old s_dev before kill super and not allow
685 to restore from callback - restore it before call kill_anon_super.
689 Description: buffer overruns could theoretically occur
690 Details : llapi_semantic_traverse() modifies the "path" argument by
691 appending values to the end of the origin string, and a
692 overrun may occur. Adding buffer overrun check in liblustreapi.
696 Description: Fix error on 'ls .' at the top of the Lustre mount.
697 Details : Don't revalidate dentry if it is a root dentry.
701 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
702 Details : Modify vm-tunables-rhel4.patch.
706 Description: Kernel config for 2.6.18-vanilla.
707 Details : Modify targets/2.6-vanilla.target.in.
708 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
709 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
710 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
711 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
715 Description: improve handling recoverable errors
716 Details : if request processig with error which can be recoverable on server
717 request should be resend, otherwise page released from cache and
722 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
723 Details : Modify target file & which_patch.
724 A flaw was found in the IA32 system call emulation provided
725 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
726 value could be stored in the %RAX register, which could trigger an
727 out-of-bounds system call table access. An untrusted local user
728 could exploit this flaw to run code in the kernel
729 (ie a root privilege escalation). (CVE-2007-4573).
733 Description: O_DIRECT bypasses client statistics.
734 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
735 read_ahead_stats were updated. copied the stats section from
736 osc_send_oap_rpc() into async_internal().
740 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
744 Description: change order of libsysio includes
745 Details : '#include sysio.h' should always come before '#include xtio.h'
749 Description: handle "serious error: objid * already exists" more gracefully
750 Details : If LAST_ID value on disk is smaller than the objects existing in
751 the O/0/d* directories, it indicates disk corruption and causes an
752 LBUG(). If the object is 0-length, then we should use the existing
753 object. This will help to avoid a full fsck in most cases.
755 Severity : enhancement
757 Description: adapt the lustre_config script to support the upgrade case
758 Details : Add "-u" option for lustre_config script to support upgrading 1.4
759 server targets to 1.6 in parallel.
763 Description: To avoid grant space > avaible space when the disk is almost
764 full. Without this patch you might see the error "grant XXXX >
765 available" or some LBUG about grant, when the disk is almost
767 Details : In filter_check_grant, for non_grant cache write, we should
768 check the left space by if (*left > ungranted + bytes), instead
769 of (*left > ungranted), because only we are sure the left space
770 is enough for another "bytes", then the ungrant space should be
771 increase. In client, we should update cl_avail_grant only there
772 is OBD_MD_FLGRANT in the reply.
776 Description: Update RHEL 4 kernel to fix local root privilege escalation.
777 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
778 described in CVE-2007-4573. This problem could allow untrusted
779 local users to gain root access.
782 Frequency : when using O_DIRECT and quotas
784 Description: Incorrect file ownership on O_DIRECT output files
785 Details : block usage reported by 'lfs quota' does not take into account
786 files that have been written with O_DIRECT.
791 Description: touch file failed when fs is not full
792 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
793 otherwise we can get ENOSP while fs is not full.
797 Description: parallel lock callbacks
798 Details : Instead of sending blocking and completion callbacks as separated
799 requests, adding them to a set and sending in parallel.
802 Frequency : only for Cray XT3
803 Bugzilla : 12829/13455
804 Description: Changing primary group doesn't change the group lustre assigns to
806 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
807 overridden with the primary group provided by the group upcall,
808 whereas the supplied fsgid can be trusted if it is in the list of
809 supplementary groups returned by the group upcall.
811 Severity : enhancement
813 Description: Allow masking D_WARNING, D_ERROR messages from console
814 Details : Console messages can now be disabled via lnet.printk.
818 Description: User code with malformed file open parameter crashes client node
819 Details : Before packing join_file req, all the related reference should be
820 checked carefully in case some malformed flags cause fake
821 join_file req on client.
825 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
826 Details : ldlm_completion_ast() assumes that a lock is granted when the req
827 mode is equal to the granted mode. However, it should also check
828 that LDLM_FL_CP_REQD is not set.
832 Description: Heavy nfs access might result in deadlocks
833 Details : After ELC code landed, it is now improper to enqueue any mds
834 locks under och_sem, because enqueue might want to decide to
835 cancel open locks for same inode we are holding och_sem for.
839 Description: Client eviction while running blogbench
840 Details : A lot of unlink operations with concurrent I/O can lead to a
841 deadlock causing evictions. To address the problem, the number of
842 oustanding OST_DESTROY requests is now throttled to
843 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
844 callbacks are processed in priority.
848 Description: enable ACLs on MDS by default
849 Details : ACLs must be enabled on MDS by default.
852 Frequency : PPC/PPC64 only
854 Description: conflicts between asm-ppc64/types.h and lustre_types.h
855 Details : fix duplicated definitions between asm-ppc64/types.h and
856 lustre_types.h on PPC.
859 Frequency : PPC/PPC64 only
861 Description: asm-ppc/segment.h does not exist
862 Details : fix compile issue on PPC.
866 Description: better handle error messages in extents code
869 Frequency : RHEL4 only
871 Description: mkfs is very slow on IA64/RHEL4
872 Details : A performance regression has been discovered in the MPT Fusion
873 driver between versions 3.02.73rh and 3.02.99.00rh. As a
874 consequence, we have downgraded the MPT Fusion driver in the RHEL4
875 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
877 Severity : enhancement
879 Description: SNMP support enhancement
880 Details : Adding total number of sampled request for an MDS node in snmp
883 Severity : enhancement
885 Description: Optimize ldlm waiting list processing for PR extent locks
886 Details : When processing waiting list for read extent lock and meeting read
887 lock that is same or wider to it that is not contended, skip
888 processing rest of the list and immediatelly return current
889 status of conflictness, since we are guaranteed there are no
890 conflicting locks in the rest of the list.
894 Description: Time out and refuse to reconnect
895 Details : When the failover node is the primary node, it is possible
896 to have two identical connections in imp_conn_list. We must
897 compare not conn's pointers but NIDs, otherwise we can defeat
898 connection throttling.
902 Description: port llog fixes from b1_6 into HEAD
903 Details : Port llog reference couting and some llog cleanups from b1_6
904 (bug 10800) into HEAD, for protect from panic and access to already
905 free llog structures.
909 Description: Detect stride IO mode in read-ahead
910 Details : When a client does stride read, read-ahead should detect that and
911 read-ahead pages according to the detected stride pattern.
915 Description: data checksumming impacts single node performance
916 Details : add support for several checksum algorithm. Currently, only CRC32
917 and Adler-32 are supported. The checksum type can be changed on
918 the fly via /proc/fs/lustre/osc/*/checksum_type.
922 Description: use adler32 for page checksums
923 Details : when available, use the Adler-32 algorithm instead of CRC32 for
928 Description: build for x2 fails
929 Details : fix compile issue on Cray systems.
933 Description: Properly match for duplicate locks
934 Details : Due to different lock order from skiplists code, we need to
935 traverse entire list for now
938 Frequency : only on PPC/SLES10
940 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
941 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
942 be defined. Add a hack in mkfs_lustre.c to work around this compile
947 Description: LASSERT on MDS when client holding flock lock dies
948 Details : ldlm pool logic depends on number of granted locks equal to
949 number of released locks which is not true for flock locks, so
950 just exclude such locks from consideration.
954 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
955 Details : Use fsfilt_sync() for both the whole filesystem sync and
956 individual file sync to eliminate dangerous inode locking
957 with I_LOCK that can lead to a deadlock.
961 Description: performance in 1.6.3
962 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
963 sure that 1MB requests are not fragmented by the block layer.
965 Severity : enhancement
967 Description: organize the server-side client stats on per-nid basis
968 Details : Change the structure of stats under obdfilter and mds to
976 The "uuid"s file would list the uuids of _active_ exports.
977 And the clear entry is to clear all stats and stale nids.
979 Severity : enhancement
981 Description: eliminate client locks in face of contention
982 Details : file contention detection and lockless i/o implementation
987 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
989 --------------------------------------------------------------------------------
991 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
993 * Support for kernels:
994 2.6.5-7.283 (SLES 9),
995 2.6.9-55.EL (RHEL 4),
996 2.6.16.46-0.14 (SLES 10),
997 2.6.18.8 vanilla (kernel.org)
998 * Client support for unpatched kernels:
999 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1000 2.6.16 - 2.6.22 vanilla (kernel.org)
1001 * Due to recently discovered recovery problems, we do not recommend
1002 using patchless RHEL 4 clients with this or any earlier release.
1003 * Recommended e2fsprogs version: 1.39.cfs8
1004 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1005 * Starting with this release, the ldiskfs backing filesystem required
1006 by Lustre is now in its own package, lustre-ldiskfs. This package
1007 should be installed. It is versioned separately from Lustre and
1008 may be released separately in future.
1012 Description: block reactivating mgc import until all deactivates complete
1013 Details : Fix race when failing back MDT/MGS to itself (testing)
1016 Frequency : at statup only
1018 Description: mds_lov_synchronize race leads to various problems
1019 Details : simultaneous MDT->OST connections at startup can cause the
1020 sync to abort, leaving the OSC in a bad state.
1022 Severity : enhancement
1024 Description: add optional extra BUILD_VERSION info
1025 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1026 to override the lustre version.
1029 Frequency : 2.6.18 servers only
1031 Description: ll_kern_mount() doesn't release the module reference
1032 Details : The ldiskfs module reference count never drops down to 0
1033 because ll_kern_mount() doesn't release the module reference.
1038 Description: server LBUG when using old ost_num_threads parameter
1039 Details : Accept the old ost_num_threads parameter but warn that it
1040 is deprecated, and fix an off-by-one error that caused an LBUG.
1045 Description: Transient SCSI error results in persistent IO issue
1046 Details : iobuf->dr_error is not reinitialized to 0 between two
1050 Frequency : sometimes when underlying device returns I/O errors
1052 Description: OSTs not going read-only during write failures
1053 Details : OSTs are not remounted read-only when the journal commit threads
1054 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1055 instead of the ext3 wrappers.
1059 Description: poor connect scaling with increasing client count
1060 Details : Don't run filter_grant_sanity_check for more than 100 exports
1061 to improve scaling for large numbers of clients.
1064 Frequency : SLES10 only
1066 Description: sanity-quota.sh quotacheck failed: rc = -22
1067 Details : Quotas cannot be enabled on SLES10.
1070 Frequency : liblustre clients only
1072 Description: getdirentries does not give error when run on compute nodes
1073 Details : getdirentries does not fail when the size specified as an argument
1074 is too small to contain at least one entry
1076 Severity : enhancement
1078 Description: Add LNET router traceability for debug purposes
1079 Details : If a checksum failure occurs with a router as part of the
1080 IO path, the NID of the last router that forwarded the bulk data
1081 is printed so it can be identified.
1086 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1087 Details : Due to a race condition, liblustre clients were occasionally
1088 evicted incorrectly.
1090 Severity : enhancement
1092 Description: lfs setstripe use optional parameters instead of postional
1095 Severity : enhancement
1097 Description: Nanosecond timestamp support for ldiskfs
1098 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1099 resolution timestamps. There is not yet support for this at
1100 the Lustre filesystem level.
1103 Frequency : during server recovery
1105 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1106 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1110 Frequency : nfs export on patchless client
1112 Description: connectathon hang when test nfs export over patchless client
1113 Details : Disconnected dentry cannot be found with lookup, so we do not need
1114 to unhash it or make it invalid
1117 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1118 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1119 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1124 Description: open req refcounting wrong on reconnect
1125 Details : If reconnect happened between getting open reply from server and
1126 call to mdc_set_replay_data in ll_file_open, we will schedule
1127 replay for unreferenced request that we are about to free.
1128 Subsequent close will crash in variety of ways.
1129 Check that request is still eligible for replay in
1130 mdc_set_replay_data().
1135 Description: disable writes to filesystem when reading health_check file
1136 Details : the default for reading the health_check proc file has changed
1137 to NOT do a journal transaction and write to disk, because this
1138 can cause reads of the /proc file to hang and block HA state
1139 checking on a healthy but otherwise heavily loaded system. It
1140 is possible to return to the previous behaviour during configure
1141 with --enable-health-write.
1143 Severity : enhancement
1145 Description: 64-bit inode version
1146 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1147 to the inode. This will be required for version-based recovery.
1152 Description: MDS fails to start if a duplicate client export is detected
1153 Details : in some rare cases it was possible for a client to connect to
1154 an MDS multiple times. Upon recovery the MDS would detect this
1155 and fail during startup. Handle this more gracefully.
1157 Severity : enhancement
1159 Description: Add -o localflock option to simulate old noflock
1161 Details : This will achieve local-only flock/fcntl locks
1167 Description: log_commit_thread vs filter_destroy race leads to crash
1168 Details : Take import reference before releasing llog record semaphore
1173 Description: Wrong request locking in request set processing
1174 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1175 imp_delayed_list, in this place should be used imp_lock.
1178 Frequency : when reconnection
1180 Description: Grant Leak when osc reconnect to OST
1181 Details : When osc reconnect ost, OST(filter) should check whether it
1182 should grant more space to client by comparing fed_grant and
1183 cl_avail_grant, and return the granted space to client instead
1184 of "new granted" space, because client will call osc_init_grant
1185 to update the client grant space info.
1188 Frequency : when client reconnect to OST
1190 Description: Grant Leak when osc do resend and replay bulk write
1191 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1192 bulk write request, because the grant info while be sync between
1193 OSC and OST when reconnect, and we should ignore the grant info
1194 these of resend/replay write req.
1199 Description: Grant space more than avaiable left space sometimes.
1200 Details : When then OST is about to be full, if two bulk writing from
1201 different clients came to OST. Accord the avaliable space of the
1202 OST, the first req should be permitted, and the second one
1203 should be denied by ENOSPC. But if the seconde arrived before
1204 the first one is commited. The OST might wrongly permit second
1205 writing, which will cause grant space > avaiable space.
1208 Frequency : when client is evicted
1210 Description: Grant might be wrongly erased when osc is evicted by OST
1211 Details : when the import is evicted by server, it will fork another
1212 thread ptlrpc_invalidate_import_thread to invalidate the
1213 import, where the grant will be set to 0. While the original
1214 thread will update the grant it got when connecting. So if
1215 the former happened latter, the grant will be wrongly errased
1216 because of this race.
1221 Description: Checking Stale with correct fid
1222 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1223 to check whether it is stale, because sometimes, we want the
1224 enqueue happened anyway, and op_data.fid2 will not be initialized.
1226 Severity : enhancement
1228 Description: update patchless client
1229 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1232 Frequency : only with 2.4 kernel
1234 Description: random memory corruption
1235 Details : size of struct ll_inode_info is to big for union inode.u and this
1236 can be cause of random memory corruption.
1241 Description: Memory leak in recovery
1242 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1243 It should also check obd_fail before fsfilt_start, otherwise if
1244 fsfilt_start return -EROFS,(failover mds during mds recovery).
1245 then the req will return with repmsg->transno = 0 and rc = EROFS.
1246 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1247 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1248 be freed no matter whether fsfilt_commit success or not.
1251 Frequency : only with huge count clients
1253 Description: Prevents from taking the superblock lock in llap_from_page for
1255 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1256 for llap_from_page call in ll_removepage prevents from taking the
1257 superblock lock for a soon died page.
1262 Description: Not check open intent error before release open handle
1263 Details : in some rare cases, the open intent error is not checked before
1264 release open handle, which may cause
1265 ASSERTION(open_req->rq_transno != 0), because it tries to release
1266 the failed open handle.
1271 Description: Set cat log bitmap only after create log success.
1272 Details : in some rare cases, the cat log bitmap is set too early. and it
1273 should be set only after create log success.
1277 Description: Accessing a block bevice can re-enable I/O when Lustre is
1278 tearing down a device.
1279 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1284 Description: service threads may hog cpus when there are a lot of requests
1285 Details : Insert cond_resched to give other threads a chance to use some CPU
1290 Description: the cat log was not initialized in recovery
1291 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1292 unlink log on mds will not be initialized until mds post
1293 recovery. And also in mds post recovery, the unlink log will
1294 initialization will be done asynchronausly, so there will be race
1295 between add unlink log and unlink log initialization.
1299 Description: brw_stats were being printed incorrectly
1300 Details : brw_stats were being printed as log2 but all of them were not
1301 recorded as log2. Also remove some code duplication arising from
1302 filter_tally_{read,write}.
1306 Frequency : rare, only in recovery.
1307 Description: ASSERTION(req->rq_type != LI_POISON) failed
1308 Details : imp_lock should be held while iterating over imp_sending_list for
1309 prevent destroy request after get timeout in ptlrpc_queue_wait.
1313 Description: replay-single.sh test 52 fails
1314 Details : A lock's skiplist need to be cleanup when it being unlinked
1315 from its resource list.
1319 Description: Short directio read returns full requested size rather than
1321 Details : Direct I/O operations should return actual amount of bytes
1322 transferred rather than requested size.
1324 Severity : enhancement
1326 Description: metadata RPC reduction (e.g. for rm performance)
1327 Details : decrease the amount of synchronous RPC between clients and servers
1328 by canceling conflicing lock before the operation on the client
1329 and packing thier handles into the main operation RPC to server.
1331 Severity : enhancement
1333 Description: Async OSC create to avoid the blocking unnecessarily.
1334 Details : If a OST has no remain object, system will block on the creating
1335 when need to create a new object on this OST. Now, ways use
1336 pre-created objects when available, instead of blocking on an
1337 empty osc while others are not empty. If we must block, we block
1338 for the shortest possible period of time.
1340 Severity : enhancement
1342 Description: refine locking for avoid write wrong info into lov_objid file
1343 Details : fix possible races with add new target and write/update data in
1346 --------------------------------------------------------------------------------
1348 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1353 Frequency : on some architectures
1355 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1356 Details : Uninitialized flags sometimes cause configuration commands to
1360 Frequency : patchless clients only
1362 Description: missing __iget() symbol export
1363 Details : The __iget() symbol export is missing. To avoid the need for
1364 this on patchless clients the deathrow inode reaper is turned
1365 off, and we depend on the VM to clean up old inodes. This
1366 dependency was during via the fix for bug 12181.
1368 --------------------------------------------------------------------------------
1370 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1372 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1373 INTEROPERATE with older versions automatically. In many cases a
1374 special upgrade step is needed. Please read the
1375 user documentation before upgrading any part of a live system.
1376 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1377 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1378 * WARNING: Lustre configuration and startup changes are required with
1379 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1385 Severity : enhancement
1387 Description: MountConf
1388 Details : Lustre configuration is now managed via mkfs and mount
1389 commands instead of lmc and lconf. New obd types (MGS, MGC)
1390 are added for dynamic configuration management. See
1391 https://mail.clusterfs.com/wikis/lustre/MountConf for
1394 Severity : enhancement
1396 Description: dynamic OST addition
1397 Details : OSTs can now be added to a live filesystem
1399 Severity : enhancement
1401 Description: startup order invariance
1402 Details : MDTs and OSTs can be started in any order. Clients only
1403 require the MDT to complete startup.
1405 Severity : enhancement
1407 Description: parallel, asynchronous orphan cleanup
1408 Details : orphan cleanup is now performed in separate threads for each
1409 OST, allowing parallel non-blocking operation.
1411 Severity : enhancement
1413 Description: optimized stripe assignment
1414 Details : stripe assignments are now made based on ost space available,
1415 ost previous usage, and OSS previous usage, in order to try
1416 to optimize storage space and networking resources.
1418 Severity : enhancement
1420 Description: Permanently set tunables
1421 Details : All writable /proc/fs/lustre tunables can now be permanently
1422 set on a per-server basis, at mkfs time or on a live system.
1424 Severity : enhancement
1426 Description: Lustre message v2
1427 Details : Add lustre message format v2.
1429 Severity : enhancement
1431 Description: client OST exclusion list
1432 Details : Clients can be started with a list of OSTs that should be
1433 declared "inactive" for known non-responsive OSTs.
1437 Description: SPEC SFS validation failure on NFS v2 over lustre.
1438 Details : Changes the blocksize for regular files to be 2x RPC size,
1439 and not depend on stripe size.
1441 Severity : enhancement
1443 Description: Multiple MD RPCs in flight.
1444 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1445 To avoid overly-overloading MDS, introduce a limit on number of
1446 MDS RPCs in flight for a single client and add /proc controls
1447 to adjust this limit.
1449 Severity : enhancement
1451 Description: client read/write statistics
1452 Details : Add client read/write call usage stats for performance
1453 analysis of user processes.
1454 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1455 file access. extents_stats shows chunk size distribution.
1456 extents_stats_per_process show chunk size distribution per
1459 Severity : enhancement
1461 Description: mds statistics
1462 Details : Add detailed mds operations statistics in
1463 /proc/fs/lustre/mds/*/stats.
1467 Description: Failure of copying files with lustre special EAs.
1468 Details : Client side always return success for setxattr call for lustre
1469 special xattr (currently only "trusted.lov").
1474 Description: Refcount LNET uuids
1475 Details : The global LNET uuid list grew linearly with every startup;
1476 refcount repeated list entries instead of always adding to
1479 Severity : enhancement
1481 Description: Dynamic service threads
1482 Details : Within a small range, start extra service threads
1483 automatically when the request queue builds up.
1486 Frequency : mixed-endian client/server environments
1488 Description: mixed-endian crashes
1489 Details : The new msg_v2 system had some failures in mixed-endian
1492 Severity : enhancement
1494 Description: Easy OST removal
1495 Details : OSTs can be permanently deactivated with e.g. 'lctl
1496 conf_param lustre-OST0001.osc.active=0'
1498 Severity : enhancement
1500 Description: MGS proc entries
1501 Details : Added basic proc entries for the MGS showing what filesystems
1504 Severity : enhancement
1506 Description: provide MGS failover
1507 Details : Added config lock reacquisition after MGS server failover.
1509 Severity : enhancement
1511 Description: add Linux 2.4 support
1512 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1516 Description: a large application tries to do I/O to the same resource and dies
1517 in the middle of it.
1518 Details : Check the req->rq_arrival time after the call to
1519 ost_brw_lock_get(), but before we do anything about
1520 processing it & sending the BULK transfer request. This
1521 should help move old stale pending locks off the queue as
1522 quickly as obd_timeout.
1525 Frequency : when an incorrect nid is specified during startup
1527 Description: ptlrpc connect to non-existant node causes kernel crash
1528 Details : LNET can't be re-entered from an event callback, which
1529 happened when we expire a message after the export has been
1530 cleaned up. Instead, hand the zombie cleanup off to another
1533 Severity : enhancement
1535 Description: plain/inodebits lock performance improvement
1536 Details : Grouping plain/inodebits in granted list by their request modes
1537 and bits policy, thus improving the performance of search through
1541 Frequency : only if OST filesystem is corrupted
1543 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1544 Details : for a short time RPCs with bulk IO are in the replay list,
1545 but replay of bulk IOs is unimplemented. If the OST filesystem
1546 is corrupted due to disk cache incoherency and then replay is
1547 started it is possible to trip an assertion. Avoid putting
1548 committed RPCs into the replay list at all to avoid this issue.
1551 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1553 Description: System hang on startup
1554 Details : This bug allowed the liblustre (e.g. catamount) client to
1555 return to the app before handling all startup RPCs. This
1556 could leave the node unresponsive to lustre network traffic
1557 and manifested as a server ptllnd timeout.
1559 Severity : enhancement
1561 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1562 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1563 boolean to control whether to print peer debug info when a
1564 client's RPC times out.
1567 Frequency : only for kernels with patches from Lustre below 1.4.3
1569 Description: Remove old rdonly API
1570 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1573 Frequency : only for devices with external journals
1575 Description: Set external device read-only also
1576 Details : During a commanded failover stop, we set the disk device
1577 read-only while the server shuts down. We now also set any
1578 external journal device read-only at the same time.
1581 Frequency : when upgrading from 1.4 while trying to change parameters
1583 Description: The wrong (new) MDC name was used when setting parameters for
1584 upgraded MDT's. Also allows changing of OSC (and MDC)
1585 parameters if --writeconf is specified at tunefs upgrade time.
1588 Frequency : when setting specific ost indicies
1590 Description: QOS code breaks on skipped indicies
1591 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1592 created with --index need not be sequential.
1594 Severity : enhancement
1596 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1597 Details : The uninit_groups feature works in conjunction with the kernel
1598 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1599 up the pass1 processing of e2fsck. This is a read-only feature
1600 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1601 mount filesystems that have had this feature enabled.
1603 Severity : enhancement
1605 Description: Improve multi-block allocation algorithm to avoid fragmentation
1606 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1607 allocation locality and avoid filesystem fragmentation.
1609 ------------------------------------------------------------------------------
1611 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1613 * Support for kernels:
1614 2.4.21-47.0.1.EL (RHEL 3)
1615 2.6.5-7.283 (SLES 9)
1616 2.6.9-42.0.10.EL (RHEL 4)
1617 2.6.12.6 vanilla (kernel.org)
1618 2.6.16.27-0.9 (SLES 10)
1619 * Recommended e2fsprogs version: 1.39.cfs5
1621 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1625 Frequency : occasional, depends on client load and configuration
1626 Bugzilla : 12181, 12203
1627 Description: data loss for recently-modified files
1629 Details : In some cases it is possible that recently written or created
1630 files may not be written to disk in a timely manner (this should
1631 normally be within 30s unless client IO load is very high).
1632 The problem appears as zero-length files or files that are a
1633 multiple of 1MB in size after a client crash or client eviction
1634 that are missing data at the end of the file.
1636 This problem is more likely to be hit on clients where files are
1637 repeatedly created and unlinked in the same directory, clients
1638 have a large amount of RAM, have many CPUs, the filesystem has
1639 many OSTs, the clients are rebooted frequently, and/or the files
1640 are not accessed by other nodes after being written.
1642 The presence of the problem can be detected by looking at
1643 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1644 smaller than the second (nr_unused) then dirty files will not
1645 be flushed automatically to disk. "sync; sleep 10" should be
1646 run several times on the node before unmounting it to update
1647 Lustre (this is also safe to run on nodes without this problem).
1649 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1650 that can cause this same problem, so customers using that kernel
1651 also need to update the kernel in addition to Lustre. In order
1652 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1654 It is normal that files written just before a client crash (less
1655 than 30s) may not yet have been flushed to disk, even for local
1659 Frequency : frequent on thin XT3 nodes
1661 Description: UUID collision on thin XT3 Linux nodes
1662 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1663 randomly, since we relied on an insufficiently-seeded PRNG.
1668 Description: OSS hangs after "All ost request buffers busy"
1669 Details : A deadlock between quota and journal operations caused OSS
1670 hangs after printing "All ost request buffers busy."
1673 Frequency : always on liblustre builds
1675 Description: Cleanup compiler warnings on liblustre
1678 Frequency : always on liblustre builds on XT3
1680 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1684 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1685 > 32000 subdirectories
1686 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1687 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1688 subdirectory count crosses 32000. This will aid e2fsck to
1689 correctly handle more than 32000 subdirectories.
1692 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1694 Description: System hang on startup
1695 Details : This bug allowed the liblustre (e.g. catamount) client to
1696 return to the app before handling all startup RPCs. This
1697 could leave the node unresponsive to lustre network traffic
1698 and manifested as a server ptllnd timeout.
1700 Severity : enhancement
1702 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1703 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1704 boolean to control whether to print peer debug info when a
1705 client's RPC times out.
1710 Description: make O_SYNC working on 2.6 kernels
1711 Details : 2.6 kernels use different method for mark pages for write,
1712 so need add a code to lustre for O_SYNC work.
1717 Description: Failure to close file and release space on NFS
1718 Details : Put inode details into lock acquired in ll_intent_file_open.
1719 Use mdc_intent_lock in ll_intent_open to properly
1720 detect all kind of errors unhandled by mdc_enqueue.
1725 Description: proc file read during shutdown sometimes raced obd removal,
1727 Details : Add lock to prevent obd access after proc file removal.
1730 Frequency : Only for files larger than 4GB on 32-bit clients.
1732 Description: improperly doing page alignment of locks
1733 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1734 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1739 Description: RPCs being resent when they shouldn't be
1740 Details : Some RPCs that should not be resent are being resent. This
1741 can cause inconsistencies in the RPC state machine. Do not
1742 resend such requests.
1745 Frequency : rare, only with NFS export
1747 Description: Crash on NFS re-export node
1748 Details : under very unusual load conditions an assertion is hit in
1749 ll_intent_file_open()
1752 Frequency : only if OST filesystem is corrupted
1754 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1755 Details : for a short time RPCs with bulk IO are in the replay list,
1756 but replay of bulk IOs is unimplemented. If the OST filesystem
1757 is corrupted due to disk cache incoherency and then replay is
1758 started it is possible to trip an assertion. Avoid putting
1759 committed RPCs into the replay list at all to avoid this issue.
1764 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1765 Details : Large single O_DIRECT read and write calls can fail to allocate
1766 a sufficiently large buffer to process the request. In case of
1767 allocation failure the allocation is retried with a smaller
1768 buffer and broken into smaller requests.
1770 Severity : enhancement
1772 Description: Add -o localflock option to simulate old noflock behaviour.
1773 Details : This will achieve local-only flock/fcntl locks coherentness.
1778 Description: versioning check is incomplete
1779 Details : Checking the version difference of client vs. server, report
1780 error if the gap is too big.
1785 Description: add support PG_writeback bit
1786 Details : add support for PG_writeback bit for Lustre, for more carefull
1787 work with page cache in 2.6 kernel. This also fix some deadlocks
1788 and remove hack for work O_SYNC with 2.6 kernel.
1790 Severity : enhancement
1792 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1793 Details : The uninit_groups feature works in conjunction with the kernel
1794 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1795 up the pass1 processing of e2fsck. This is a read-only feature
1796 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1797 mount filesystems that have had this feature enabled.
1799 Severity : enhancement
1801 Description: Improve multi-block allocation algorithm to avoid fragmentation
1802 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1803 allocation locality and avoid filesystem fragmentation.
1805 ------------------------------------------------------------------------------
1807 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1809 * Support for kernels:
1810 2.6.9-42.0.3.EL (RHEL 4)
1811 2.6.5-7.276 (SLES 9)
1812 2.4.21-47.0.1.EL (RHEL 3)
1813 2.6.12.6 vanilla (kernel.org)
1814 2.6.16.21-0.8 (SLES10)
1815 * Recommended e2fsprogs version: 1.39.cfs2-0
1817 * The backwards-compatible /proc/sys/portals symlink has been removed
1818 in this release. Before upgrading, please ensure that you change
1819 any configuration scripts or /etc/sysctl.conf files that access
1820 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1821 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1822 in advance of the upgrade on any system running Lustre 1.4.6 or
1823 newer, since /proc/sys/lnet was added in that version.
1824 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1828 Frequency : only when quota is used
1830 Description: avoid scanning export list for quota master
1831 Details : Change the algorithms to avoid scanning export list in order
1832 to improve the efficiency.
1835 Frequency : MDS failover only, very rarely
1837 Description: "went back in time" messages on mds failover
1838 Details : The greatest transno may be lost when the current operation
1839 finishes with an error (transno==0) and the client's last_rcvd
1840 record is over-written. Save the greatest transno in the
1841 mds_last_transno for this case.
1844 Frequency : always for specific kernels and striping counts
1846 Description: client may get "Matching packet too big" without ACL support
1847 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1848 when trying to access files in certain configurations. The
1849 clients should in fact be denied when mounting because they do
1850 not understand ACLs.
1853 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1855 Description: many clients connecting with IO in progress causes connect timeouts
1856 Details : Avoid synchronous journal commits to avoid delays caused by many
1857 clients connecting/disconnecting when bulk IO is in progress.
1858 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1859 OST_IO_PORTAL to avoid delays behind potentially many pending
1863 Frequency : occasionally with multiple writers to a single file
1865 Description: shared writes to file may result in wrong size reported by stat()
1866 Details : Allow growing of kms when extent lock is cancelled
1869 Frequency : always with random mmap IO to multi-striped file
1871 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1872 Details : Only if the hole is at the end of OST object so that kms is too
1873 small. Fix is to increase kms accordingly in ll_nopage.
1876 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1878 Description: writes to a missing object would leak memory on the OST
1879 Details : If there is an inconsistency between the MDS and OST filesystems,
1880 such that the MDS references an object that doesn't exist, writes
1881 to that object will leak memory due to incorrect cleanup in the
1882 error handling path, eventually running out of memory on the OST.
1887 Description: Creating too long symlink causes lustre errors
1888 Details : Check symlink and name lengths before sending requests to MDS.
1891 Frequency : only if flock is enabled (not on by default)
1893 Description: posix locks not released on fd closure on 2.6.9+
1894 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1895 kernels, this caused such locks not to be released on fd close and
1896 then assertions on fs unmount about still used locks.
1899 Frequency : MDS failover only, very rarely
1901 Description: clients may get ASSERTION(granted_lock != NULL)
1902 Details : When request was taking a long time, and a client was resending
1903 a getattr by name lock request. The were multiple lock requests
1904 with the same client lock handle and
1905 mds_getattr_name->fixup_handle_for_resent_request found one of the
1906 lock handles but later failed with ASSERTION(granted_lock != NULL).
1911 Description: handle->h_buffer_credits > 0, assertion failure
1912 Details : h_buffer_credits is zero after truncate, causing assertion
1913 failure. This patch extends the transaction or creates a new
1917 Frequency : NFS re-export or patchless client
1918 Bugzilla : 11179, 10796
1919 Description: Crash on NFS re-export node (__d_move)
1920 Details : We do not want to hash the dentry if we don't have a lock.
1921 But if this dentry is later used in d_move, we'd hit uninitialised
1922 list head d_hash, so we just do this to init d_hash field but
1923 leave dentry unhashed.
1926 Frequency : NFS re-export or patchless client
1928 Description: NFS exports has problem with symbolic link
1929 Details : lustre client didn't properly install dentry when re-exported
1930 to NFS or running patchless client.
1933 Frequency : NFS re-export or patchless client
1935 Description: Various nfs/patchless fixes.
1936 Details : fixes reuse disconected alias for lookup process - this fixes
1937 warning "find_exported_dentry: npd != pd",
1938 fix permission error with open files at nfs.
1939 fix apply umask when do revalidate.
1942 Frequency : occasional
1944 Description: Crash on NFS re-export node
1945 Details : calling clear_page() on the wrong pointer triggered oops in
1946 generic_mapping_read().
1949 Frequency : rarely, using O_DIRECT IO
1951 Description: unaligned directio crashes client with LASSERT
1952 Details : check for unaligned buffers before trying any requests.
1955 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1957 Description: stale data returned from RAID cache
1958 Details : If only a small amount of IO is done to the RAID device before
1959 reading it again it is possible to get stale data from the RAID
1960 cache instead of reading it from disk.
1963 Frequency : always for sles10 kernel
1965 Description: sles10 support
1966 Details : ll_follow_link: compile fixes and using of nd_set_link
1967 under newer kernels.
1970 Frequency : depends on arch, kernel and compiler version, always on sles10
1973 Description: recursive or deep enough symlinks cause stack overflow
1974 Details : getting rid of large stack-allocated variable in
1978 Frequency : depends on hardware
1980 Description: lustre write performance loss in the SLES10 kernel
1981 Details : the performance loss is caused by using of write barriers in the
1982 ext3 code. The SLES10 kernel turns barrier support on by
1983 default. The fix is to undo that change for ldiskfs.
1985 ------------------------------------------------------------------------------
1987 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1989 * Support for kernels:
1990 2.6.9-42.0.3EL (RHEL 4)
1991 2.6.5-7.276 (SLES 9)
1992 2.4.21-47.0.1.EL (RHEL 3)
1993 2.6.12.6 vanilla (kernel.org)
1997 Frequency : quota enabled and large files being deleted
1999 Description: releasing more than 4GB of quota at once hangs OST
2000 Details : If a user deletes more than 4GB of files on a single OST it
2001 will cause the OST to spin in an infinite loop. Release
2002 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2007 Description: statfs data retrieved from /proc may be stale or zero
2008 Details : When reading per-device statfs data from /proc, in the
2009 {kbytes,files}_{total,free,avail} files, it may appear
2010 as zero or be out of date.
2013 Frequency : systems with MD RAID1 external journal devices
2015 Description: lconf's call to blkid is confused by RAID1 journal devices
2016 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2017 of returning all block devices that match the journal UUID.
2020 Frequency : always, for aggregate stripe size over 4GB
2022 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2023 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2024 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2027 Frequency : always if "lfs find" used on a local file/directory
2029 Description: "lfs find" segfaults if used on a local file/directory
2030 Details : The case where a directory component was not specified wasn't
2031 handled correctly. Handle this properly.
2034 Frequency : always on ppc64
2036 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2037 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2038 perform endianness conversion.
2041 Frequency : rarely (truncate to non-zero file size after write under load)
2042 Bugzilla : 10730, 10687
2043 Description: Files padded with zeros to next 4K multiple
2044 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2045 it is possible that files that are truncated to a non-zero size
2046 immediately after being written are filled with zero bytes beyond
2047 the truncated size. No file data is lost.
2049 Severity : enhancement
2051 Description: Allow recovery/failover for liblustre clients.
2052 Details : liblustre clients were unaware of failover configurations until
2055 Severity : enhancement
2057 Description: user file locks should fail when not mounting with flock option
2058 Details : Set up an error-returning stub in ll_file_operations.lock field
2059 to prevent incorrect behaviour when client is mounted without
2060 flock option. Also, set up properly f_op->flock field for
2064 Frequency : always on ia64
2066 Description: "lfs df" loops on printing out MDS statfs information
2067 Details : The obd_ioctl_data was not initialized and in some systems
2068 this caused a failure during the ioctl that did not return
2069 an error. Initialize the struct and return an error on failure.
2072 Frequency : SLES 9 only
2074 Description: Error of copying files with lustre special EAs as root
2075 Details : Client side always return success for setxattr call for lustre
2076 special xattr (currently only "trusted.lov").
2079 Frequency : rarely on clusters with both ia64+i386 clients
2081 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2082 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2083 clients are doing concurrent writes to the same file it is
2084 possible that the ia64 clients may LASSERT because the OST
2085 extent locks are not PAGE_SIZE aligned. Ensure that grown
2086 locks are always aligned on the request boundary.
2089 Frequency : specific use, occasional
2091 Description: Overwriting in use executable truncates on-disk binary image
2092 Details : If one node attempts to overwrite an executable in use by
2093 another node, we now correctly return ETXTBSY instead of
2094 truncating the file.
2096 Severity : enhancement
2098 Description: Async OSC create to avoid the blocking unnecessarily.
2099 Details : If a OST has no remain object, system will block on the creating
2100 when need to create a new object on this OST. Now, ways use
2101 pre-created objects when available, instead of blocking on an
2102 empty osc while others are not empty. If we must block, we block
2103 for the shortest possible period of time.
2108 Description: chmod on Lustre root is propagated to other clients
2109 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2110 invalid by the follow_mount time.
2113 Frequency : liblustre clients only
2115 Description: Race in 'instant cancel' lock handling could lead to such locks
2116 never to be granted in case of SMP MDS
2117 Details : Do not destroy not yet granted but cbpending locks in
2121 Frequency : replay/resend of open
2123 Description: non null lock assetion failure in mds_intent_policy
2124 Details : Trying to replay/resend lockless open requests resulted in
2125 mds_open() returning 0 with no lock. Now it sets a flag if
2126 a lock is going to be returned.
2128 Severity : enhancement
2130 Description: Checksum enhancements
2131 Details : New checksum enhancements allow for resending RPCs that failed
2134 Severity : enhancement
2136 Description: Tunables on number of dirty pages in cacche
2137 Details : Allow to set limit on number of dirty pages cached.
2142 Description: client crash on unmount - lock still has references
2143 Details : In some error handling cases it was possible to leak a lock
2144 reference on a client while accessing a file. This was not
2145 harmful to the client during operation, but would cause the
2146 client to crash when the filesystem is unmounted.
2149 Frequency : specific case, rare
2151 Description: ETXTBSY on mds though file not in use by client
2152 Details : ETXTBSY is no longer incorrectly returned when attempting to
2153 chmod or chown a directory that the user previously tried to
2154 execute or a currently-executing binary.
2157 Frequency : extremely rare except on liblustre-based clients
2159 Description: Lustre space not freed when files are deleted
2160 Details : Clean up open-unlinked files after client eviction. Previously
2161 the unlink was skipped and the files remained as orphans.
2166 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2167 Details : In some cases it was possible to send a blocking callback to
2168 a client doing a glimpse, even though that client didn't get
2169 a lock granted. When the glimpse lock is cancelled on the OST
2170 the freed lock is left on the waiting list and corrupted the list.
2173 Frequency : all core dumps
2175 Description: Broke core dumps to lustre
2176 Details : Negative dentry may be unhashed if parent does not have UPDATE
2177 lock, but some callers, e.g. do_coredump, expect dentry to be
2178 hashed after successful create, hash it in ll_create_it.
2180 ------------------------------------------------------------------------------
2182 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2184 * Support for kernels:
2185 2.6.9-42.0.2.EL (RHEL 4)
2186 2.6.5-7.276 (SLES 9)
2187 2.4.21-40.EL (RHEL 3)
2188 2.6.12.6 vanilla (kernel.org)
2192 Frequency : always on RHEL 3
2194 Description: Number of open files grows over time
2195 Details : The number of open files grows over time, whether or not
2196 Lustre is started. This was due to a filp leak introduced
2197 by one of our kernel patches.
2199 ------------------------------------------------------------------------------
2201 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2203 * Support for kernels:
2204 2.6.9-42.EL (RHEL 4)
2205 2.6.5-7.276 (SLES 9)
2206 2.4.21-40.EL (RHEL 3)
2207 2.6.12.6 vanilla (kernel.org)
2212 Bugzilla : 5719, 9635, 9792, 9684
2213 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2214 Details : If a server is under heavy load and cannot reply to new
2215 connection requests before the client resends the (re)connect,
2216 the connection handling code can behave badly if two service
2217 threads are concurrently handing separate (re)connections from
2218 the same client. Add better locking to the connection handling
2219 code, and ensure that only a single connection will be processed
2220 for a given client UUID, even if the lock is dropped.
2222 Severity : enhancement
2224 Description: add TCP zero-copy support to kernel
2225 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2226 sends if the hardware supports scatter-gather and checksumming.
2227 This allows socklnd to do client-write and server-read more
2228 efficiently and reduce CPU utilization from skbuf copying.
2231 Frequency : only if NFS exporting from client
2233 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2234 Details : If setting attributes on a file created under NFS that had
2235 never been opened it would be possible to oops the client
2236 if the file had no objects.
2239 Frequency : always for liblustre
2241 Description: liblustre client does MDS+OSTs setattr RPC for each write
2242 Details : When doing a write from a liblustre client, the client
2243 incorrectly issued an RPC to the MDS and each OST the file was
2244 striped over in order to update the timestamps. When writing
2245 with small chunks and many clients this could overwhelm the MDS
2246 with RPCs. In all cases it would slow down the write because
2247 these RPCs are unnecessary.
2249 Severity : enhancement
2251 Description: allow number of MDS service threads to be changed at module load
2252 Details : It is now possible to change the number of MDS service threads
2253 running. Adding "options mds mds_num_threads={N}" to the MDS's
2254 /etc/modprobe.conf will set the number of threads for the next
2255 time Lustre is restarted (assuming the "mds" module is also
2256 reloaded at that time). The default number of threads will
2257 stay the same, 32 for most systems.
2262 Description: OST crash if filesystem is unformatted or corrupt
2263 Details : If an OST is started on a device that has never been formatted
2264 or if the filesystem is corrupt and cannot even mount then the
2265 error handling cleanup routines would dereference a NULL pointer.
2270 Description: NULL pointer deref in llap_from_page.
2271 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2272 incorrect) mapping if the page was truncated/reclaimed while it was
2273 searched for. Check for this condition and skip such pages when
2274 doing readahead. Introduce extra check to llap_from_page() to
2275 verify page->mapping->host is non-NULL (so page is not anonymous).
2278 Frequency : Sometimes when using sys_sendfile
2280 Description: "page not covered by a lock" warnings from ll_readpage
2281 Details : sendfile called ll_readpage without right page locks present.
2282 Now we introduced ll_file_sendfile that does necessary locking
2283 around call to generic_file_sendfile() much like we do in
2287 Frequency : with certain MDS communication failures at client mount time
2289 Description: NULL pointer deref after failed client mount
2290 Details : a client connection request may delayed by the network layer
2291 and not be sent until after the PTLRPC layer has timed out the
2292 request. If the client fails the mount immediately it will try
2293 to clean up before the network times out the request. Add a
2294 reference from the request import to the obd device and delay
2295 the cleanup until the network drops the request.
2298 Frequency : occasionally during client (re)connect
2300 Description: assertion failure during client (re)connect
2301 Details : processing a client connection request may be delayed by the
2302 client or server longer than the client connect timeout. This
2303 causes the client to resend the connection request. If the
2304 original connection request is replied in this interval, the
2305 client may trip an assertion failure in ptlrpc_connect_interpret()
2306 which thought it would be the only running connect process.
2309 Frequency : only with obd_echo servers and clients that are rebooted
2311 Description: kernel BUG accessing uninitialized data structure
2312 Details : When running an obd_echo server it did not start the ping_evictor
2313 thread, and when a client was evicted an uninitialized data
2314 structure was accessed. Start the ping_evictor in the RPC
2315 service startup instead of the OBD startup.
2317 Severity : enhancement
2318 Bugzilla : 10193 (patchless)
2319 Description: Remove dependency on various unexported kernel interfaces.
2320 Details : No longer need reparent_to_init, exit_mm, exit_files,
2321 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2324 Frequency : rare (only users of deprecated and unsupported LDAP config)
2326 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2327 Details : LDAP apparently contains 'lustreName' attributes instead of
2328 'name'. A simple remapping of the name is sufficient.
2331 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2333 Description: waiting_locks_callback trips kernel BUG if client is evicted
2334 Details : Running with the dump_on_timeout debug flag turned on makes
2335 it possible that the waiting_locks_callback() can try to dump
2336 the Lustre kernel debug logs from an interrupt handler. Defer
2337 this log dumping to the expired_lock_main() thread.
2339 Severity : enhancement
2341 Description: Support NFS exporting on 2.6 kernels.
2342 Details : Implement non-rawops metadata methods for NFS server to use without
2343 changing NFS server code.
2346 Frequency : very rare (synthetic metadata workload only)
2348 Description: two racing renames might cause an MDS thread to deadlock
2349 Details : Running the "racer" program may cause one MDS thread to rename
2350 a file from being the source of a rename to being the target of
2351 a rename at exactly the same time that another thread is doing
2352 so, and the second thread has already enqueued these locks after
2353 doing a lookup of the target and is trying to relock them in
2354 order. Ensure that we don't try to re-lock the same resource.
2357 Frequency : only very large systems with liblustre clients
2359 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2360 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2361 all OSTs in parallel. This allows the eviction of stale liblustre
2362 clients to proceed much faster than if they were done in series,
2363 and also offers similar improvements for other set_info RPCs.
2368 Description: excessive CPU usage during initial read phase on client
2369 Details : During the initial read phase on a client, it would agressively
2370 retry readahead on the file, consuming too much CPU and impacting
2371 performance (since 1.4.5.8). Improve the readahead algorithm
2372 to avoid this, and also improve some other common cases (read
2373 of small files in particular, where "small" is files smaller than
2374 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2379 Description: MDS crash when receiving packet with unknown intent.
2380 Details : Do not LBUG in unknown intent case, just return -EFAULT
2382 Severity : enhancement
2383 Bugzilla : 9293, 9385
2384 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2385 Details : Do not serialize getattr (non-intent version) and statfs.
2388 Frequency : occasional, when OST network is overloaded/intermittent
2390 Description: client evicted by OST after bulk IO timeout
2391 Details : If a client sends a bulk IO request (read or write) the OST
2392 may evict the client if it is unresposive to its data GET/PUT
2393 request. This is incorrect if the network is overloaded (takes
2394 too long to transfer the RPC data) or dropped the OST GET/PUT
2395 request. There is no need to evict the client at all, since
2396 the pinger and/or lock callbacks will handle this, and the
2397 client can restart the bulk request.
2400 Frequency : Always when mmapping file with no objects
2402 Description: client crashes when mmapping file with no objects
2403 Details : Check that we actually have objects in a file before doing any
2404 operations on objects in ll_vm_open, ll_vm_close and
2410 Description: Request leak when working with deleted CWD
2411 Details : Introduce advanced request refcount tracking for requests
2412 referenced from lustre intent.
2414 Severity : Enhancement
2416 Description: Cache open file handles on client.
2417 Details : MDS now will return special lock along with openhandle, if
2418 requested and client is allowed to hold openhandle, even if unused,
2419 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2420 closing files for every read/write openration.
2422 Severity : Enhancement
2424 Description: Cache open negative dentries on client when possible.
2425 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2426 negative dentries on lock revocation.
2431 Description: Remounting a client read-only wasn't possible with a zconf mount
2432 Details : It wasn't possible to remount a client read-only with llmount.
2434 Severity : enhancement
2435 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2436 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2437 MPICH is included to add Lustre-specific ADIO interfaces.
2438 This is based closely on the UFS ADIO layer and only differs
2439 in file creation, in order to allow the OST striping to be set.
2440 This is user-contributed code and not supported by CFS.
2445 Description: extended inode attributes (immutable, append-only) work improperly
2446 when 2.4 and 2.6 kernels are used on client/server or vice versa
2447 Details : Introduce kernel-independent values for these flags.
2449 Severity : enhancement
2452 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2453 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2454 granularity. Now, Fractional MB granularity is be supported,
2455 this is very useful for low memory system.
2457 Severity : enhancement
2459 Description: Getattr by fid
2460 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2461 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2462 have locks and avoids getattr rpc requests for it on every stat.
2465 Frequency : Always, for filesystems larger than 2TB
2467 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2468 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2469 filesystems larger than 2TB to be mounted with mballoc enabled.
2472 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2474 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2475 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2476 filesystems larger than 2TB is mounted then the kernel will
2477 silently corrupt the start of the filesystem. CONFIG_LBD is
2478 enabled for all CFS-supported kernels, but the possibility of
2479 this happening with a modified kernel config exists.
2481 Severity : enhancement
2483 Description: add client O_DIRECT support for 2.6 kernels
2484 Details : It is now possible to do O_DIRECT reads and writes to files
2485 in the Lustre client mountpoint on 2.6 kernel clients.
2487 Severity : enhancement
2489 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2490 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2491 parallel, not waiting for response from every OST before sending
2492 a rpc to the next OST.
2497 Description: setattr vs write race when updating file timestamps
2498 Details : Client processes that update a file timestamp into the past
2499 right after writing to the file (e.g. tar) it is possible that
2500 the updated file modification time can be reset to the current
2501 time due to a race between processing the setattr and write RPC.
2503 Severity : enhancement
2505 Description: Bring 'lfs find' closer in line with regular Linux find.
2506 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2507 -print0 options and obtains all the needed info through the lustre
2510 Severity : enhancement
2512 Description: support up to 1024 configured devices on one node
2513 Details : change obd_dev array from statically allocated to dynamically
2514 allocated structs as they are first used to reduce memory usage
2519 Description: Flush dirty partially truncated pages during truncate
2520 Details : Immediatelly flush partially truncated pages in filter_setattr,
2521 this way we completely avoid having any pages in page cache on OST
2522 and can retire ugly workarounds during writes to flush such pages.
2527 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2528 Details : protect inode from truncation within vfs_unlink() context
2529 just take a reference before calling vfs_unlink() and release it
2530 when parent's i_sem is free.
2535 Description: last_id value checked outside lock on OST caused LASSERT failure
2536 Details : If there were multiple MDS->OST object precreate requests in
2537 flight, it was possible that the OST's last object id was checked
2538 outside a lock and incorrectly tripped an assertion. Move checks
2539 inside locks, and discard old precreate requests.
2542 Frequency : always, if extents are used on OSTs
2544 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2545 Details : OSTs using the extents format would not zero the high 16 bits of
2546 the index physical block number. This is not a problem for any
2547 OST filesystems smaller than 16TB, and no kernels support ext3
2548 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2549 new/modified files) and can be fixed for existing filesystems
2550 with e2fsprogs-1.39-cfs1.
2555 Description: import connection selection may be incorrect if timer wraps
2556 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2557 import connections to be ignored if the 32-bit jiffies counter
2558 wraps. Use a 64-bit jiffies counter.
2561 Frequency : very large clusters immediately after boot
2563 Description: LNET request buffers exhausted under heavy short-term load
2564 Details : If a large number of client requests are generated on a service
2565 that has previously never seen so many requests it is possible
2566 that the request buffer growth cannot keep up with the spike in
2567 demand. Instead of dropping incoming requests, they are held in
2568 the LND until the RPC service can accept more requests.
2571 Frequency : Sometimes during replay
2573 Description: Assertion failure in ll_local_open after replay.
2574 Details : If replay happened on an open request reply before we were able
2575 to set replay handler, reply will become not swabbed tripping the
2576 assertion in ll_local_open. Now we set the handler right after
2577 recognising of open request
2580 Frequency : very rare
2582 Description: kernel reports "badness in vsnprintf"
2583 Details : Reading from the "recovery_status" /proc file in small chunks
2584 may cause a negative length in lprocfs_obd_rd_recovery_status()
2585 call to vsnprintf() (which is otherwise harmless). Exit early
2586 if there is no more space in the output buffer.
2588 Severity : enhancement
2590 Description: clear OBD RPC statistics by writing to them
2591 Details : It is now possible to clear the OBD RPC statistics by writing
2592 to the "stats" file.
2597 Description: Client mtime is not the same on different clients after utimes
2598 Details : In some cases, the client was using the utimes() syscall on
2599 a file cached on another node. The clients now validate the
2600 ctime from the MDS + OSTs to determine which one is right.
2605 Description: Inability to activate failout mode
2606 Details : lconf script incorrectly assumed that in pythong string's numeric
2607 value is used in comparisons.
2610 Frequency : always with multiple stripes per file
2612 Description: Inefficient object allocation for mutli-stripe files
2613 Details : When selecting which OSTs to stripe files over, for files with
2614 a stripe count that divides evenly into the number of OSTs,
2615 the MDS is always picking the same starting OST for each file.
2616 Return the OST selection heuristic to the original design.
2621 Description: mount failures may take full timeout to return an error
2622 Details : Under some heavy load conditions it is possible that a
2623 failed mount can wait for the full obd_timeout interval,
2624 possibly several minutes, before reporting an error.
2625 Instead return an error as soon as the status is known.
2627 Frequency : quota enabled and large files being deleted
2629 Description: releasing more than 4GB of quota at once hangs OST
2630 Details : If a user deletes more than 4GB of files on a single OST it
2631 will cause the OST to spin in an infinite loop. Release
2632 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2637 Description: statfs data retrieved from /proc may be stale or zero
2638 Details : When reading per-device statfs data from /proc, in the
2639 {kbytes,files}_{total,free,avail} files, it may appear
2640 as zero or be out of date.
2643 Frequency : systems with MD RAID1 external journal devices
2645 Description: lconf's call to blkid is confused by RAID1 journal devices
2646 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2647 of returning all block devices that match the journal UUID.
2650 Frequency : always, for aggregate stripe size over 4GB
2652 Description: assertion fails when trying to use 4GB stripe size
2653 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2654 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2657 Frequency : always on ppc64
2659 Description: the first write on an ext3 filesystem with mballoc got stuck
2660 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2661 perform endianness conversion.
2663 ------------------------------------------------------------------------------
2665 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2667 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2668 INTEROPERATE with older versions automatically. Please read the
2669 user documentation before upgrading any part of a live system.
2670 * WARNING: Lustre networking configuration changes are required with
2671 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2674 * Support for kernels:
2675 2.6.9-22.0.2.EL (RHEL 4)
2676 2.6.5-7.244 (SLES 9)
2677 2.6.12.6 vanilla (kernel.org)
2680 Severity : enhancement
2681 Bugzilla : 7981/8208
2682 Description: Introduced Lustre Networking (LNET)
2683 Details : LNET is new networking infrastructure for Lustre, it includes
2684 a reorganized network configuration mode (see the user
2685 documentation for full details) as well as support for routing
2686 between different network fabrics. Lustre Networking Devices
2687 (LNDs) for the supported network fabrics have also been
2688 created for this new infrastructure.
2690 Severity : enhancement
2691 Description: Introduced Access control lists
2692 Details : clients can set ACLs on files and directories in order to have
2693 more fine-grained permissions than the standard Unix UGO+RWX.
2694 The MDS must be started with the "-o acl" mount option.
2696 Severity : enhancement
2697 Description: Introduced filesystem quotas
2698 Details : Administrators may now establish per-user quotas on the
2701 Severity : enhancement
2703 Description: Configuration change for the XT3
2704 The PTLLND is now used to run Lustre over Portals on the XT3
2705 The configure option(s) --with-cray-portals are no longer used.
2706 Rather --with-portals=<path-to-portals-includes> is used to
2707 enable building on the XT3. In addition to enable XT3 specific
2708 features the option --enable-cray-xt3 must be used.
2713 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2714 Details : A race between allocating a new llcd and re-getting the llcd_lock
2715 allowed another thread to grab newly-allocated llcd.
2717 Severity : enhancement
2719 Description: 2.6 OST async journal commit and locking fix to improve performance
2720 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2721 2.6 kernels are now async as they already were in 2.4 kernels so
2722 that they can commit concurrently with the network bulk transfer.
2723 For block-allocated files the filter allocation semaphore is held
2724 to avoid filesystem fragmentation during allocation. BKL lock
2725 removed for 2.6 xattr operations where it is no longer needed.
2730 Description: lconf incorrectly determined whether two IP networks could talk
2731 Details : In some more complicated routing and multiple-network
2732 configurations, lconf will avoid trying to make a network
2733 connection to a disjoint part of the IP space. It was doing the
2734 math incorrectly for one set of cases.
2739 Description: Fix for potential infinite loop processing records in an llog.
2740 Details : If an llog record is corrupted/zeroed, it is possible to loop
2741 forever in llog_process(). Validate the llog record length
2742 and skip the remainder of the block on error.
2745 Frequency : occasional (liblustre only)
2747 Description: liblustre could not open files whose last component is a symlink
2748 Details : sysio_path_walk() would incorrectly pass the open intent to
2749 intermediate path components.
2752 Frequency : rare (liblustre only with non-standard tuning)
2753 Bugzilla : 7201 (7350)
2754 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2755 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2756 after it has been released, only do this for referenced locks
2758 Severity : enhancement
2760 Description: specifying an (invalid) directory default stripe_size of -1
2761 would reset the directory default striping
2762 Details : stripe_size -1 was used internally to signal directory stripe
2763 removal, now use "all default" to signal dir stripe removal
2764 as a directory striping of "all default" is not useful
2767 Frequency : common for large clusters running liblustre clients
2769 Description: doing an ls when liblustre clients are running is slow
2770 Details : sending a glimpse AST to a liblustre client waits for every AST
2771 to time out, as liblustre clients will not respond. Since they
2772 cannot cache data we refresh the OST lock LVB from disk instead.
2774 Severity : enhancement
2776 Description: doing an ls at the same time as file IO can be slow
2777 Details : enqueue and other "small" requests can be blocked behind many
2778 large IO requests. Create a new OST IO portal for non-IO
2779 requests so they can be processed faster.
2782 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2784 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2785 created file with mode 000
2786 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2787 allow an ftruncate() as a non-root user to a file with mode 000.
2788 Lustre now respects this flag to disable mode checks when
2789 truncating a file owned by the user
2792 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2795 Description: Revoking locks from clients that went dead or catatonic might take
2797 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2798 cancellation of such locks instant on servers without waiting for
2799 any reply from clients. Clients drops these locks when cancel
2800 notification from server is received without replying.
2803 Frequency : liblustre-only, when liblustre client dies or becomes busy
2805 Description: Doing ls on Linux clients can take a long time with active
2807 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2808 granting such locks to it in the first place if possible. Locks
2809 are taken by proxy on the OST during the read or write and
2810 dropped immediately afterward. Add connect flags handling, do
2811 not grant locks to liblustre clients for glimpse ASTs.
2813 Severity : enhancement
2815 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2816 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2817 sequential reads and behaves badly with applications doing random
2818 reads. Improve it by reading ahead at least read region, and
2819 avoiding excessive large RPC for small reads.
2821 Severity : enhancement
2823 Description: Creating more than 1000 files for a single job may cause a load
2824 imbalance on the OSTs if there are also a large number of OSTs.
2825 Details : qos_prep_create() uses an OST index reseed value that is an
2826 even multiple of the number of available OSTs so that if the
2827 reseed happens in the middle of the object allocation it will
2828 still utilize the OSTs as uniformly as possible.
2833 Description: OST or MDS may oops in ping_evictor_main()
2834 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2835 but doesn't restart at beginning of obd_exports_timed list
2838 Severity : enhancement
2840 Description: improve by-nid export eviction on the MDS and OST
2841 Details : allow multiple exports with the same NID to be evicted at one
2842 time without re-searching the exports list.
2845 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2847 Description: MDS may oops in groups_free()
2848 Details : in rare race conditions a newly allocated group_info struct is
2849 freed again, and this can be NULL. The 2.4 compatibility code
2850 for groups_free() checked for a NULL pointer, but 2.6 did not.
2853 Frequency : common for liblustre clients doing little filesystem IO
2854 Bugzilla : 9352, 7313
2855 Description: server may evict liblustre clients accessing contended locks
2856 Details : if a client is granted a lock or receives a completion AST
2857 with a blocking AST already set it would not reply to the AST
2858 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2861 Frequency : lfs setstripe, only systems with more than 160 OSTs
2863 Description: unable to set striping with a starting offset beyond OST 160
2864 Details : llapi_create_file() incorrectly limited the starting stripe
2865 index to the maximum single-file stripe count.
2868 Frequency : LDAP users only
2870 Description: lconf did not handle in-kernel recovery with LDAP properly
2871 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2873 Severity : enhancement
2875 Description: bind OST threads to NUMA nodes to improve performance
2876 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2877 node and do their allocations there to localize memory access
2879 Severity : enhancement
2881 Description: llmount can determine client NID directly from Myrinet (GM)
2882 Details : the client NID code from gmnalnid was moved directly into
2883 llmount, removing the need to use this or specifying the
2884 client NID explicitly when mounting GM clients with zeroconf
2887 Frequency : if client is started with down MDS
2889 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2890 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2891 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2892 ptlrpc_import_delayed_req() because we didn't check if the
2893 request was interrupted, and we also didn't break out of the
2894 event loop if there was no timeout
2899 Description: data loss during non-page-aligned writes to a single file from
2900 both multiple nodes and multiple threads on one node at same time
2901 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2902 inconsistency led to false short-reads, that were cached and later
2903 used by ->prepare_write() to fill in partially written page,
2904 leading to data loss.
2907 Frequency : always, if lconf --abort_recovery used
2909 Description: lconf --abort_recovery fails with 'Operation not supported'
2910 Details : lconf was attempting to abort recovery on the MDT device and not
2913 Severity : enhancement
2915 Description: remove cleanup logs
2916 Details : replace lconf-generated cleanup logs with lustre internal
2917 cleanup routines. Eliminates the need for client-cleanup and
2920 Severity : enhancement
2922 Description: add support for EAs (user and system) on lustre filesystems
2923 Details : it is now possible to store extended attributes in the Lustre
2924 client filesystem, and with the user_xattr mount option it
2925 is possible to allow users to store EAs on their files also
2927 Severity : enhancement
2929 Description: Add possibility (config option) to show minimal available OST free
2931 Details : When compiled with --enable-mindf configure option, statfs(2)
2932 (and so, df) will return least minimal free space available from
2933 all OSTs as amount of free space on FS, instead of summary of
2934 free spaces of all OSTs.
2936 Severity : enhancement
2938 Description: do not expand extent locks acquired on OST-side
2939 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2940 by server: they are not cached anyway.
2943 Frequency : when mmap is used/binaries executed from Lustre
2945 Description: Unmmap pages before throwing them away from read cache.
2946 Details : llap_shrink cache now attempts to unmap pages before discarding
2947 them (if unmapping failed - do not discard). SLES9 kernel has
2948 extra checks that trigger if this unmapping is not done first.
2953 Description: lconf didn't resolve symlinks before checking to see whether a
2954 given mountpoint was already in use
2957 Frequency : when migrating failover services
2958 Bugzilla : 6395, 9514
2959 Description: When migrating a subset of services from a node (e.g. failback
2960 from a failover service node) the remaining services would
2961 time out and evict clients.
2962 Details : lconf --force (implied by --failover) sets the global obd_timeout
2963 to 5 seconds in order to quickly disconnect, but this caused
2964 other RPCs to time out too quickly. Do not change the global
2965 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2967 Severity : enhancement
2968 Frequency : if MDS is started with down OST
2969 Bugzilla : 9439,5706
2970 Description: Allow startup/shutdown of an MDS without depending on the
2971 availability of the OSTs.
2972 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2973 Add appropriate locking and lov-osc refcounts for safe
2974 cleaning. Add osc abort_inflight calls in case the
2975 synchronize never started.
2978 Frequency : occasional (Cray XT3 only)
2980 Description: root not authorized to access files in CRAY_PORTALS environment
2981 Details : The client process capabilities were not honoured on the MDS in
2982 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2983 been accessed by an authorized user then root was able to access
2984 the file on the local client also. The root user capabilities
2985 are now allowed on the MDS, as this environment has secure UID.
2988 Frequency : occasional
2990 Description: ldiskfs "too long searching" message happens too often
2991 Details : A debugging message (otherwise harmless) prints too often on
2992 the OST console. This has been reduced to only happen when
2993 there are fragmentation problems on the filesystem.
2998 Description: Division by zero in statfs when all OSCs are inactive
2999 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3000 lov_statfs divides by value returned by lov_get_stripecnt().
3004 Bugzilla : 9489, 3273
3005 Description: First write from each client to each OST was only 4kB in size,
3006 to initialize client writeback cache, which caused sub-optimal
3007 RPCs and poor layout on disk for the first writen file.
3008 Details : Clients now request an initial cache grant at (re)connect time
3009 and so that they can start streaming writes to the cache right
3010 away and always do full-sized RPCs if there is enough data.
3011 If the OST is rebooted the client also re-establishes its grant
3012 so that client cached writes will be honoured under the grant.
3017 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3018 Details : Now I/O RPCs go to different portal number and (presumably) fast
3019 lock requests (and glimses) and other RPCs get their own service
3020 threads pool that should be able to service those RPCs
3023 Severity : enhancement
3025 Description: Ability to exchange lustre version between client and servers and
3026 issue warnings at client side if client is too old. Also for
3027 liblustre clients there is ability to refuse connection of too old
3029 Details : New 'version' field is added to connect data structure that is
3030 filled with version info. That info is later checked by server and
3034 Frequency : rare, liblustre only.
3035 Bugzilla : 9296, 9581
3036 Description: Two simultaneous writes from liblustre at offset within same page
3037 might proceed at the same time overwriting eachother with stale
3039 Details : I/O lock withing llu_file_prwv was released too early, before data
3040 actually was hitting the wire. Extended lock-holding time until
3041 server acknowledges receiving data.
3044 Frequency : extremely rare. Never observed in practice.
3046 Description: avoid generating lustre_handle cookie of 0.
3047 Details : class_handle_hash() generates handle cookies by incrementing
3048 global counter, and can hit 0 occasionaly (this is unlikely, but
3049 not impossible, because initial value of cookie counter is
3050 selected randonly). Value of 0 is used as a sentinel meaning
3051 "unassigned handle" --- avoid it. Also coalesce two critical
3052 sections in this function into one.
3054 Severity : enhancement
3056 Description: allow liblustre clients to delegate truncate locking to OST
3057 Details : To avoid overhead of locking, liblustre client instructs OST to
3058 take extent lock in ost_punch() on client's behalf. New connection
3059 flag is added to handle backward compatibility.
3061 Severity : enhancement
3062 Bugzilla : 4928, 7341, 9758
3063 Description: allow number of OST service threads to be specified
3064 Details : a module parameter allows the number of OST service threads
3065 to be specified via "options ost ost_num_threads={N}" in the
3066 OSS's /etc/modules.conf or /etc/modprobe.conf.
3070 Bugzilla : 6146, 9635, 9895
3071 Description: servers crash with bad pointer in target_handle_connect()
3072 Details : In rare cases when a client is reconnecting it was possible that
3073 the connection request was the last reference for that export.
3074 We would temporarily drop the export reference and get a new
3075 one, but this may have been the last reference and the export
3076 was just destroyed. Get new reference before dropping old one.
3078 Severity : enhancement
3079 Frequency : if client is started with failover MDS
3081 Description: Allow multiple MDS hostnames in the mount command
3082 Details : Try to read the configuration from all specified MDS
3083 hostnames during a client mount in case the "primary"
3086 Severity : enhancement
3088 Description: Stop sending data to evicted clients as soon as possible.
3089 Details : Check if the client we are about to send or are sending data to
3090 was evicted already. (Check is done every second of waiting,
3091 for which l_wait_event interface was extended to allow checking
3092 of exit condition at specified intervals).
3095 Frequency : rare, normally only when NFS exporting is done from client
3097 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3099 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3102 Frequency : always, if srand() is called before liblustre initialization
3104 Description: Liblustre uses system PRNG disturbing its usage by user application
3105 Details : Introduce internal to lustre fast and high-quality PRNG for
3106 lustre usage and make liblustre and some other places in generic
3107 lustre code to use it.
3109 Severity : enhancement
3110 Bugzilla : 9477, 9557, 9870
3111 Description: Verify that the MDS configuration logs are updated when xml is
3112 Details : Check if the .xml configuration logs are newer than the config
3113 logs stored on the MDS and report an error if this is the case.
3114 Request --write-conf, or allow starting with --old_conf.
3116 Severity : enhancement
3118 Description: Handle symlinks in the path when checking if Lustre is mounted.
3119 Details : Resolve intermediate symlinks when checking if a client has
3120 mounted a filesystem to avoid duplicate client mounts.
3125 Description: lconf can hit an error exception but still return success.
3126 Details : The lconf command catches the Command error exception at the top
3127 level script context and will exit with the associated exit
3128 status, but doesn't ensure that this exit status is non-zero.
3133 Description: failure of ptlrpc thread startup can cause oops
3134 Details : Starting a ptlrpc service thread can fail if there are a large
3135 number of threads or the server memory is very fragmented.
3136 Handle this without oopsing.
3139 Frequency : always, only if liblustre and non-default acceptor port was used
3141 Description: liblustre cannot connect to servers with non-default acceptor port
3142 Details : tcpnal_set_default_params() was not called and was therefore
3143 ignoring the environment varaible TCPNAL_PORT, as well as other
3144 TCPNAL_ environment variables
3149 Description: two objects could be created on the same OST for a single file
3150 Details : If an OST is down, in some cases it was possible to create two
3151 objects on a single OST for a single file. No problems other
3152 than potential performance impact and spurious error messages.
3156 Bugzilla : 5681, 9562
3157 Description: Client may oops in ll_unhash_aliases
3158 Details : Client dcache may become inconsistent in race condition.
3159 In some cases "getcwd" can fail if the current directory is
3165 Description: Inode refcounting problems in NFS export code
3166 Details : link_raw functions used to call d_instantiate without obtaining
3167 extra inode reference first.
3171 Bugzilla : 9942, 9903
3172 Description: Referencing freed requests leading to crash, memleaks with NFS.
3173 Details : We used to require that call to ll_revalidate_it was always
3174 followed by ll_lookup_it. Also with revalidate_special() it is
3175 possible to call ll_revalidate_it() twice for the same dentry
3176 even if first occurence returned success. This fix changes semantic
3177 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3178 reference on a request referred from the intent.
3179 ll_intent_release() then releases such a request.
3182 Frequency : rare, normally benchmark loads only
3184 Description: unlinked inodes were kept in memory on the client
3185 Details : If a client is repeatedly creating and unlinking files it
3186 can accumulate a lot of stale inodes in the inode slab cache.
3187 If there is no other client load running this can cause the
3188 client node to run out of memory. Instead flush old inodes
3189 from client cache that have the same inode number as a new inode.
3192 Frequency : SLES9 2.6.5 kernel and long filenames only
3193 Bugzilla : 9969, 10379
3194 Description: utime reports stale NFS file handle
3195 Details : SLES9 uses out-of-dentry names in some cases, which confused
3196 the lustre dentry revalidation. Change it to always use the
3200 Frequency : rare, unless heavy write-truncate concurrency is continuous
3201 Bugzilla : 4180, 6984, 7171, 9963, 9331
3202 Description: OST becomes very slow and/or deadlocked during object unlink
3203 Details : filter_destroy() was holding onto the parent directory lock
3204 while truncating+unlinking objects. For very large objects this
3205 may block other threads for a long time and slow overall OST
3206 responsiveness. It may also be possible to get a lock ordering
3207 deadlock in this case, or run out of journal credits because of
3208 the combined truncate+unlink. Solution is to do object truncate
3209 first in one transaction without parent lock, and then do the
3210 final unlink in a new transaction with the parent lock. This
3211 reduces the lock hold time dramatically.
3214 Frequency : rare, 2.4 kernels only
3216 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3217 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3218 before unmounting their filesystem in order to avoid deadlock.
3219 The kernel_locked() function in 2.4 kernels only checks whether
3220 the kernel lock is held, not whether it is this process that is
3221 holding it as 2.6 kernels do.
3226 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3227 Details : The client ptlrpc code may be trying to reconnect to a down
3228 server before a previous connection attempt has timed out.
3229 Increase the reconnect interval to be longer than the connection
3230 timeout interval to avoid sending duplicate connections to
3234 Frequency : echo_client brw_test command
3236 Description: fix echo_client to work with OST preallocated code
3237 Details : OST preallocation code (5137) didn't take echo_client IO path
3238 into account: echo_client calls filter methods outside of any
3239 OST thread and, hence, there is no per-thread preallocated
3240 pages and buffers to use. Solution: hijack pga pages for IO. As
3241 a byproduct, this avoids unnecessary data copying.
3245 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3246 Description: Client can oops in mdc_commit_close() after open replay
3247 Details : It was possible for the MDS to return an open request with no
3248 transaction number in mds_finish_transno() if the client was
3249 evicted, but without actually returning an error. Clients
3250 would later try to replay that open and may trip an assertion
3251 Simplify the client close codepath, and always return an error
3252 from the MDS in case the open is not successful.
3255 Frequency : rare, 2.6 OSTs only
3257 Description: OST may deadlock under high load on fragmented files
3258 Details : If there was a heavy load and highly-fragmented OST filesystems
3259 it was possible to have all the OST threads deadlock waiting on
3260 allocation of biovecs, because the biovecs were not released
3261 until the entire RPC IO was completed. Instead, release biovecs
3262 as soon as they are complete to ensure forward IO progress.
3264 Severity : enhancement
3266 Description: Support for specifying external journal device at mount
3267 Details : If an OST or MDS device is formatted with an external journal
3268 device, this device major/minor is stored in the ext3 superblock
3269 and may not be valid for failover. Allow detecting and
3270 specifying the external journal at mount time.
3275 Description: Mounting an MDS with pending unlinked files may cause oops
3276 Details : target_finish_recovery() calls mds_postrecov() which returned
3277 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3278 considers this an error and immediately begins cleaning up the
3279 lov, just after starting the mds_lov process
3281 Severity : enhancement
3283 Description: Implement 'lfs df' to report actual free space on per-OST basis
3284 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3285 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3286 usage of objects; '-h' to report in human readable format.
3288 ------------------------------------------------------------------------------
3290 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3297 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3298 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3299 at startup during orphan unlinking. Instead, skip buddy update
3300 before setup, as it will be regenerated after recovery is complete.
3305 Description: If an OST is inactive, its locks might reference stale inodes.
3306 Details : lov_change_cbdata() must iterate over all namespaces, even if
3307 they are inactive to clear inode references from the lock.
3309 Severity : enhancement
3310 Frequency : occasional, if non-standard max_dirty_mb used
3312 Description: Client will block write RPCs if not enough grant
3313 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3314 then the client will block writes while waiting for another RPC
3315 to complete instead of consuming its dirty limit. With change
3316 we get improved performance when max_dirty_mb is small.
3318 Severity : enhancement
3319 Bugzilla : 3389, 6253
3320 Description: Add support for supplementary groups on the MDS.
3321 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3322 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3323 lookups for user supplementary groups into a cache.
3327 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3328 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3329 for liblustre O_EXCL is enforced by MDS.
3333 Description: readdir never returns NULL in liblustre.
3334 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3335 dentry in struct dirent.
3339 Frequency : liblustre only, depends on application IO pattern
3340 Description: liblustre clients evicted if not contacting servers
3341 Details : Don't put liblustre clients into the ping_evictor list, so
3342 they will not be evicted by the pinger ever.
3344 Severity : enhancement
3346 Description: Add ability to evict clients by NID from MDS.
3347 Details : By echoing "nid:$NID" string into
3348 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3349 $NID would be instantly evicted from this MDS and from all active
3350 OSTs connected to it.
3354 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3355 Details : lookup_it_finish() used to query file size from OSTs that was not
3360 Description: service threads change working directory to that of init
3361 Details : Starting lustre service threads may pin the working directory
3362 of the parent thread, making that filesystem busy. Threads
3363 now change to the working directory of init to avoid this.
3367 Frequency : during shutdown only
3368 Description: shutdown with a failed MDS or OST can cause unmount to hang
3369 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3374 Frequency : 2.6 only
3375 Description: chmod/chown may include an extra supplementary group
3376 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3377 supplementary group and if none is specified this is used.
3380 Bugzilla : 5479 (6816)
3382 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3383 Details : If lookup is in progress on a file that is unlinked we might try
3384 to revalidate the inode and fail in revalidate after lookup is
3385 complete and ll_file_open() enqueues the open again but
3386 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3389 Frequency : always, if lconf --abort_recovery used
3391 Description: lconf --abort_recovery fails with 'Operation not supported'
3392 Details : lconf was attempting to abort recovery on the MDT device and not
3395 ------------------------------------------------------------------------------
3397 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3402 Frequency : rare (only unsupported configurations with a node running as an
3404 Bugzilla : 6514, 5137
3405 Description: Mounting a Lustre file system on a node running as an OST could
3407 Details : OSTs now preallocates memory needed to write out data at
3408 startup, instead of when needed, to avoid having to
3409 allocate memory in possibly low memory situations.
3410 Specifically, if the file system is mounted on on OST,
3411 memory pressure could force it to try to write out data,
3412 which it needed to allocate memory to do. Due to the low
3413 memory, it would be unable to do so and the node would
3414 become unresponsive.
3416 Severity : enhancement
3418 Description: Addition of lconf --service command line option
3419 Details : lconf now accepts a '--service <arg>' option, which is
3420 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3422 Severity : enhancement
3424 Description: Failover mode is now the default for OSTs.
3425 Details : By default, OSTs will now run in failover mode. To return to
3426 the old behaviour, add '--failout' to the lmc line for OSTs.
3428 Severity : enhancement
3430 Description: Health checks are now provided for MDS and OSTs
3431 Details : Additional detailed health check information on MSD and OSTs
3432 is now provided through the procfs health_check value.
3435 Frequency : occasional, depends on IO load
3437 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3438 after numerous create/delete cycles
3439 Details : The ext3 inode allocation policy would not allocate new inodes
3440 very well on the OSTs because there are no new directories
3441 being created. Instead we look for groups with free space if
3442 the parent directories are nearly full.
3447 Description: Network or server problems during mount may cause partially
3448 mounted clients instead of returning an error.
3449 Details : The config llog parsing code may overwrite the error return
3450 code during mount error handling, returning success instead
3456 Description: MDS can fail to allocate large reply buffers
3457 Details : After long uptimes the MDS can fail to allocate large reply
3458 buffers (e.g. zconf client mount config records) due to memory
3459 fragmentation or consumption by the buffer cache. Preallocate
3460 some large reply buffers so that these replies can be sent even
3461 under memory pressure.
3465 Frequency : rare (liblustre)
3466 Description: fsx running with liblustre complained that using truncate() to
3467 extend the file doesn't work. This patch corrects that issue.
3468 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3469 ATTR_SIZE and lsm use in llu_setattr_raw.
3473 Frequency : rare, only 2.6 kernels
3474 Description: Unusual file access patterns on the MDS may result in inode
3475 data being lost in very rare circumstances.
3476 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3477 code in the 2.6 kernel caused the inode and/or EA data not to
3478 be read from disk, causing single-file corruption.
3482 Frequency : rare, only 2.6 filesystems using extents
3483 Description: Heavy concurrent write and delete load may cause data corruption.
3484 Details : It was possible under high-load situations to have an extent
3485 metadata block in the block device cache from a just-unlinked
3486 file overwrite a newly-allocated data block. We now unmap any
3487 metadata buffers that alias just-allocated data blocks.
3491 Frequency : filesystems with default stripe_count larger than 77
3492 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3493 Details : lconf specifies an inode size of 4096 bytes when the default
3494 stripe_count is larger than 77. This conflicts with the default
3495 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3496 this case to avoid pinning too much memory for large EAs.
3498 ------------------------------------------------------------------------------
3500 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3505 Frequency : rare (extremely heavy IO load with hundreds of clients)
3507 Description: Client is evicted, gets IO error writing to file
3508 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3509 caused clients to be evicted for AST timeouts. The fixes in
3510 bug 5192 mean we no longer need to have such short AST timeouts
3511 so ldlm_timeout has been increased.
3514 Frequency : occasional during --force or --failover shutdown under load
3515 Bugzilla : 5949, 4834
3516 Description: Server oops/LBUG if stopped with --force or --failover under load
3517 Details : a collection of import/export refcount and cleanup ordering
3518 issues fixed for safer force cleanup
3521 Frequency : only filesystems larger than 120 OSTs
3522 Bugzilla : 5990, 6223
3523 Description: lfs getstripe would oops on a very large filesystem
3524 Details : lov_getconfig used kfree on vmalloc'd memory
3527 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3528 Bugzilla : 6242, 6243
3529 Description: reading from files that had been truncated to a non-zero size
3530 but never opened returned no data
3531 Details : ll_file_read() reads zeros from no-object files to EOF
3536 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3537 Details : There's little harm in aborting MDS/OSS recovery and letting it
3538 try again, so I removed the LASSERT and return an error instead.
3540 Severity : enhancement
3542 Description: New debugging infrastructure for tracking down data corruption
3543 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3544 (b) cover more of the client-side code path, and (c) try to narrow
3545 down where problems occurred
3549 Bugzilla : 3819, 4364, 4397, 6313
3550 Description: Racing close and eviction MDS could cause assertion in mds_close
3551 Details : It was possible to get multiple mfd references during close and
3552 client eviction, leading to one thread referencing a freed mfd.
3554 Severity: : enhancement
3555 Bugzilla : 3262, 6359
3556 Description: Attempts to reconnect to servers are now more aggressive.
3557 Details : This builds on the enhanced upcall-less recovery that was added
3558 in 1.4.2. When trying to reconnect to servers, clients will
3559 now try each server in the failover group every 10 seconds. By
3560 default, clients would previously try one server every 25 seconds.
3565 Description: After recovery, certain operations trigger a failed
3566 assertion on a client.
3567 Details : Failing over an mds, using lconf -d --failover, while a
3568 client was doing a readdir() call would cause the client to
3569 LBUG after recovery completed and the readdir() was resent.
3571 Severity : enhancement
3573 Description: Default groups are now added by lconf
3574 Details : You can now run lconf --group <servicename> without having to
3575 manually add groups with lmc.
3578 Frequency : occasional
3580 Description: Nodes with an elan id of 0 trigger a failed assertion
3583 Frequency : always when accessing e.g. tty/console device nodes
3585 Description: tty and some other devices nodes cannot be used on lustre
3586 Details : file's private_data field is used by device data and lustre
3587 values in there got lost. New field was added to struct file to
3588 store fs-specific private data.
3591 Frequency : when exporting Lustre via NFS
3593 Description: NFSD failed occasionally when looking up a path component
3594 Details : NFSD is looking up ".." which was broken in ext3 directories
3595 that had grown large enough to become hashed.
3598 Frequency : Clusters with multiple interfaces not on the same subnet
3600 Description: Nodes will repeatedly try to reconnect to an interface which it
3601 cannot reach and report an error to the log.
3602 Details : Extra peer list entries will be created by lconf with some peers
3603 unreachable. lconf now validates the peer before adding it.
3606 Frequency : Only if a default stripe is set on the filesystem root.
3608 Description: Setting a default stripe on the filesystem root prevented the
3609 filesystem from being remounted.
3610 Details : The client was sending extra request flags in the root getattr
3611 request and did not allocate a reply buffer for the dir EA.
3614 Frequency : occasional, higher if lots of files are accessed by one client
3615 Bugzilla : 6159, 6097
3616 Description: Client trips assertion regarding lsm mismatch/magic
3617 Details : While revalidating inodes the VFS looks up inodes with ifind()
3618 and in rare cases can find an inode that is being freed.
3619 The ll_test_inode() code will free the lsm during ifind()
3620 when it finds an existing inode and then the VFS later attaches
3621 this free lsm to a new inode.
3625 Bugzilla : 6422, 7030
3626 Description: MDS deadlock between mkdir and client eviction
3627 Details : Creating a new file via mkdir or mknod (starting a transaction
3628 and getting the ns lock) can deadlock with client eviction
3629 (gets ns lock and trying to finish a synchronous transaction).
3632 Frequency : occasional
3633 Description: While starting a server, the fsfilt_ext3 module could not be
3635 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3636 kernels. Previously, lconf would still use the ext3 name
3637 when trying to load modules. Now, it will correctly use
3638 ext3 on 2.4 and ldiskfs on 2.6.
3640 Severity : enhancement
3641 Description: The default stripe count has been changed to 1
3642 Details : The interpretation of the default stripe count (0, to lfs
3643 or lmc) has been changed to mean striping across a single
3644 OST, rather than all available. For general usage we have
3645 found a stripe count of 1 or 2 works best.
3647 Severity : enhancement
3648 Description: Add support for compiling against Cray portals.
3649 Details : Conditional compiling for some areas that are different
3653 Frequency : occasional
3654 Bugzilla : 6409, 6834
3655 Description: Creating files with an explicit stripe count may lead to
3656 a failed assertion on the MDS
3657 Details : If some OSTs are full or unavailable, creating files may
3658 trigger a failed assertion on the MDS. Now, Lustre will
3659 try to use other servers or return an error to the
3663 Frequency : occasional
3665 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3666 caused lock timeout during readahead (since 1.4.2).
3667 Details : Processes doing readahead might match a lock that hasn't been
3668 granted yet if there are overlapping and conflicting lock
3669 requests. The readahead process waits on ungranted lock
3670 (original lock is CBPENDING), while OST waits for that process
3671 to cancel CBPENDING read lock and eventually evicts client.
3673 Severity : enhancement
3675 Description: Initial enabling of flock support for clients
3676 Details : Implements fcntl advisory locking and file status functions.
3677 This feature is provided as an optional mount flag (default
3678 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3679 locking are implemented yet, and those that are are not guaranteed
3680 to be completely correct in production environments.
3681 mount -t lustre -o [flock|noflock] ...
3684 Frequency : occasional
3686 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3687 assertion in the ext3 JBD (journaling) layer.
3688 Details : The b_committed_data struct is protected by the big kernel lock
3689 in 2.4 kernels, serializing journal_commit_transaction() and
3690 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3691 access to this struct. In 2.6 kernels there is finer grained
3692 locking to improve SMP performance of the JBD layer.
3696 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3701 Description: merge and cleanup kernel patches.
3704 -----------------------------------------------------------------------------
3706 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3708 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3709 versions of Lustre. Please update all servers and clients to
3710 version 1.4.2 or later at the same time. You must also run
3711 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3712 to update the configuration logs.
3714 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3715 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3716 - swab llog records of type '0' so we get proper header size/idx (5861)
3717 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3718 - fix rename of one directory over another leaking an inode (5953)
3719 - avoid SetPageDirty on 2.6 (5981)
3720 - don't re-add just-being-destroyed locks to the waiting list (5653)
3721 - when creating new directories, inherit the parent's custom
3722 striping settings if present parent (3048)
3723 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3724 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3725 - don't hold client locks on temporary worklist from l_lru (5666)
3726 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3727 - automatically evict dead clients (5921)
3728 - Update file size properly in create+truncate+fstat case (6196)
3729 - Do not unhash mountpoint dentries, do not allow removal of
3731 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3732 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3733 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3734 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3735 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3736 - Correctly handle reads of files with no objects (6243)
3737 - lctl recover will also mark a device active if deactivate used (5933)
3739 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3740 - allow --write-conf on an MDS with different nettype than client (5619)
3741 - don't write config llogs to MDS for mounts not from that MDS (5617)
3742 - lconf should create multiple TCP connections from a client (5201)
3743 - init scripts are now turned off by default; run chkconfig --on
3744 lustre and chkconfig --on lustrefs to use them
3745 - upcalls are no longer needed for clients to recover to failover
3747 - add --abort-recovery option to lconf to abort recovery on device
3749 - add support for an arbitrary number of OSTs (3026)
3750 - Quota support protocol changes.
3751 - forward compatibility changes to wire structs (6007)
3752 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3753 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3754 - improved "lustre status" script
3755 - initialize blocksize for non-regular files (6062)
3756 - added --disable-server and --disable-client configure options (5782)
3757 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3758 - Vanilla 2.4.29 support
3759 - increase maximum number of obd devices to 520 (6242)
3760 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3761 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3762 SLES 9 2.6.5 kernels
3763 - limit stripes per file to 160 (the maximum EA size) (6093)
3765 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3768 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3769 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3770 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3771 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3772 - small ext3 extents cleanups and fixes (5733)
3773 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3774 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3775 - avoid needless client->OST connect, fix handle mismatch (5317)
3776 - fix DLM error path that led to out-of-sync client, long delays (5779)
3777 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3778 - fix several locking issues related to i_size (5492,5624,5654,5672)
3779 - don't move pending lock onto export if it is already evicted (5683)
3780 - fix kernel oops when creating .foo in unlinked directory (5548)
3781 - fix deadlock in obdfilter statistics vs. object create (5811)
3782 - use time_{before,after} to avoid timer jiffies wrap (5882)
3783 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3784 - Do not leak request if server was not able to process it (5154)
3785 - If mds_open unable to find parent dir, make that negative lookup(5154)
3786 - don't create new directories with extent-mapping (5909, 5936)
3788 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3789 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3790 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3791 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3792 - support for RHEL4 kernel on i686 (5773)
3793 - provide error messages when incompatible logs are encountered (5898)
3795 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3796 * version 1.4.0.10 (1.4.1 release candidate 1)
3798 - don't keep a lock reference when lock is not granted (4238)
3799 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3800 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3801 - drop import inflight refcount on signal_completed_replay error (5255)
3802 - unlock page after async write error during send (3677)
3803 - handle missing objects in filter_preprw_read properly (5265)
3804 - no transno return for symlink open, don't save no-trasno open (3440)
3805 - don't try to complete elan receive that already failed (4012)
3806 - free RPC server reply state on error (5406)
3807 - clean up thread from ptlrpc_start_thread() on error (5160)
3808 - readahead could read extra page into cache that wasn't ejected (5388)
3809 - prevent races in class_attach/setup/cleanup/detach (5260)
3810 - don't dereference de->d_inode after l_dput of de (5458)
3811 - use "int" for stripe value returned from lock_to_stripe (5544)
3812 - mballoc allocation and error-checking fixes in 2.6 (5504)
3813 - block device patches to fix I/O request sizes in 2.6 (5482)
3814 - look up hostnames for IB nals (5602)
3815 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3816 - don't start multiple acceptors for the same port (5277)
3817 - fix incorrect LASSERT in mds_getattr_name (5635)
3818 - export a proc file for general "ping" checking (5628)
3819 - fix "lfs check" to not block when the MDS is down (5628)
3821 - service request history (4965)
3822 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3823 - create an "evict_client" /proc entry on OSTs, like the MDS has
3824 - fix mount usage message, return errors per mount(8) (5168)
3825 - change grep [] to grep "[]" in tests so they work in more UMLs
3826 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3827 - remove ancient LOV_MAGIC_V0 EA support (5047)
3828 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3829 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3830 - allow readahead allocations to fail when low on memory (5383)
3831 - mmap locking landed again, after considerable improvement (2828)
3832 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3834 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3837 - send OST transaction number in read/write reply to free req (4966)
3838 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3839 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3840 - the watchdog thread now runs as interruptible (5246)
3841 - flock/lockf fixes (but it's still disabled, pending 5135)
3842 - don't use EXT3 constants in llite code (5094)
3843 - memory shortage at startup could cause assertion (5176)
3845 - reorganization of lov code
3846 - single portals codebase
3848 - add extents/mballoc support (5025)
3849 - direct I/O reads in the obdfilter (4048)
3850 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3852 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3855 - send OST transaction number in read/write reply to free req (4966)
3856 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3857 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3858 - the watchdog thread now runs as interruptible (5246)
3859 - handle missing objects in filter_preprw_read properly (5265)
3860 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3861 - drop import inflight refcount on signal_completed_replay error (5255)
3862 - unlock page after async write error during send (3677)
3863 - return original error code on reconstructed replies (3761)
3864 - no transno return for symlink open, don't save no-trasno open (3440)
3866 - add pid to ldlm debugging output (4922)
3867 - bump the watchdog timeouts -- we can't handle 30sec yet
3868 - extra debugging for orphan dentry/inode bug (5259)
3870 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3873 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3874 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3875 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3876 - fix NULL dereference in /proc/sys/portals/routes (4827)
3877 - allow failed mdc_close() operations to be interrupted (4561)
3878 - stop precreate on OST before MDS would time out on it (4778)
3879 - don't send partial-page writes before EOF from client (4410)
3880 - discard client grant for sub-page writes on large-page clients (4520)
3881 - don't free dentries not owned by NFS code, check generation (4806)
3882 - fix lsm leak if mds_create_objects() fails (4801)
3883 - limit debug_daemon file size, always print CERROR messages (4789)
3884 - use transno after validating reply (3892)
3885 - process timed out requests if import state changes (3754)
3886 - update mtime on OST during writes, return in glimpse (4829)
3887 - add mkfsoptions to LDAP (4679)
3888 - use ->max_readahead method instead of zapping global ra (5039)
3889 - don't interrupt __l_wait_event() during strace
3891 - add software watchdogs to catch hung threads quickly (4941)
3892 - make lustrefs init script start after nfs is mounted
3893 - fix CWARN/ERROR duplication (4930)
3894 - return async write errors to application if possible (2248)
3895 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3896 - print NAL number in %x format (4645)
3897 - update barely-supported suse-2.4.21-171 series (4842)
3898 - support for sles 9 %post scripts
3899 - support for building 2.6 kernel-source packages
3900 - support for sles km_* packages
3902 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3905 - ignore -ENOENT errors in osc_destroy (3639)
3906 - notify osc create thread that OSC is being cleaned up (4600)
3907 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3908 - reconstruct ost_handle() like mds_handle() (4657)
3909 - create a new thread to do import eviction to avoid deadlock (3969)
3910 - let lconf resolve symlinked-to devices (4629)
3911 - don't unlink "objects" from directory with default EA (4554)
3912 - hold socknal file ref over connect in case target is down (4394)
3913 - allow more than 32000 subdirectories in a single directory (3244)
3914 - fix blocks count for O_DIRECT writes (3751)
3915 - OST returns ENOSPC from object create when no space left (4539)
3916 - don't send truncate RPC if file size isn't changing (4410)
3917 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3918 - bind to privileged port in socknal and tcpnal (3689)
3920 - rate limit CERROR/CWARN console message to avoid overload (4519)
3921 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3922 - basic mmap support (3918)
3923 - kernel patch series update from b1_4 (4711)
3925 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3928 - avoid crash during MDS cleanup with OST shut down (2775)
3929 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3930 - don't use bad inodes on the MDS (3744)
3931 - dynamic object preallocation to improve recovery speed (4236)
3932 - don't hold spinlock over lock dumping or change debug flags (4401)
3933 - don't zero obd_dev when it is force cleaned (3651)
3934 - print grants to console if they go negative (4431)
3935 - "lctl deactivate" will stop automatic recovery attempts (3406)
3936 - look for existing locks in ldlm_handle_enqueue() (3764)
3937 - don't resolve lock handle twice in recovery avoiding race (4401)
3938 - revalidate should check working dir is a directory (4134)
3940 - don't always mark "slow" obdfilter messages as errors (4418)
3942 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3945 - don't close LustreDB during write_conf until it is done (3860)
3946 - fix typo in lconf for_each_profile (3821)
3947 - allow dumping logs from multiple threads at one time (3820)
3948 - don't allow multiple threads in OSC recovery (3812)
3949 - fix debug_size parameters (3864)
3950 - fix mds_postrecov to initialize import for llog ctxt (3121)
3951 - replace config semaphore with spinlock (3306)
3952 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3953 - don't allow enqueue to complete on a destroyed export (3822)
3954 - down write_lock before checking llog header bitmap (3825)
3955 - recover from lock replay timeout (3764)
3956 - up llog sem before sending rpc (3652)
3957 - reduce ns lock hold times when setting kms (3267)
3958 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3959 - fix NULL deref and obd_dev leak on setup error (3312)
3960 - replace some LBUG about llog ops with error handling (3841)
3961 - don't match INVALID dentries from d_lookup and spin (3784)
3962 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3963 - fix invalid assertion in ptlrpc_set_wait (3880)
3965 - add libwrap support for the TCP acceptor (3996)
3966 - add /proc/sys/portals/routes for non-root route listing (3994)
3967 - allow setting MDS UUID in .xml (2580)
3968 - print the stack of a process that LBUGs (4228)
3970 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3973 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3974 - make sure to unset replay flag from failed open requests (3440)
3975 - if default stripe count is 0, use OST count for inode size (3636)
3976 - update parent mtime/ctime on client for create/unlink (2611)
3977 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3978 - free recovery state on server during a forced cleanup (3571)
3979 - unregister_reply for resent reqs (3063)
3980 - loop back devices mounting and status check on 2.6 (3563)
3981 - fix resource-creation race that can provoke i_size == 0 (3513)
3982 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3983 - more debugging for page-accounting assertion (3746)
3984 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3985 - avoid deadlock after precreation failure (3758)
3986 - fix race and lock order deadlock in orphan handling (3450, 3750)
3987 - add validity checks when grabbing inodes from l_ast_data (3599)
3989 - add /proc/.../recovery_status to obdfilter (3428)
3990 - lightweight CDEBUG infrastructure, debug daemon (3668)
3991 - change default OSC RPC parameters to be better on small clusters
3992 - turn off OST read cache for files smaller than 32MB
3993 - install man pages and include them in rpms (3100)
3994 - add new init script for (un)mounting lustre filesystems (2593)
3995 - run chkconfig in %post for init scripts (3701)
3996 - drop scimac NAL (unmaintained)
3998 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4001 - clean kiobufs before and after use (3485)
4002 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4003 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4004 - change init script to fail more clearly if not run as root (1528)
4005 - allow clients to reconnect during replay (1742)
4006 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4007 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4008 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4009 - fix cond_resched (3554)
4010 - only evict unfinished clients after recovery (3515)
4011 - allow bulk resend, prevent data loss (3570)
4012 - dynamic ptlrpc request buffer allocation (2102)
4013 - don't allow unlinking open directory if it isn't empty (2904)
4014 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4015 - remove extraneous obd dereference causing LASSERT failure (3334)
4016 - don't use get_cycles() when creating temp. files on the mds (3156)
4017 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4018 - handle EEXIST for set-stripe, set proper directory name (3336)
4020 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4021 - fix ksocknal_fmb_callback() error messages (2918)
4023 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4026 - don't copy lvb into (possibly NULL) reply on error (2983)
4027 - don't deref dentry after dput, don't free lvb on error (2922)
4028 - use the kms to determine writeback rpc length (2947)
4029 - increment oti_logcookies when osc is inactive (2948)
4030 - update client's i_blocks count via lvb messages (2543)
4031 - handle intent open/close of special files properly (1557)
4032 - mount MDS with errors=remount-ro, like obdfilter (2009)
4033 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4034 - don't use cancelling-locks' kms values (2947)
4035 - use highest lock extent for kms, not last one (2925)
4036 - don't dereference ERR_PTR() dentry in error handling path (3107)
4037 - fix thread race in portals_debug_dumplog() (3122)
4038 - create lprocfs device entries at setup instead of at attach (1519)
4039 - common AST error handler, don't evict client on completion race (3145)
4040 - zero nameidata in detach_mnt in 2.6 (3118)
4041 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4042 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4043 - initialize RPC timeout timer earlier for 2.6 (3219)
4044 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4045 - print nal/nid for unknown nid (3258)
4046 - additional checks for oscc recovery before doing precreate (3284)
4047 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4048 - don't crash in mdc_close for bad permissions on open (3285)
4049 - zero i_rdev for non-device files (3147)
4050 - clear page->private before handing to FS, better assertion (3119)
4051 - tune the read pipeline (3236)
4052 - fix incorrect decref of invalidated dentry (2350)
4053 - provide read-ahead stats and refine rpc in flight stats (3328)
4054 - don't hold journal transaction open across create RPC (3313)
4055 - update atime on MDS at close time (3265)
4056 - close LDAP connection when recovering to avoid server load (3315)
4057 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4058 - don't leak open file on MDS after open resend (3325)
4059 - serialize filter_precreate and filter_destroy_precreated (3329)
4060 - loop device shouldn't call sync_dev() for nul device (3092)
4061 - clear page cache after eviction (2766)
4062 - resynchronize MDS->OST in background (2824)
4063 - refuse to mount the same filesystem twice on same mountpoint (3394)
4064 - allow llmount to create routes for mounting behind routers (3320)
4065 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4066 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4067 - fix rare problem with rename on htree directories (3417)
4069 - allow default OST striping configuration per directory (1414)
4070 - fix compilation for qswnal for 2.6 kernels (3125)
4071 - increase maximum number of MDS request buffers for large systems
4072 - change liblustreapi to be useful for external progs like lfsck (3098)
4073 - increase local configuration timeout for slow disks (3353)
4074 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4076 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4079 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4080 - don't overwrite extent policy data in reply if lock was blocked (2901)
4081 - drop filter export grants atomically with removal from device (2663)
4082 - del obd_self_export from work_list in class_disconnect_exports (2908)
4083 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4084 - swab reply message in mdc_close, other PPC fixes (2464)
4085 - fix destroying of named logs (2325)
4086 - overwrite old logs when running lconf --write_conf (2264)
4087 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4088 - fix race in target_handle_connect (2898)
4089 - mds_reint_create() should take same inode create lock (2926)
4090 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4091 - don't close files for self_export to avoid uninitialized obd (2936)
4092 - allow MDS with the same name as client node (2939)
4093 - hold dentry reference for closed log files for unlink (2325)
4094 - reserve space for all logs during transactions (2059)
4095 - don't evict page beyond end of stripe extent (2925)
4096 - don't oops on a deleted current working directory (2399)
4097 - handle hard links to targets without a parent properly (2517)
4098 - don't dereference NULL lock when racing during eviction (2867)
4099 - don't grow lock extents when lots of conflicting locks (2919)
4101 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4104 - account for cache space usage on clients to avoid data loss (974)
4105 - lfsck support in lustre kernel code (2349)
4106 - reduce journal credits needed for BRW writes (2370)
4107 - orphan handling to avoid losing space on client/server crashes
4108 - ptlrpcd can be blocked, stopping ALL progress (2477)
4109 - use lock value blocks to assist in proper KMS, faster stat (1021)
4110 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4111 - recovery for initial connections (2355)
4112 - fixes for mds_cleanup_orphans (1934)
4113 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4114 - block all file creations until orphan recovery completes (1901)
4115 - client remove rq_connection from request struct (2423)
4116 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4117 - recovery timer race (2670)
4118 - mdc_close recovey bug (2532)
4119 - ptlrpc cleanup bug (2710)
4120 - mds timeout on local locks (2588)
4121 - namespace lock held during RPCs (2431)
4122 - handle interrupted sync write properly (2503)
4123 - don't try to handle a message that hasn't been replied to (2699)
4124 - client assert failure during cleanup after abort recovery (2701)
4125 - leak mdc device after failed mount (2712)
4126 - ptlrpc_check_set allows timedout requests to complete (2714)
4127 - wait for inflight reqs when ptlrpcd finishes (2710)
4128 - make sure unregistered services are removed from the srv_list
4129 - reset bulk XID's when resending them (caught by 1138 test)
4130 - unregister_bulk after timeout
4131 - fix lconf error (2694)
4132 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4133 - readahead locks pages, leaves pending causing memory pressure (2673)
4134 - increase OST request buffers to 4096 on large machines (2729)
4135 - fix up permission of existing directories in simple_mkdir (2661)
4136 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4137 - don't assign transno to errored transactions (2742)
4138 - don't delete objects on OST if given a bogus objid from MDS (2751)
4139 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4140 - if rq_no_resend, then timeout request after recovery (2432)
4141 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4142 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4143 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4144 - MDS recovery shouldn't delete objects that it has given out (2730)
4145 - if enqueue arrives after completion, don't clobber LVB (2819)
4146 - don't unlock pages twice when trigger_group_io returns error (2814)
4147 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4148 - don't write pages to disk if there was an error (1450)
4149 - don't ping imports that have recovery disabled (2676)
4150 - take buffered bytes into account when balancing socknal conn (2817)
4151 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4152 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4153 - remove little-endian swabbing of llog records (1987)
4154 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4155 - retry reposting request buffers if they fail (1191)
4156 - grow extent at grant time to avoid granting a revoked lock (2809)
4157 - lock revoke doesn't evict page if covered by a second lock (2765)
4158 - disable VM readahead to avoid reading outside lock extents (2805)
4160 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4161 - updated kernel patches for hp-2.4.20 kernel (2681)
4163 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4166 - fix truncated write corruption (2366)
4167 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4169 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4170 - ensure len(uuid) < 37 in lmc (1171)
4171 - fix ia64 OOPS in llog_test (2255)
4172 - zero end of page at obdfilter for partial page writes (2648)
4173 - don't leave stale dentries around after renames (bug 2428)
4174 - fix timeouts when evicting a client with a single lock held (2642)
4175 - set deadline for the initial HELLO message to drain (2634)
4176 - print out dotted-quad IP addresses in the socknal (2302)
4178 - additional debugging for MDS client eviction problem (2443)
4179 - fix mkfsoptions support for osts (2603, 2604)
4181 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4184 - add series for the vanilla 2.6.0 kernel
4185 - add series for the vanilla 2.4.24 kernel
4186 - add series for a cray x86/64 UL kernel drop
4187 - fix xattr patches for the vanilla 2.4.19 series
4189 - generate true UUIDs in lmc (1171)
4190 - have portals stack dumping break in UML (2466)
4191 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4192 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4193 - wake callback waiting threads on client eviction (2460)
4194 - Add --ptldebug and --subsystem to lmc (1719)
4195 - update assertion to allow safe interrupt allocation
4196 - set rq_no_resend for cancel requests (2432)
4197 - recalculate ptlrpcd timeout after resend (2494)
4198 - call vfs_rmdir when removing pending directories (2368)
4199 - fix renaming a file to itself (2429)
4200 - lmc creates a default one-stripe lov (2454)
4201 - expand procfs space to handle large clusters (2326)
4202 - increase UML stack to avoid overflow
4203 - update lconf's list of debug and subsystem masks
4204 - fix lfs find --obd (2510)
4205 - /proc tunable for disabling filter read caching (2591)
4206 - stop rpm packages from altering slapd.conf (2301)
4207 - disable nagle in the socknal under 0conf (2578)
4208 - choose mds inode size based on stripe count (2572)
4209 - fix kernel-source rpm problems (2516)
4211 - add --disable-doc to avoid pdf generation (2421)
4212 - update documentation, tests, type-os, comments
4213 - avoid format warnings on ia64
4214 - remove the TOE NAL
4215 - tiny code cleanups by removing unused fields
4217 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4220 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4221 - remove the most blatant lies from BUILDING file (2371)
4222 - change default debug level to reasonable production setting
4223 - reduce client side cache size to reduce cache flush time
4224 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4225 - make TCP zerocopy and pinger support enabled by default (2476)
4226 - sync writes completed after process exits caused crashes (2319)
4227 - maintain correct mount count on the MDS (2356)
4228 - backout 1557, because 2316 wasn't really fixed
4229 - better file I/O statistics gathering in /proc
4230 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4231 - another kernel patch to fix zero-copy TCP function export
4232 - don't take duplicate lock when processing re-sent getattr (2420)
4233 - lctl uses obd_self_export instead of creating new conn (2353)
4234 - MDS/OST recovery case which requires object creation asserted (2425)
4235 - move lfs from /usr/sbin to /usr/bin in packages
4236 - fix race between mds_client_add and mds_client_free (2417)
4237 - use kmalloc instead of slabs in portals (2430)
4238 - don't create duplicate records when a failover MDS is present (2442)
4239 - remove unnecessary mount age check (2332)
4240 - don't remove directory inodes from locks prematurely (2451)
4241 - don't break if MDS service name is the same as hostname (2103)
4242 - fix races in client write RPC generation when cache full (2482)
4244 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4247 - remove now-unused request->rq_obd (278)
4248 - if an allocation fails, print out how much memory we've used (1933)
4249 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4250 - add the "configurable stack size" patch to most series files (1256)
4251 - ability to write large log records, for 100+ OST configs (2306)
4252 - fix NULL deref when filter_prep fails (2314)
4253 - fix operator precedence error in filter_sync
4254 - dynamic allocation of socknal TX descriptors (2315)
4255 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4256 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4257 - ensure that CWARN messages in Portals always get to the syslog
4258 - __init/__exit are not for prototype decls (ldlm_init/exit)
4259 - x86-64 compile warning fixes
4260 - fix gateway LMC keyword conflict (2318)
4261 - fix MDS lock inversions in getattr/reint paths (1844)
4262 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4263 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4264 - DLM race condition prevented some lock evictions (2328)
4265 - ENOMEM detection and retry on socknal sends (2230)
4266 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4267 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4268 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4269 - better error messages when a client is rejected during recovery (1505)
4270 - avoid cancelling locks which were never granted, after failure (2330)
4271 - fix i_sem/journal inversion in mds_client_add (2333)
4272 - fix truncate/getattr lock cycle deadlock (2334)
4273 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4274 - fix two rare exit paths which could leak an l_lock() ref (2321)
4275 - fencepost error in MDS/OST orphan recovery (2226)
4276 - make log record alignment 8 bytes (1988)
4277 - lstripe now fails when requested offset > ost_count (2237)
4278 - ensure that all kernel series have a complete list.h (1607)
4279 - fix crashes in special-file operations (2316)
4280 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4282 - allow configurable automake binary, for testing new versions
4283 - small update to the lfs documentation
4285 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4287 * fix negative export reference count in fsfilt_sync (2312)
4289 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4290 * release candidate 0.9.1
4292 - orphans are moved into the PENDING directory for possible recovery
4293 - replayed opens now open by fid for orphan/rename safety (1042)
4294 - last close of an orphan inode generates a transno (683)
4295 - chdir() and mount() now pin the directory entry (1020)
4296 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4297 - discard very old requests without processing them (1502)
4298 - remove some common, well-understood CERRORs (1505)
4299 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4300 - clear "grant" flags in OST replies until OST grant code lands (1644)
4301 - fix read performance by not clobbering i_blksize on client (1598)
4302 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4303 - make LRU size a /proc tunable, clears locks when reduced (707)
4304 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4305 - server locks take references on exports now (1558)
4306 - build fixes for 2.4.20-rh trees (1663)
4307 - return an error from lov_create if all OSCs are inactive (1751)
4308 - fix import levels when a reconnect happens without a timeout (1597)
4309 - exit early from mds_open if we get a lookup error (1749)
4310 - partial page read at EOF wouldn't wait for disk before sending (1642)
4311 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4312 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4313 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4314 - write/truncate lock inversion (1639)
4315 - Don't auto-load obdclass, portals modules during cleanup (1495)
4316 - fix timestamps from jumping to "now" (1763)
4317 - extra journal assertions (1648)
4318 - add an extra multiunlink test (1771)
4319 - fix read_record/write_record API (1776)
4320 - fix leak of offset_extent, possible incorrect i_size later (1772)
4321 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4322 - Debugging for the kqswnal_get_idle_tx problems (1820)
4323 - Allow recovery to be attempted multiple times (1536)
4324 - Write out MDS last_rcvd file after it is first created (1600)
4325 - Fix tx_descriptor leak in failed transmit situations (1827)
4326 - ext3 journaling fixes for assertion failure after IO error (1871)
4327 - class_export_put() on freed export after completion AST error (1896)
4328 - Fix revalidate looping in VFS (1322)
4329 - Don't access a freed export during MDS_REINT timeout (1521)
4330 - Add open-unlink recovery support on the MDS (1673,1764)
4331 - Return an error if no MDS data was read from last_rcvd (1946)
4332 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4333 - Don't setup a disk device that doesn't match exported UUID (317)
4334 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4335 - avoid committing NULL handle in force close
4336 - local.sh is now a one-stripe LOV configuration
4337 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4338 - fix ext3 htree duplicate directory entry corruption (1516)
4339 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4340 - update to new LOV EA format (2097)
4341 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4343 2003-06-15 Phil Schwan <phil@clusterfs.com>
4346 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4347 - per-import/export recovery handling (958, 931, 959)
4348 - multiple last-rcvd slots, for serving multiple FSes (949)
4349 - connections are again shared between multiple imp/exports (963, 964)
4350 - "umount -f" would hang if any requests needed to be sent (393, 978)
4351 - avoid pinning large req buffer by copying for queued messages (989)
4352 - add "uuid" to "lctl device" command to help upcalls (991)
4353 - "open" RPCs with transnos would confuse recovery counters (1037)
4354 - do proper endian conversion of all wire messages (288, 340, 891)
4355 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4356 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4357 - fix ptlrpc_connection leak in target_handle_connect (1174)
4358 - fix import refcounting bug in OST and MDS cleanup (1134)
4359 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4360 - fix very unlikely obd_types race condition (501)
4361 - remove osc_open hack for echo_client (1187)
4362 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4363 - a failure in read_inode2 leads to deadlock (1139)
4364 - cancel ack-locks as soon as transaction is committed (1072)
4365 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4366 - make sure to commitrw after any preprw to avoid deadlock (1162)
4367 - failing to execute a file in a lustre FS would lock inode (1203)
4368 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4369 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4370 - obd_uuid2tgt would walk off the end of the list (1255)
4371 - on IA64 the osc would give portals incorrect bulk size (1258)
4372 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4373 - fix lock inversion caused by new llite matching code (1282)
4374 - limit the number of dirty pages on a client to 10MB (1286)
4375 - timed out locks were not being corrected cancelled (1289)
4376 - fix O_DIRECT above 4GB on IA-32 (1292)
4377 * major user-visible changes
4378 - fail out/fail over policy now controlled by the upcall (993)
4380 - add OBD_PING to check server availability and failure (954)
4381 - lustre messages are now sent in sending host order (288, 340, 891)
4382 - add eadatalen to MDS getattr reply (340)
4383 - OST read replies may contain second buffer, with per-page status (593)
4385 2003-03-11 Phil Schwan <phil@clusterfs.com>
4388 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4389 - failed open()s could cause deadlock; fixed (867, 869)
4390 - stop cancelling OST locks when files are closed (481)
4391 - overlapping XID spaces caused network corruption (851, 853)
4392 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4393 - setattr_raw added, to do single-RPC, server-side setattrs
4394 - lmc/lconf syntax change for OST UUIDs
4395 - fix crashy race condition between ptlrpc_free_req and osc_close
4396 - don't use request in mdc_enqueue if we hit a timeout (889)
4397 - don't set the inode i_size for regular files from the MDS (896)
4398 - handle out of order completion AST (842)
4399 - don't LBUG if a lock request times out after receiving AST (913)
4400 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4401 - if a bad lock AST arrives, send an error instead of dropping entirely
4402 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4403 - fix leak in bulk IO when only partially completed (899, 900, 926)
4404 - fix O_DIRECT for ia64 (55)
4405 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4406 - C-z after timeout could hang a process forever; fixed (977)
4408 - client-side I/O cache (678, 924, 929, 941, 970)
4410 - READPAGE and SETATTRs which don't take server-side locks get
4413 2003-02-11 Phil Schwan <phil@clusterfs.com>
4416 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4417 - Fix fsfilt_extN_readpage() to read a full page of directory
4418 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4419 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4420 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4421 - Most intent operations were removed, in favour of a new RPC mode
4422 that does a single RPC to the server and bypasses most of the VFS
4423 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4424 - Aggressively cancel local locks on DLM servers
4425 - mds_reint_unlink sends EA to the client if it's the last nlink.
4426 client uses that EA to unlink OST objects.
4427 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4428 - recursive symlinks were fixed (439)
4429 - fixed NULL deref in DEBUG_REQ
4430 - filter_update_lastobjid no longer calls sync, which annoyed extN
4431 - fixed multi-client small-writes to a single file problem (445)
4432 - fixed mtime updates during file writes (607)
4433 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4434 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4435 - fixed Linux OST size reporting problem (444, 656)
4436 - OST now updates object mtime with writes or setattr (607, 619)
4437 - client verifies file size before zeroing page past EOF (445)
4438 - OST now writes last allocated objid to disk with allocation (108)
4439 - LOV on echo now works (409)
4441 - mds_reint_unlink sends a new buffer, with the EA included. this
4442 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4443 set if a regular file was being unlinked, and it was the last link
4444 - use PtlGet from the target for bulk writes (315)
4445 - OST now updates object mtime with writes or setattr (607, 619)
4446 - LDLM now has a grant-time callback to revalidate locked items, if
4448 - Many MDS operations were reorganized to combat race conditions
4450 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4451 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4453 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4456 - Fully reactivate OST imports after reconnection (512, others)
4457 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4458 - More graceful error handling for truncating on dead OST (515)
4459 - Don't error out unless we're actually accessing dead stripes (474)
4460 - Fix garbage sizes when stripes are missing (410)
4461 - LRU counters were broken, causing constant lock purge (433, 432)
4462 - garbage on read from stripes with failed OSTs (441)
4463 - mark OSCs as active before reconnecting during recovery (438)
4464 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4465 - lfind did not preserve OST order in output (443)
4466 - symlinks cause hung clients, incorrect data (439)
4467 - stop dereferencing request after dropping refcount (457)
4468 - don't LASSERT(spin_is_locked) on non-SMP (455)
4469 - fixes for many rename() bugs
4470 - fstat didn't correctly synchronize attributes (399)
4471 - server must handle lock cancellation during blocking AST prep (487)
4472 - bulk descriptors were free()d too soon (511)
4473 - fix paths in lconf, which would load incorrect modules (451, 507)
4474 - fix confusing lconf 'host not found' error message (386)
4475 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4476 - fix race condition in mdc_blocking_ast() for inode access (526)
4477 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4478 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4479 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4480 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4481 - fix lstripe in several ways (kernel side) (527)
4482 - fix request leak in ldlm_cli_enqueue (262)
4483 - incorrect OSC was marked inactive after OST failure
4484 - call mds_fs_cleanup before unmounting filesystem (524)
4485 - fix races between taking ns_lock and ldlm_lock_change_resource
4486 - fix races updating LOV export open file list
4487 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4488 - fix recovery NULL deref in ldlm_cli_cancel_unused
4489 - fix some DLM races by using new hash table for lock handles (419)
4490 - permit the client to specify desired inodes, at replay
4491 - duplicate requests when we queue them for replay reintegration
4492 - fix last_rcvd offset calculation
4493 - sync after each recovered transaction, so we always make progress
4494 - never, not always, ERESTART requests without transnos
4495 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4497 - skip replay if the MDS says that the client is already connected
4498 - don't check for a recovery-enabled export to match lctl's UUID
4499 - don't INC_USE_COUNT for phantom exports
4500 - don't crash when cleaning up phantom exports (567)
4501 - don't double-finish or set replay data for errored mdc_open requests
4502 - abort requests when they time out, so we don't get old replies
4503 - send/receive replies for AST messages again
4504 - if the client says that it doesn't have the lock, cancel it on the
4506 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4507 mark it as destroyed, it will all work out when decref is called
4508 - fix module use counts (22, 581)
4510 - ASTs now expect a reply (server cancels lock on error reply)
4512 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4515 - fix many simultaneous client startup (392)
4516 - fix dentry->d_it clobbering
4517 - credentials weren't being shipped for readdir/getattr operations
4518 - remove invalid assertions triggered during some concurrent MD
4520 - proper Lustre versions added (336, 389)
4521 - fix memory leak for create error case (398)
4522 - fix LOV locking bug that would get cli/srv out of sync
4523 - fix echo client over LOV (409)
4524 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4525 - fix double-O_EXCL intent crash (424)
4526 - avoid sending multiple lock CANCELs (352)
4528 - MDS can do multi-client recovery (modulo bugs in new code)
4530 - many updates, edits, cleanups
4532 2002-11-18 Phil Schwan <phil@clusterfs.com>
4535 - fix null d_it dereference (346)
4536 - fix full OST/dbench hang (333)
4537 - fix permission problem with file removal (286)
4538 - fix removal of OSCs from LOV when they fail
4539 - fix NULL deref during bulk timeout (214)
4540 - fix problems related to multiple filesystems on one MDS (241)
4541 - fixed serious subtle metadata locking bugs
4542 - free locks on clients when inodes are removed due to memory
4544 - fix inode pointer in lock data (285)
4545 - partial support for multiple MDS on a single host (241)
4546 - data locks weren't cancelled at clear_inode time (290, 311)
4547 - intent locks could lead to unbounded lock growth (205)
4548 - added a maximum lock count, an LRU list, and a flusher
4549 - fix multiple rename (365)
4550 - properly abstracted the echo client
4551 - OSC locked 1 byte too many; fixed
4552 - rewrote brw callback code:
4553 - fixed recovery bugs related to LOVs (306)
4554 - fixed too-many-pages-in-one-write crash (191)
4555 - fixed (again) crash in sync_io_timeout (214)
4556 - probably fixed callback-related race (385)
4558 - Add capability to MDS protocol
4559 - LDLM cancellations and callbacks on different portals
4561 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4564 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4565 - defer open object destruction to close time (601981, 138)
4566 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4567 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4568 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4569 - don't call obd_setattr in ll_file_release for destroyed objects
4570 * protocol change to lustre_msg: move |version| and add |flags|
4571 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4572 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4573 * added replay of create, unlink, link and rename operations during
4574 MDS failover; recovery should be much more robust now
4575 * remove failed OSCs from LOVs (only lov_create uses this so far)
4576 * the lustre-HOWTO was brought (more) up to date (582544)
4578 2002-10-23 Phil Schwan <phil@clusterfs.com>
4581 - in-use dentries weren't being reused properly (617851)
4582 - prevent multiple LDLM setup (599178)
4583 - fix LOV size calculations for truncate (617853)
4584 - fix client handling of MDS intent errors (POSIX)
4585 - fix permission bug in lovstripe.c test (624321)
4586 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4587 - truncate past end of file could corrupt data
4588 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4589 - a race in recovery could return ETIMEDOUT to apps (623947)
4590 - building outside the source directory was fixed
4591 * the lustre-HOWTO was brought (more) up to date (582544)
4592 * major progress was made on recovery functionality
4594 2002-10-10 Phil Schwan <phil@clusterfs.com>
4597 - recovery deadlock fix
4598 - rm -rf causes LBUG fix (617817)
4599 - file open by multiple tasks fix (618962)
4600 - directory permissions bugs (602707 and 620007)
4601 - journal_stop fixed with locking (611313)
4602 - O_APPEND failures resolved (618273, perhaps 614459)
4603 - lconf PATH fix (619770)
4604 - IA64 build fix (621450)
4605 - RPC buffer sizes scale with amount of memory
4607 2002-10-01 Phil Schwan <phil@clusterfs.com>
4610 - locks would be cancelled without throwing away data pages,
4611 resulting in inconsistent data (605627)
4612 - inode attributes were not always being refreshed (605627, 612449)
4613 - lconf now continues to cleanup after lctl reports an error
4614 - MDS now enforces user permissions (602707)
4615 - lprocfs cleanup fixed, but not yet enabled (614157)
4616 - fixed infinite server hang, should a client not respond to an AST
4617 - avoid going into recovery if user calls readlink() with a buffer
4618 that's too small (613941)
4619 - AST RPCs no longer require replies (614867) -- this may be changed
4620 - don't crash server if client sends an IOV that's too big (611336)
4621 - fixed lock conversion deadlock (611892)
4622 - fixed the following of symlinks (614622)
4623 * recovery: the server can remove locks from a client that dies, other
4624 clients can make progress
4625 * more extN patch fixes
4626 * compile-time configurable ptlrpc buffer allocations
4628 - collaborative read cache document
4629 - Lustre Lite Performance CDR document-in-progress
4631 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4634 - fix typo in patch-2.4.18
4636 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4639 - clear ptlrpc request each time in handle_incoming_request()
4640 - unlink of files now destroys the object on the OST
4642 2002-09-19 Peter Braam <braam@clusterfs.com>
4644 * add hard link support
4645 * change obdfile creation method
4646 * kernel patch changed
4648 2002-09-19 Peter Braam <braam@clusterfs.com>
4651 - stack overflow bug in extN fixed
4653 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4655 * documentation updates
4656 - add man pages for config tools
4657 - update tests/README to describe testing with new config tools
4658 - finish metadata API descriptions
4659 * bug fixes and cleanups
4660 - statfs workaround for 16TB limit
4661 - LOV stripe allocation improved, can stripe on subset of OSTs
4662 - LOV file size/IO offset was wrong for files > 4GB in size
4663 - object EA data was being dropped, caused files to be unreadable
4664 - memory overflow with non-LOV OST caused memory corruption
4665 - fixed regression tests to work with new config tools, obdfilter
4666 - fixed bug when directory size became larger than 1 block
4667 - fixed bug (for single client case) when PWD was deleted
4668 - invalidate local directory pages when doing intent-based ops
4669 - avoid LDLM oops when lock callback contained bad data
4671 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4673 * documentation updates
4674 * bug fixes and cleanups
4675 - configuration tools
4678 - 64-bit compile warnings
4679 - 64-bit internal statfs data
4681 * test_brw on persistent OST devices
4683 * lprocfs (disabled)
4685 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4687 * documentation updates
4688 * bug fixes and cleanups
4689 * configuration tools
4691 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4694 * many small fixes to 0_5_4
4695 * io/network handling
4696 * thinkos in MDS operations
4698 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4701 * crucial basic fixes to 0.5.3
4702 * IOR, Iozone work over Elan
4705 2002-08-07 Phil Schwan <phil@clusterfs.com>
4706 * version 0_5_3, our first alpha
4707 * we use the new Portals iovs
4708 * documentation updates
4709 * bug fixes and cleanups
4710 * small changes in the DLM wire protocol
4712 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4713 * version 0_5_1 with some initial stability,
4714 * locking on MD and file I/O.
4715 * documentation updates
4716 * several bug fixes since 0.5.0
4717 * small changes in wire protocol
4719 2002-07-18 Phil Schwan <phil@clusterfs.com>
4721 * delivered as Lustre Light Alpha
4722 * fixed a crash after handling invalid MDS requests
4723 * fixed directory pages for architectures with non-4k pages sizes
4725 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4726 * release version v0_4_4
4727 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4729 * Fixes a number of interruption problems with OST operations.
4730 * Update documentation for portals header changes
4731 * Move all wire protocol structs/defines to lustre_idl.h
4732 * Fixes symlink length bug.
4733 * Add tcpdump to repository.
4735 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4736 * release version v0_4_3
4737 * Fixes statfs for inodes on extN.
4738 * Fixes bug in runtests which would delete /etc/hosts.
4739 * Use 64-bit object IDs wherever possible (not into VFS though)
4740 Remove ost_get_info, which is unused by lustre, and out of date.
4742 2002-07-03 Peter Braam <braam@clusterfs.com>
4743 * release version v0_4_2 Fixes a lookup error (type not passed)
4744 * move forward to head of Portals
4745 * move forward to latest Lustre kernel
4747 2002-06-25 Peter Braam <braam@clusterfs.com>
4748 * release version v0_4_1. Hopefully stable on single node use.