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: Stack overflow during MDS log replay
18 Details : ease stack pressure by using a thread dealing llog_process.
22 Description: wait until IO finished before start new when do lock cancel.
23 Details : VM protocol want old IO finished before start new, in this case
24 need wait until PG_writeback is cleared until check dirty flag and
25 call writepages in lock cancel callback.
27 Severity : enhancement
29 Description: using special macro for print time and cleanup in includes.
33 Description: mds_mfd_close() ASSERTION(rc == 0)
34 Details : In mds_mfd_close(), we need protect inode's writecount change
35 within its orphan write semaphore to prevent possible races.
39 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
43 Frequency : rare, on shutdown ost
44 Description: don't hit live lock with umount ost.
45 Details : shrink_dcache_parent can be in long loop with destroy dentries,
46 use shrink_dcache_sb instead.
50 Description: don't panic with use echo client
51 Details : echo client pass NULL as client nid pointer and this produce null
56 Description: fix build on ppc32
57 Details : compile code with -m64 flag produce wrong object file for ppc32.
61 Description: add message levels for liblustreapi
65 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
66 Details : Move the 'good_osts' check before the 'total_bavail' check. This
67 will result in an -EAGAIN and in the exit call path we call
68 alloc_rr() which will with increasing aggressiveness attempt to
69 aquire precreated objects on the minimum number of required OSCs.
73 Description: Use old size assignment to avoid deadlock
74 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
75 the scheduling while holding a spinlock. We do not need locking
76 for size in ll_update_inode() because size is only updated from
77 the MDS for directories or files without objects, so there is no
78 other place to do the update, and concurrent access to such inodes
79 are protected by the inode lock.
83 Description: resolve "_IOWR redefined" build error on SLES10
87 Description: dump the memory debugging after all modules are unloaded to
88 suppress false negative in conf_sanity test 39
90 Severity : enhancement
92 Description: build kernel-ib packages for OFED 1.3 in our release cycle
97 Description: fix SLES kernel versioning
98 Details : the kernel version for our SLES 10 kernel did not include a "-"
99 before the "smp" at the end. while this was not a problem in
100 general, it did mean that software trying to use the kernel
101 version to try to detect a vendor specific kernel would fail.
102 this was most evident by the OFED build scripts.
106 Description: Don't update lov_desc members until making sure they are valid
107 Details : When updating lov_desc members via proc fs, need fix their
108 validities before doing the real update.
112 Description: don't put request into delay list while invalidate in flight.
113 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
114 import in flight. this produce timeout for invalidate and sometimes
115 can cause stale data.
119 Frequency : on ppc only
120 Description: not convert ost objects for directory because it's not exist.
121 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
125 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
127 Severity : enhancement
129 Description: Update to support 2.6.22.14 vanilla kernel.
133 Frequency : rare, on recovery
134 Description: read procfs can produce deadlock in some situation
135 Details : Holding lprocfs lock which send rpc can produce block for destroy
136 obd objects and this also block reconnect with -EALREADY. This isn't
137 fix all lprocfs bugs - but make it rare.
139 Severity : enhancement
141 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
144 Frequency : frequent on X2 node
146 Description: mdc_set_open_replay_data LBUG
147 Details : Set replay data for requests that are eligible for replay.
151 Description: lustre_mgs: operation 101 on unconnected MGS
152 Details : When MGC is disconnected from MGS long enough, MGS will evict the
153 MGC, and late on MGC cannot successfully connect to MGS and a lot
154 of the error messages complaining that MGS is not connected.
157 Frequency : on start mds
159 Description: Implement get_info(last_id) in obdfilter.
162 Frequency : occasional
164 Description: Correctly check stale fid, not start epoch if ost not support SOM
165 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
166 overwrited with new generated fid, but mds can anwer with one of these
167 two fids and both is not stale. setattr incorectly start epoch and
168 assume will be called done_writeting, but without SOM done_writing
172 Frequency : rare, depends on device drivers and load
174 Description: MDS or OSS nodes crash due to stack overflow
175 Details : Code changes in 1.8.0 increased the stack usage of some functions.
176 In some cases, in conjunction with device drivers that use a lot
177 of stack the MDS (or possibly OSS) service threads could overflow
178 the stack. One change which was identified to consume additional
179 stack has been reworked to avoid the extra stack usage.
182 Frequency : occasional
184 Description: Do not fail import if osc_interpret_create gets -EAGAIN
185 Details : If osc_interpret_create got -EAGAIN it immediately exits and
186 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
187 oscc_has_objects and see OSC has no objests and call
188 oscc_internal_create to resend create request.
190 Severity : enhancement
192 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
194 Severity : enhancement
196 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
199 Frequency : very rare
201 Description: Fix replay if there is an un-replied request and open
202 Details : In some cases, older replay request will revert the
203 mcd->mcd_last_xid on MDS which is used to record the client's
206 Severity : enhancement
208 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
210 Severity : enhancement
212 Description: Add rhel5 support to HEAD.
214 Serverity : enhancement
216 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
221 Description: Don't allow skipping OSTs if index has been specified.
222 Details : Don't allow skipping OSTs if index has been specified, make locking
223 in internal create lots better.
227 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
228 Details : access to bitfield in structure is always rounded to long
229 and this produce problem with not atomic change any bit.
233 Description: Lustre make rpms failed.
234 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
235 builds Lustre rpms from tarball.
238 Frequency : rare on shutdown ost
240 Description: If llog cancel was not send before clean_exports phase, this can
241 produce deadlock in llog code.
242 Details : If llog thread has last reference to obd and call class_import_put
243 this produce deadlock because llog_cleanup_commit_master wait when
244 last llog_commit_thread exited, but this never success because was
245 called from llog_commit_thread.
249 Description: allow userland application know is lost one of stripes.
250 Details : fill lvb_blocks with error code on ost and return it to
251 application if error flag found.
255 Description: NULL lov_tgts causing MDS oops
256 Details : more safe checks for NULL lov_tgts for avoid oops.
258 Severity : enhancement
260 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
264 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
265 Details : If an OST is down the MDS will hang indefinitely in
266 obd_statfs_rqset() waiting for the statfs data. While for
267 MDS QOS usage of statfs, it should not stuck in waiting.
269 Severity : enhancement
271 Description: remote_acl support
272 Details : Support ACL-based permission check for remote user.
273 Support setfacl/getfacl for remote user with the utils
274 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
275 the system "{s,g}etfacl" utils.
277 Severity : enhancement
279 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
281 Severity : enhancement
283 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
287 Description: make mgs_setparam() handle fsname containing dash
288 Details : fsname containing a dash does not work with lctl conf_param
290 Severity : enhancement
292 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
294 Severity : enhancement
296 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
300 Frequency : rare, at shutdown
301 Description: access already free / zero obd_namespace.
302 Details : if client_disconnect_export was called without force flag set,
303 and exist connect request in flight, this can produce access to
304 NULL pointer (or already free pointer) when connect_interpret
305 store ocd flags in obd_namespace.
309 Frequency : only at startup
310 Description: not alloc memory with spinlock held.
311 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
312 if any spinlock held.
314 Severity : enhancement
316 Description: make lustre randomly fail allocating memory
317 Details : Make lustre randomly failed allocating memory for testing purpose.
319 Severity : enhancement
321 Description: lost problems with lov objid file
322 Details : Fixes some scability and access to not inited memory problems
323 in work with lov objdid file.
328 Description: lfs find does not continue on file error
329 Details : Continue other files processing when a file/dir is absent.
333 Description: Inconsistent usage of lustre_pack_reply()
334 Details : Standardize the usage of lustre_pack_reply() such that it
335 always generate a CERROR on failure.
338 Frequency : occasional
340 Description: MDS hang or stay in waiting lock
341 Details : If client receive lock with CBPENDING flag ldlm need send lock
342 cancel as separate rpc, to avoid situation when cancel request
343 can't processed due all i/o threads stay in wait lock.
347 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
351 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
355 Description: Update to SLES9 kernel-2.6.5-7.287.3.
359 Description: Update to RHEL4 latest kernel.
361 Severity : enhancement
363 Description: Build SLES10 patchless client fails
364 Details : The configure was broken by run ./configure with
365 --with-linux-obj=.... argument for patchless client. When the
366 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
367 can't search header adequately. Use absolute path such as
368 -I($LINUX)/include instead.
372 Description: interrupt oig_wait produce painc on resend.
373 Details : brw_redo_request can be used for resend requests from ptlrpcd and
374 private set, and this produce situation when rq_ptlrpcd_data not
375 copyed to new allocated request and triggered LBUG on assert
376 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
377 ptlrpcd set if request is changed and can be safety changed to use
382 Description: LASSERT_{REQ,REP}SWAB macros are buggy
383 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
384 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
385 these macros and replace them with their difinitions instead.
389 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
390 Details : Update which_patch & target file for SLES10 latest kernel.
393 Description: Remove client patches from SLES 10 kernel.
394 Details : This causes SLES 10 clients to behave as patchless clients
395 even on a Lustre-patched (server) kernel.
397 Severity : enhancement
399 Description: self-adjustable client's lru lists
400 Details : use adaptive algorithm for managing client cached locks lru
401 lists according to current server load, other client's work
402 pattern, memory activities, etc. Both, server and client
403 side namespaces provide number of proc tunables for controlling
406 Severity : enhancement
408 Description: light-weight GSS support
409 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
410 and reduce performance overhead.
412 Severity : enhancement
414 Description: Linux keyring support
415 Details : Support using service of Linux keyring for Lustre GSS internal
416 context refresh/cache mechanism.
420 Description: Fix errors in lfs documentation
421 Details : Fixes man pages
425 Description: don't use GFP_* in generic Lustre code.
426 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
430 Description: obdclass is limited by single OBD_ALLOC(idarray)
431 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
435 Description: warnings with build patchless client with vanila 2.6.19 and up
436 Details : change old ctl_table style and replace ctl_table/ctl_table_header
437 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
441 Frequency : Only for SLES
442 Description: sanity_quota fail test_1
443 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
444 file, which makes TSTID[2] inunique.
448 Frequency : Only for SLES9
449 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
450 Details : Update target/ChangeLog/which_patch .
454 Frequency : Only for test_10 in sanity-quota.sh
455 Description: a bug in quota test code
456 Details : checking a test flag in a wrong place
460 Frequency : Only for test_13 in sanity-quota.sh
461 Description: a bug in quota test script
462 Details : shouldn't check overfull allocated quota
466 Description: jbd statistics
467 Details : Port older jbd statistics patch for sles10
469 Severity : enhancement
471 Description: Add printing inode info into message about error in writepage.
476 Description: reply_lock_interpret crash due to race with it and lock cancel.
477 Details : Do not replay locks that are being cancelled. Do not reference
478 locks by their address during replay, just by their handle.
482 Frequency : When flocks are used.
483 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
484 Details : Flock locks might destroy just granted lock if it could be merged
485 with another existing flock, this is done in completion handler,
486 so teach ldlm_cli_enquque_fini that this is a valid case for
492 Description: Oops in read and write path when failing to allocate lock.
493 Details : Check if lock allocation failed and return error back.
497 Description: lstripe command fails for valid OST index
498 Details : The stripe offset is compared to lov->desc.ld_tgt_count
499 instead of lov->desc.ld_active_tgt_count.
503 Description: sanity.sh failed test 103
504 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
509 Description: df doesn't work properly if diskfs blocksize != 4K
510 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
514 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
515 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
519 Description: obd_health_check_timeout too short
520 Details : set obd_health_check_timeout as 1.5x of obd_timeout
524 Description: llapi_file_create() does not allow some changes
525 Details : add llapi_file_open() that allows specifying the mode and
526 open flags, and also returns an open file handle.
530 Description: merge and cleanup kernel patches.
531 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
535 Description: Add journal checksum support.(Kernel part)
536 Details : The journal checksum feature adds two new flags i.e
537 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
538 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
539 indicates that the commit block contains the checksum for
540 the blocks described by the descriptor blocks. Now commit
541 record can be sent to disk without waiting for descriptor
542 blocks to be written to disk. This behavior is controlled
543 using JBD2_FEATURE_ASYNC_COMMIT flag.
547 Description: OSS needs mutliple precreate threads
548 Details : Add ability to start more than one create thread per OSS.
552 Description: Sanity.sh test_65k failed: setstripe should have succeeded
553 Details : execute lfs setstripe on client
557 Description: mds_obd_create error creating tmp object
558 Details : When the user sets quota on root, llog will be affected and can't
559 create files and write files.
562 Frequency : Always on ia64 patchless client, and possibly others.
564 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
565 Details : This allows the patchless client to be loaded on architectures
570 Description: RedHat Update kernel for RHEL5
571 Details : Add the kernel config file for RHEL5.
575 Description: RedHat Update kernel for RHEL5
576 Details : Modify the kernel config file more closer RHEL5.
580 Description: Build failure against Centos5 (RHEL5)
581 Details : Define PAGE_SIZE when it isn't present.
585 Description: test_7 on sanity-quota.sh can't be used on separate machines
586 Details : add facet to handle it
590 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
591 Details : with NFS, the anon dentry's parent was set to itself in
592 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
593 corresponding dentry other than use rec->ur_name.
595 Severity : enhancement
597 Description: lfs setstripe enhancement
598 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
602 Description: enable data checksumming by default
603 Details : enable checksum by default, allow --disable-checksum
604 configure option and "-o nochecksum" mount option. Checksums
605 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
609 Description: lustre support for RHEL5
610 Details : Add support for RHEL5.
614 Description: Client eviction due to failover config
615 Details : after a connection loss, the lustre client should attempt to
616 reconnect to the last active server first before trying the
617 other potential connections.
621 Description: when mds and osts use different quota unit(32bit and 64bit),
622 quota will be released repeatly.
623 Details : void sending multiple quota reqs to mds, which will keep the
624 status between the reqs.
628 Description: osts not allocated evenly to files
629 Details : change the condition to increase offset_idx
632 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
634 Description: Data corruption for OSTs that are formatted larger than 2TB
636 Details : When generating the bio request for lustre file writes the
637 sector number would overflow a temporary variable before being
638 used for the IO. The data reads correctly from Lustre (which
639 will overflow in a similar manner) but other file data or
640 filesystem metadata may be corrupted in some cases.
644 Description: Tune the kernel for good SCSI performance.
645 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
646 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
651 Description: rewrite ext2-derived code in obdclass/uuid.c
652 Details : rewrite inherited code (uuid parsing code from ext2 utils)
653 from scratch preserving functionality.
657 Description: "lfs find -obd UUID" prints directories
658 Details : "lfs find -obd UUID" will return all directory names instead
659 of just file names. It is incorrect because the directories
660 do not reside on the OSTs.
664 Description: use i_size_read and i_size_write in 2.6 port
665 Details : replace inode->i_size access with i_size_read/write()
668 Frequency : when removing large files
670 Description: scheduling issue during removal of large Lustre files
671 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
672 It causes scheduling issues when removing large files (17TB in the
676 Frequency : only with liblustre clients on XT3
678 Description: evictions taking too long
679 Details : allow llrd to evict clients directly on OSTs
682 Frequency : only on ppc
684 Description: /proc/fs/lustre/devices broken on ppc
685 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
686 We should make sure the type of 'index' is loff_t and then cast
687 explicitly as needed below. Do not assign an explicitly cast
691 Frequency : only for rhel5
693 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
694 Details : Modify the target file & which_kernel.
696 Severity : enhancement
698 Description: omit set fsid for export NFS
699 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
700 as FS_REQUIRES_DEV for avoid problems with generate fsid.
704 Frequency : Always, for kernels after 2.6.16
705 Description: Fix warning idr_remove called for id=.. which is not allocated.
706 Details : Last kernels save old s_dev before kill super and not allow
707 to restore from callback - restore it before call kill_anon_super.
711 Description: buffer overruns could theoretically occur
712 Details : llapi_semantic_traverse() modifies the "path" argument by
713 appending values to the end of the origin string, and a
714 overrun may occur. Adding buffer overrun check in liblustreapi.
718 Description: Fix error on 'ls .' at the top of the Lustre mount.
719 Details : Don't revalidate dentry if it is a root dentry.
723 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
724 Details : Modify vm-tunables-rhel4.patch.
728 Description: Kernel config for 2.6.18-vanilla.
729 Details : Modify targets/2.6-vanilla.target.in.
730 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
731 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
732 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
733 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
737 Description: improve handling recoverable errors
738 Details : if request processig with error which can be recoverable on server
739 request should be resend, otherwise page released from cache and
744 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
745 Details : Modify target file & which_patch.
746 A flaw was found in the IA32 system call emulation provided
747 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
748 value could be stored in the %RAX register, which could trigger an
749 out-of-bounds system call table access. An untrusted local user
750 could exploit this flaw to run code in the kernel
751 (ie a root privilege escalation). (CVE-2007-4573).
755 Description: O_DIRECT bypasses client statistics.
756 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
757 read_ahead_stats were updated. copied the stats section from
758 osc_send_oap_rpc() into async_internal().
762 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
766 Description: change order of libsysio includes
767 Details : '#include sysio.h' should always come before '#include xtio.h'
771 Description: handle "serious error: objid * already exists" more gracefully
772 Details : If LAST_ID value on disk is smaller than the objects existing in
773 the O/0/d* directories, it indicates disk corruption and causes an
774 LBUG(). If the object is 0-length, then we should use the existing
775 object. This will help to avoid a full fsck in most cases.
777 Severity : enhancement
779 Description: adapt the lustre_config script to support the upgrade case
780 Details : Add "-u" option for lustre_config script to support upgrading 1.4
781 server targets to 1.6 in parallel.
785 Description: To avoid grant space > avaible space when the disk is almost
786 full. Without this patch you might see the error "grant XXXX >
787 available" or some LBUG about grant, when the disk is almost
789 Details : In filter_check_grant, for non_grant cache write, we should
790 check the left space by if (*left > ungranted + bytes), instead
791 of (*left > ungranted), because only we are sure the left space
792 is enough for another "bytes", then the ungrant space should be
793 increase. In client, we should update cl_avail_grant only there
794 is OBD_MD_FLGRANT in the reply.
798 Description: Update RHEL 4 kernel to fix local root privilege escalation.
799 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
800 described in CVE-2007-4573. This problem could allow untrusted
801 local users to gain root access.
804 Frequency : when using O_DIRECT and quotas
806 Description: Incorrect file ownership on O_DIRECT output files
807 Details : block usage reported by 'lfs quota' does not take into account
808 files that have been written with O_DIRECT.
813 Description: touch file failed when fs is not full
814 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
815 otherwise we can get ENOSP while fs is not full.
819 Description: parallel lock callbacks
820 Details : Instead of sending blocking and completion callbacks as separated
821 requests, adding them to a set and sending in parallel.
824 Frequency : only for Cray XT3
825 Bugzilla : 12829/13455
826 Description: Changing primary group doesn't change the group lustre assigns to
828 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
829 overridden with the primary group provided by the group upcall,
830 whereas the supplied fsgid can be trusted if it is in the list of
831 supplementary groups returned by the group upcall.
833 Severity : enhancement
835 Description: Allow masking D_WARNING, D_ERROR messages from console
836 Details : Console messages can now be disabled via lnet.printk.
840 Description: User code with malformed file open parameter crashes client node
841 Details : Before packing join_file req, all the related reference should be
842 checked carefully in case some malformed flags cause fake
843 join_file req on client.
847 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
848 Details : ldlm_completion_ast() assumes that a lock is granted when the req
849 mode is equal to the granted mode. However, it should also check
850 that LDLM_FL_CP_REQD is not set.
854 Description: Heavy nfs access might result in deadlocks
855 Details : After ELC code landed, it is now improper to enqueue any mds
856 locks under och_sem, because enqueue might want to decide to
857 cancel open locks for same inode we are holding och_sem for.
861 Description: Client eviction while running blogbench
862 Details : A lot of unlink operations with concurrent I/O can lead to a
863 deadlock causing evictions. To address the problem, the number of
864 oustanding OST_DESTROY requests is now throttled to
865 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
866 callbacks are processed in priority.
870 Description: enable ACLs on MDS by default
871 Details : ACLs must be enabled on MDS by default.
874 Frequency : PPC/PPC64 only
876 Description: conflicts between asm-ppc64/types.h and lustre_types.h
877 Details : fix duplicated definitions between asm-ppc64/types.h and
878 lustre_types.h on PPC.
881 Frequency : PPC/PPC64 only
883 Description: asm-ppc/segment.h does not exist
884 Details : fix compile issue on PPC.
888 Description: better handle error messages in extents code
891 Frequency : RHEL4 only
893 Description: mkfs is very slow on IA64/RHEL4
894 Details : A performance regression has been discovered in the MPT Fusion
895 driver between versions 3.02.73rh and 3.02.99.00rh. As a
896 consequence, we have downgraded the MPT Fusion driver in the RHEL4
897 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
899 Severity : enhancement
901 Description: SNMP support enhancement
902 Details : Adding total number of sampled request for an MDS node in snmp
905 Severity : enhancement
907 Description: Optimize ldlm waiting list processing for PR extent locks
908 Details : When processing waiting list for read extent lock and meeting read
909 lock that is same or wider to it that is not contended, skip
910 processing rest of the list and immediatelly return current
911 status of conflictness, since we are guaranteed there are no
912 conflicting locks in the rest of the list.
916 Description: Time out and refuse to reconnect
917 Details : When the failover node is the primary node, it is possible
918 to have two identical connections in imp_conn_list. We must
919 compare not conn's pointers but NIDs, otherwise we can defeat
920 connection throttling.
924 Description: port llog fixes from b1_6 into HEAD
925 Details : Port llog reference couting and some llog cleanups from b1_6
926 (bug 10800) into HEAD, for protect from panic and access to already
927 free llog structures.
931 Description: Detect stride IO mode in read-ahead
932 Details : When a client does stride read, read-ahead should detect that and
933 read-ahead pages according to the detected stride pattern.
937 Description: data checksumming impacts single node performance
938 Details : add support for several checksum algorithm. Currently, only CRC32
939 and Adler-32 are supported. The checksum type can be changed on
940 the fly via /proc/fs/lustre/osc/*/checksum_type.
944 Description: use adler32 for page checksums
945 Details : when available, use the Adler-32 algorithm instead of CRC32 for
950 Description: build for x2 fails
951 Details : fix compile issue on Cray systems.
955 Description: Properly match for duplicate locks
956 Details : Due to different lock order from skiplists code, we need to
957 traverse entire list for now
960 Frequency : only on PPC/SLES10
962 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
963 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
964 be defined. Add a hack in mkfs_lustre.c to work around this compile
969 Description: LASSERT on MDS when client holding flock lock dies
970 Details : ldlm pool logic depends on number of granted locks equal to
971 number of released locks which is not true for flock locks, so
972 just exclude such locks from consideration.
976 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
977 Details : Use fsfilt_sync() for both the whole filesystem sync and
978 individual file sync to eliminate dangerous inode locking
979 with I_LOCK that can lead to a deadlock.
983 Description: performance in 1.6.3
984 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
985 sure that 1MB requests are not fragmented by the block layer.
987 Severity : enhancement
989 Description: organize the server-side client stats on per-nid basis
990 Details : Change the structure of stats under obdfilter and mds to
998 The "uuid"s file would list the uuids of _active_ exports.
999 And the clear entry is to clear all stats and stale nids.
1001 Severity : enhancement
1003 Description: eliminate client locks in face of contention
1004 Details : file contention detection and lockless i/o implementation
1005 for contended files.
1009 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1012 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1014 Description: MDS may lose file striping (and hence file data) in some cases
1015 Details : If there are additional extended attributes stored on the MDS,
1016 in particular ACLs, SELinux, or user attributes (if user_xattr
1017 is specified for the client mount options) then there is a risk
1018 of attribute loss. Additionally, the Lustre file striping
1019 needs to be larger than default (e.g. striped over all OSTs),
1020 and an additional attribute must be stored initially in the
1021 inode and then increase in size enough to be moved to the
1022 external attribute block (e.g. ACL growing in size) for file
1025 --------------------------------------------------------------------------------
1027 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1029 * Support for kernels:
1030 2.6.5-7.283 (SLES 9),
1031 2.6.9-55.EL (RHEL 4),
1032 2.6.16.46-0.14 (SLES 10),
1033 2.6.18.8 vanilla (kernel.org)
1034 * Client support for unpatched kernels:
1035 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1036 2.6.16 - 2.6.22 vanilla (kernel.org)
1037 * Due to recently discovered recovery problems, we do not recommend
1038 using patchless RHEL 4 clients with this or any earlier release.
1039 * Recommended e2fsprogs version: 1.39.cfs8
1040 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1041 * Starting with this release, the ldiskfs backing filesystem required
1042 by Lustre is now in its own package, lustre-ldiskfs. This package
1043 should be installed. It is versioned separately from Lustre and
1044 may be released separately in future.
1048 Description: block reactivating mgc import until all deactivates complete
1049 Details : Fix race when failing back MDT/MGS to itself (testing)
1052 Frequency : at statup only
1054 Description: mds_lov_synchronize race leads to various problems
1055 Details : simultaneous MDT->OST connections at startup can cause the
1056 sync to abort, leaving the OSC in a bad state.
1058 Severity : enhancement
1060 Description: add optional extra BUILD_VERSION info
1061 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1062 to override the lustre version.
1065 Frequency : 2.6.18 servers only
1067 Description: ll_kern_mount() doesn't release the module reference
1068 Details : The ldiskfs module reference count never drops down to 0
1069 because ll_kern_mount() doesn't release the module reference.
1074 Description: server LBUG when using old ost_num_threads parameter
1075 Details : Accept the old ost_num_threads parameter but warn that it
1076 is deprecated, and fix an off-by-one error that caused an LBUG.
1081 Description: Transient SCSI error results in persistent IO issue
1082 Details : iobuf->dr_error is not reinitialized to 0 between two
1086 Frequency : sometimes when underlying device returns I/O errors
1088 Description: OSTs not going read-only during write failures
1089 Details : OSTs are not remounted read-only when the journal commit threads
1090 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1091 instead of the ext3 wrappers.
1095 Description: poor connect scaling with increasing client count
1096 Details : Don't run filter_grant_sanity_check for more than 100 exports
1097 to improve scaling for large numbers of clients.
1100 Frequency : SLES10 only
1102 Description: sanity-quota.sh quotacheck failed: rc = -22
1103 Details : Quotas cannot be enabled on SLES10.
1106 Frequency : liblustre clients only
1108 Description: getdirentries does not give error when run on compute nodes
1109 Details : getdirentries does not fail when the size specified as an argument
1110 is too small to contain at least one entry
1112 Severity : enhancement
1114 Description: Add LNET router traceability for debug purposes
1115 Details : If a checksum failure occurs with a router as part of the
1116 IO path, the NID of the last router that forwarded the bulk data
1117 is printed so it can be identified.
1122 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1123 Details : Due to a race condition, liblustre clients were occasionally
1124 evicted incorrectly.
1126 Severity : enhancement
1128 Description: lfs setstripe use optional parameters instead of postional
1131 Severity : enhancement
1133 Description: Nanosecond timestamp support for ldiskfs
1134 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1135 resolution timestamps. There is not yet support for this at
1136 the Lustre filesystem level.
1139 Frequency : during server recovery
1141 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1142 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1146 Frequency : nfs export on patchless client
1148 Description: connectathon hang when test nfs export over patchless client
1149 Details : Disconnected dentry cannot be found with lookup, so we do not need
1150 to unhash it or make it invalid
1153 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1154 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1155 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1160 Description: open req refcounting wrong on reconnect
1161 Details : If reconnect happened between getting open reply from server and
1162 call to mdc_set_replay_data in ll_file_open, we will schedule
1163 replay for unreferenced request that we are about to free.
1164 Subsequent close will crash in variety of ways.
1165 Check that request is still eligible for replay in
1166 mdc_set_replay_data().
1171 Description: disable writes to filesystem when reading health_check file
1172 Details : the default for reading the health_check proc file has changed
1173 to NOT do a journal transaction and write to disk, because this
1174 can cause reads of the /proc file to hang and block HA state
1175 checking on a healthy but otherwise heavily loaded system. It
1176 is possible to return to the previous behaviour during configure
1177 with --enable-health-write.
1179 Severity : enhancement
1181 Description: 64-bit inode version
1182 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1183 to the inode. This will be required for version-based recovery.
1188 Description: MDS fails to start if a duplicate client export is detected
1189 Details : in some rare cases it was possible for a client to connect to
1190 an MDS multiple times. Upon recovery the MDS would detect this
1191 and fail during startup. Handle this more gracefully.
1193 Severity : enhancement
1195 Description: Add -o localflock option to simulate old noflock
1197 Details : This will achieve local-only flock/fcntl locks
1203 Description: log_commit_thread vs filter_destroy race leads to crash
1204 Details : Take import reference before releasing llog record semaphore
1209 Description: Wrong request locking in request set processing
1210 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1211 imp_delayed_list, in this place should be used imp_lock.
1214 Frequency : when reconnection
1216 Description: Grant Leak when osc reconnect to OST
1217 Details : When osc reconnect ost, OST(filter) should check whether it
1218 should grant more space to client by comparing fed_grant and
1219 cl_avail_grant, and return the granted space to client instead
1220 of "new granted" space, because client will call osc_init_grant
1221 to update the client grant space info.
1224 Frequency : when client reconnect to OST
1226 Description: Grant Leak when osc do resend and replay bulk write
1227 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1228 bulk write request, because the grant info while be sync between
1229 OSC and OST when reconnect, and we should ignore the grant info
1230 these of resend/replay write req.
1235 Description: Grant space more than avaiable left space sometimes.
1236 Details : When then OST is about to be full, if two bulk writing from
1237 different clients came to OST. Accord the avaliable space of the
1238 OST, the first req should be permitted, and the second one
1239 should be denied by ENOSPC. But if the seconde arrived before
1240 the first one is commited. The OST might wrongly permit second
1241 writing, which will cause grant space > avaiable space.
1244 Frequency : when client is evicted
1246 Description: Grant might be wrongly erased when osc is evicted by OST
1247 Details : when the import is evicted by server, it will fork another
1248 thread ptlrpc_invalidate_import_thread to invalidate the
1249 import, where the grant will be set to 0. While the original
1250 thread will update the grant it got when connecting. So if
1251 the former happened latter, the grant will be wrongly errased
1252 because of this race.
1257 Description: Checking Stale with correct fid
1258 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1259 to check whether it is stale, because sometimes, we want the
1260 enqueue happened anyway, and op_data.fid2 will not be initialized.
1262 Severity : enhancement
1264 Description: update patchless client
1265 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1268 Frequency : only with 2.4 kernel
1270 Description: random memory corruption
1271 Details : size of struct ll_inode_info is to big for union inode.u and this
1272 can be cause of random memory corruption.
1277 Description: Memory leak in recovery
1278 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1279 It should also check obd_fail before fsfilt_start, otherwise if
1280 fsfilt_start return -EROFS,(failover mds during mds recovery).
1281 then the req will return with repmsg->transno = 0 and rc = EROFS.
1282 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1283 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1284 be freed no matter whether fsfilt_commit success or not.
1287 Frequency : only with huge count clients
1289 Description: Prevents from taking the superblock lock in llap_from_page for
1291 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1292 for llap_from_page call in ll_removepage prevents from taking the
1293 superblock lock for a soon died page.
1298 Description: Not check open intent error before release open handle
1299 Details : in some rare cases, the open intent error is not checked before
1300 release open handle, which may cause
1301 ASSERTION(open_req->rq_transno != 0), because it tries to release
1302 the failed open handle.
1307 Description: Set cat log bitmap only after create log success.
1308 Details : in some rare cases, the cat log bitmap is set too early. and it
1309 should be set only after create log success.
1313 Description: Accessing a block bevice can re-enable I/O when Lustre is
1314 tearing down a device.
1315 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1320 Description: service threads may hog cpus when there are a lot of requests
1321 Details : Insert cond_resched to give other threads a chance to use some CPU
1326 Description: the cat log was not initialized in recovery
1327 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1328 unlink log on mds will not be initialized until mds post
1329 recovery. And also in mds post recovery, the unlink log will
1330 initialization will be done asynchronausly, so there will be race
1331 between add unlink log and unlink log initialization.
1335 Description: brw_stats were being printed incorrectly
1336 Details : brw_stats were being printed as log2 but all of them were not
1337 recorded as log2. Also remove some code duplication arising from
1338 filter_tally_{read,write}.
1342 Frequency : rare, only in recovery.
1343 Description: ASSERTION(req->rq_type != LI_POISON) failed
1344 Details : imp_lock should be held while iterating over imp_sending_list for
1345 prevent destroy request after get timeout in ptlrpc_queue_wait.
1349 Description: replay-single.sh test 52 fails
1350 Details : A lock's skiplist need to be cleanup when it being unlinked
1351 from its resource list.
1355 Description: Short directio read returns full requested size rather than
1357 Details : Direct I/O operations should return actual amount of bytes
1358 transferred rather than requested size.
1360 Severity : enhancement
1362 Description: metadata RPC reduction (e.g. for rm performance)
1363 Details : decrease the amount of synchronous RPC between clients and servers
1364 by canceling conflicing lock before the operation on the client
1365 and packing thier handles into the main operation RPC to server.
1367 Severity : enhancement
1369 Description: Async OSC create to avoid the blocking unnecessarily.
1370 Details : If a OST has no remain object, system will block on the creating
1371 when need to create a new object on this OST. Now, ways use
1372 pre-created objects when available, instead of blocking on an
1373 empty osc while others are not empty. If we must block, we block
1374 for the shortest possible period of time.
1376 Severity : enhancement
1378 Description: refine locking for avoid write wrong info into lov_objid file
1379 Details : fix possible races with add new target and write/update data in
1382 --------------------------------------------------------------------------------
1384 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1389 Frequency : on some architectures
1391 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1392 Details : Uninitialized flags sometimes cause configuration commands to
1396 Frequency : patchless clients only
1398 Description: missing __iget() symbol export
1399 Details : The __iget() symbol export is missing. To avoid the need for
1400 this on patchless clients the deathrow inode reaper is turned
1401 off, and we depend on the VM to clean up old inodes. This
1402 dependency was during via the fix for bug 12181.
1404 --------------------------------------------------------------------------------
1406 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1408 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1409 INTEROPERATE with older versions automatically. In many cases a
1410 special upgrade step is needed. Please read the
1411 user documentation before upgrading any part of a live system.
1412 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1413 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1414 * WARNING: Lustre configuration and startup changes are required with
1415 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1421 Severity : enhancement
1423 Description: MountConf
1424 Details : Lustre configuration is now managed via mkfs and mount
1425 commands instead of lmc and lconf. New obd types (MGS, MGC)
1426 are added for dynamic configuration management. See
1427 https://mail.clusterfs.com/wikis/lustre/MountConf for
1430 Severity : enhancement
1432 Description: dynamic OST addition
1433 Details : OSTs can now be added to a live filesystem
1435 Severity : enhancement
1437 Description: startup order invariance
1438 Details : MDTs and OSTs can be started in any order. Clients only
1439 require the MDT to complete startup.
1441 Severity : enhancement
1443 Description: parallel, asynchronous orphan cleanup
1444 Details : orphan cleanup is now performed in separate threads for each
1445 OST, allowing parallel non-blocking operation.
1447 Severity : enhancement
1449 Description: optimized stripe assignment
1450 Details : stripe assignments are now made based on ost space available,
1451 ost previous usage, and OSS previous usage, in order to try
1452 to optimize storage space and networking resources.
1454 Severity : enhancement
1456 Description: Permanently set tunables
1457 Details : All writable /proc/fs/lustre tunables can now be permanently
1458 set on a per-server basis, at mkfs time or on a live system.
1460 Severity : enhancement
1462 Description: Lustre message v2
1463 Details : Add lustre message format v2.
1465 Severity : enhancement
1467 Description: client OST exclusion list
1468 Details : Clients can be started with a list of OSTs that should be
1469 declared "inactive" for known non-responsive OSTs.
1473 Description: SPEC SFS validation failure on NFS v2 over lustre.
1474 Details : Changes the blocksize for regular files to be 2x RPC size,
1475 and not depend on stripe size.
1477 Severity : enhancement
1479 Description: Multiple MD RPCs in flight.
1480 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1481 To avoid overly-overloading MDS, introduce a limit on number of
1482 MDS RPCs in flight for a single client and add /proc controls
1483 to adjust this limit.
1485 Severity : enhancement
1487 Description: client read/write statistics
1488 Details : Add client read/write call usage stats for performance
1489 analysis of user processes.
1490 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1491 file access. extents_stats shows chunk size distribution.
1492 extents_stats_per_process show chunk size distribution per
1495 Severity : enhancement
1497 Description: mds statistics
1498 Details : Add detailed mds operations statistics in
1499 /proc/fs/lustre/mds/*/stats.
1503 Description: Failure of copying files with lustre special EAs.
1504 Details : Client side always return success for setxattr call for lustre
1505 special xattr (currently only "trusted.lov").
1510 Description: Refcount LNET uuids
1511 Details : The global LNET uuid list grew linearly with every startup;
1512 refcount repeated list entries instead of always adding to
1515 Severity : enhancement
1517 Description: Dynamic service threads
1518 Details : Within a small range, start extra service threads
1519 automatically when the request queue builds up.
1522 Frequency : mixed-endian client/server environments
1524 Description: mixed-endian crashes
1525 Details : The new msg_v2 system had some failures in mixed-endian
1528 Severity : enhancement
1530 Description: Easy OST removal
1531 Details : OSTs can be permanently deactivated with e.g. 'lctl
1532 conf_param lustre-OST0001.osc.active=0'
1534 Severity : enhancement
1536 Description: MGS proc entries
1537 Details : Added basic proc entries for the MGS showing what filesystems
1540 Severity : enhancement
1542 Description: provide MGS failover
1543 Details : Added config lock reacquisition after MGS server failover.
1545 Severity : enhancement
1547 Description: add Linux 2.4 support
1548 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1552 Description: a large application tries to do I/O to the same resource and dies
1553 in the middle of it.
1554 Details : Check the req->rq_arrival time after the call to
1555 ost_brw_lock_get(), but before we do anything about
1556 processing it & sending the BULK transfer request. This
1557 should help move old stale pending locks off the queue as
1558 quickly as obd_timeout.
1561 Frequency : when an incorrect nid is specified during startup
1563 Description: ptlrpc connect to non-existant node causes kernel crash
1564 Details : LNET can't be re-entered from an event callback, which
1565 happened when we expire a message after the export has been
1566 cleaned up. Instead, hand the zombie cleanup off to another
1569 Severity : enhancement
1571 Description: plain/inodebits lock performance improvement
1572 Details : Grouping plain/inodebits in granted list by their request modes
1573 and bits policy, thus improving the performance of search through
1577 Frequency : only if OST filesystem is corrupted
1579 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1580 Details : for a short time RPCs with bulk IO are in the replay list,
1581 but replay of bulk IOs is unimplemented. If the OST filesystem
1582 is corrupted due to disk cache incoherency and then replay is
1583 started it is possible to trip an assertion. Avoid putting
1584 committed RPCs into the replay list at all to avoid this issue.
1587 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1589 Description: System hang on startup
1590 Details : This bug allowed the liblustre (e.g. catamount) client to
1591 return to the app before handling all startup RPCs. This
1592 could leave the node unresponsive to lustre network traffic
1593 and manifested as a server ptllnd timeout.
1595 Severity : enhancement
1597 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1598 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1599 boolean to control whether to print peer debug info when a
1600 client's RPC times out.
1603 Frequency : only for kernels with patches from Lustre below 1.4.3
1605 Description: Remove old rdonly API
1606 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1609 Frequency : only for devices with external journals
1611 Description: Set external device read-only also
1612 Details : During a commanded failover stop, we set the disk device
1613 read-only while the server shuts down. We now also set any
1614 external journal device read-only at the same time.
1617 Frequency : when upgrading from 1.4 while trying to change parameters
1619 Description: The wrong (new) MDC name was used when setting parameters for
1620 upgraded MDT's. Also allows changing of OSC (and MDC)
1621 parameters if --writeconf is specified at tunefs upgrade time.
1624 Frequency : when setting specific ost indicies
1626 Description: QOS code breaks on skipped indicies
1627 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1628 created with --index need not be sequential.
1630 Severity : enhancement
1632 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1633 Details : The uninit_groups feature works in conjunction with the kernel
1634 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1635 up the pass1 processing of e2fsck. This is a read-only feature
1636 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1637 mount filesystems that have had this feature enabled.
1639 Severity : enhancement
1641 Description: Improve multi-block allocation algorithm to avoid fragmentation
1642 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1643 allocation locality and avoid filesystem fragmentation.
1645 ------------------------------------------------------------------------------
1647 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1649 * Support for kernels:
1650 2.4.21-47.0.1.EL (RHEL 3)
1651 2.6.5-7.283 (SLES 9)
1652 2.6.9-42.0.10.EL (RHEL 4)
1653 2.6.12.6 vanilla (kernel.org)
1654 2.6.16.27-0.9 (SLES 10)
1655 * Recommended e2fsprogs version: 1.39.cfs5
1657 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1661 Frequency : occasional, depends on client load and configuration
1662 Bugzilla : 12181, 12203
1663 Description: data loss for recently-modified files
1665 Details : In some cases it is possible that recently written or created
1666 files may not be written to disk in a timely manner (this should
1667 normally be within 30s unless client IO load is very high).
1668 The problem appears as zero-length files or files that are a
1669 multiple of 1MB in size after a client crash or client eviction
1670 that are missing data at the end of the file.
1672 This problem is more likely to be hit on clients where files are
1673 repeatedly created and unlinked in the same directory, clients
1674 have a large amount of RAM, have many CPUs, the filesystem has
1675 many OSTs, the clients are rebooted frequently, and/or the files
1676 are not accessed by other nodes after being written.
1678 The presence of the problem can be detected by looking at
1679 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1680 smaller than the second (nr_unused) then dirty files will not
1681 be flushed automatically to disk. "sync; sleep 10" should be
1682 run several times on the node before unmounting it to update
1683 Lustre (this is also safe to run on nodes without this problem).
1685 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1686 that can cause this same problem, so customers using that kernel
1687 also need to update the kernel in addition to Lustre. In order
1688 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1690 It is normal that files written just before a client crash (less
1691 than 30s) may not yet have been flushed to disk, even for local
1695 Frequency : frequent on thin XT3 nodes
1697 Description: UUID collision on thin XT3 Linux nodes
1698 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1699 randomly, since we relied on an insufficiently-seeded PRNG.
1704 Description: OSS hangs after "All ost request buffers busy"
1705 Details : A deadlock between quota and journal operations caused OSS
1706 hangs after printing "All ost request buffers busy."
1709 Frequency : always on liblustre builds
1711 Description: Cleanup compiler warnings on liblustre
1714 Frequency : always on liblustre builds on XT3
1716 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1720 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1721 > 32000 subdirectories
1722 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1723 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1724 subdirectory count crosses 32000. This will aid e2fsck to
1725 correctly handle more than 32000 subdirectories.
1728 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1730 Description: System hang on startup
1731 Details : This bug allowed the liblustre (e.g. catamount) client to
1732 return to the app before handling all startup RPCs. This
1733 could leave the node unresponsive to lustre network traffic
1734 and manifested as a server ptllnd timeout.
1736 Severity : enhancement
1738 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1739 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1740 boolean to control whether to print peer debug info when a
1741 client's RPC times out.
1746 Description: make O_SYNC working on 2.6 kernels
1747 Details : 2.6 kernels use different method for mark pages for write,
1748 so need add a code to lustre for O_SYNC work.
1753 Description: Failure to close file and release space on NFS
1754 Details : Put inode details into lock acquired in ll_intent_file_open.
1755 Use mdc_intent_lock in ll_intent_open to properly
1756 detect all kind of errors unhandled by mdc_enqueue.
1761 Description: proc file read during shutdown sometimes raced obd removal,
1763 Details : Add lock to prevent obd access after proc file removal.
1766 Frequency : Only for files larger than 4GB on 32-bit clients.
1768 Description: improperly doing page alignment of locks
1769 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1770 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1775 Description: RPCs being resent when they shouldn't be
1776 Details : Some RPCs that should not be resent are being resent. This
1777 can cause inconsistencies in the RPC state machine. Do not
1778 resend such requests.
1781 Frequency : rare, only with NFS export
1783 Description: Crash on NFS re-export node
1784 Details : under very unusual load conditions an assertion is hit in
1785 ll_intent_file_open()
1788 Frequency : only if OST filesystem is corrupted
1790 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1791 Details : for a short time RPCs with bulk IO are in the replay list,
1792 but replay of bulk IOs is unimplemented. If the OST filesystem
1793 is corrupted due to disk cache incoherency and then replay is
1794 started it is possible to trip an assertion. Avoid putting
1795 committed RPCs into the replay list at all to avoid this issue.
1800 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1801 Details : Large single O_DIRECT read and write calls can fail to allocate
1802 a sufficiently large buffer to process the request. In case of
1803 allocation failure the allocation is retried with a smaller
1804 buffer and broken into smaller requests.
1806 Severity : enhancement
1808 Description: Add -o localflock option to simulate old noflock behaviour.
1809 Details : This will achieve local-only flock/fcntl locks coherentness.
1814 Description: versioning check is incomplete
1815 Details : Checking the version difference of client vs. server, report
1816 error if the gap is too big.
1821 Description: add support PG_writeback bit
1822 Details : add support for PG_writeback bit for Lustre, for more carefull
1823 work with page cache in 2.6 kernel. This also fix some deadlocks
1824 and remove hack for work O_SYNC with 2.6 kernel.
1826 Severity : enhancement
1828 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1829 Details : The uninit_groups feature works in conjunction with the kernel
1830 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1831 up the pass1 processing of e2fsck. This is a read-only feature
1832 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1833 mount filesystems that have had this feature enabled.
1835 Severity : enhancement
1837 Description: Improve multi-block allocation algorithm to avoid fragmentation
1838 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1839 allocation locality and avoid filesystem fragmentation.
1841 ------------------------------------------------------------------------------
1843 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1845 * Support for kernels:
1846 2.6.9-42.0.3.EL (RHEL 4)
1847 2.6.5-7.276 (SLES 9)
1848 2.4.21-47.0.1.EL (RHEL 3)
1849 2.6.12.6 vanilla (kernel.org)
1850 2.6.16.21-0.8 (SLES10)
1851 * Recommended e2fsprogs version: 1.39.cfs2-0
1853 * The backwards-compatible /proc/sys/portals symlink has been removed
1854 in this release. Before upgrading, please ensure that you change
1855 any configuration scripts or /etc/sysctl.conf files that access
1856 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1857 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1858 in advance of the upgrade on any system running Lustre 1.4.6 or
1859 newer, since /proc/sys/lnet was added in that version.
1860 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1864 Frequency : only when quota is used
1866 Description: avoid scanning export list for quota master
1867 Details : Change the algorithms to avoid scanning export list in order
1868 to improve the efficiency.
1871 Frequency : MDS failover only, very rarely
1873 Description: "went back in time" messages on mds failover
1874 Details : The greatest transno may be lost when the current operation
1875 finishes with an error (transno==0) and the client's last_rcvd
1876 record is over-written. Save the greatest transno in the
1877 mds_last_transno for this case.
1880 Frequency : always for specific kernels and striping counts
1882 Description: client may get "Matching packet too big" without ACL support
1883 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1884 when trying to access files in certain configurations. The
1885 clients should in fact be denied when mounting because they do
1886 not understand ACLs.
1889 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1891 Description: many clients connecting with IO in progress causes connect timeouts
1892 Details : Avoid synchronous journal commits to avoid delays caused by many
1893 clients connecting/disconnecting when bulk IO is in progress.
1894 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1895 OST_IO_PORTAL to avoid delays behind potentially many pending
1899 Frequency : occasionally with multiple writers to a single file
1901 Description: shared writes to file may result in wrong size reported by stat()
1902 Details : Allow growing of kms when extent lock is cancelled
1905 Frequency : always with random mmap IO to multi-striped file
1907 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1908 Details : Only if the hole is at the end of OST object so that kms is too
1909 small. Fix is to increase kms accordingly in ll_nopage.
1912 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1914 Description: writes to a missing object would leak memory on the OST
1915 Details : If there is an inconsistency between the MDS and OST filesystems,
1916 such that the MDS references an object that doesn't exist, writes
1917 to that object will leak memory due to incorrect cleanup in the
1918 error handling path, eventually running out of memory on the OST.
1923 Description: Creating too long symlink causes lustre errors
1924 Details : Check symlink and name lengths before sending requests to MDS.
1927 Frequency : only if flock is enabled (not on by default)
1929 Description: posix locks not released on fd closure on 2.6.9+
1930 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1931 kernels, this caused such locks not to be released on fd close and
1932 then assertions on fs unmount about still used locks.
1935 Frequency : MDS failover only, very rarely
1937 Description: clients may get ASSERTION(granted_lock != NULL)
1938 Details : When request was taking a long time, and a client was resending
1939 a getattr by name lock request. The were multiple lock requests
1940 with the same client lock handle and
1941 mds_getattr_name->fixup_handle_for_resent_request found one of the
1942 lock handles but later failed with ASSERTION(granted_lock != NULL).
1947 Description: handle->h_buffer_credits > 0, assertion failure
1948 Details : h_buffer_credits is zero after truncate, causing assertion
1949 failure. This patch extends the transaction or creates a new
1953 Frequency : NFS re-export or patchless client
1954 Bugzilla : 11179, 10796
1955 Description: Crash on NFS re-export node (__d_move)
1956 Details : We do not want to hash the dentry if we don't have a lock.
1957 But if this dentry is later used in d_move, we'd hit uninitialised
1958 list head d_hash, so we just do this to init d_hash field but
1959 leave dentry unhashed.
1962 Frequency : NFS re-export or patchless client
1964 Description: NFS exports has problem with symbolic link
1965 Details : lustre client didn't properly install dentry when re-exported
1966 to NFS or running patchless client.
1969 Frequency : NFS re-export or patchless client
1971 Description: Various nfs/patchless fixes.
1972 Details : fixes reuse disconected alias for lookup process - this fixes
1973 warning "find_exported_dentry: npd != pd",
1974 fix permission error with open files at nfs.
1975 fix apply umask when do revalidate.
1978 Frequency : occasional
1980 Description: Crash on NFS re-export node
1981 Details : calling clear_page() on the wrong pointer triggered oops in
1982 generic_mapping_read().
1985 Frequency : rarely, using O_DIRECT IO
1987 Description: unaligned directio crashes client with LASSERT
1988 Details : check for unaligned buffers before trying any requests.
1991 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1993 Description: stale data returned from RAID cache
1994 Details : If only a small amount of IO is done to the RAID device before
1995 reading it again it is possible to get stale data from the RAID
1996 cache instead of reading it from disk.
1999 Frequency : always for sles10 kernel
2001 Description: sles10 support
2002 Details : ll_follow_link: compile fixes and using of nd_set_link
2003 under newer kernels.
2006 Frequency : depends on arch, kernel and compiler version, always on sles10
2009 Description: recursive or deep enough symlinks cause stack overflow
2010 Details : getting rid of large stack-allocated variable in
2014 Frequency : depends on hardware
2016 Description: lustre write performance loss in the SLES10 kernel
2017 Details : the performance loss is caused by using of write barriers in the
2018 ext3 code. The SLES10 kernel turns barrier support on by
2019 default. The fix is to undo that change for ldiskfs.
2021 ------------------------------------------------------------------------------
2023 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2025 * Support for kernels:
2026 2.6.9-42.0.3EL (RHEL 4)
2027 2.6.5-7.276 (SLES 9)
2028 2.4.21-47.0.1.EL (RHEL 3)
2029 2.6.12.6 vanilla (kernel.org)
2033 Frequency : quota enabled and large files being deleted
2035 Description: releasing more than 4GB of quota at once hangs OST
2036 Details : If a user deletes more than 4GB of files on a single OST it
2037 will cause the OST to spin in an infinite loop. Release
2038 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2043 Description: statfs data retrieved from /proc may be stale or zero
2044 Details : When reading per-device statfs data from /proc, in the
2045 {kbytes,files}_{total,free,avail} files, it may appear
2046 as zero or be out of date.
2049 Frequency : systems with MD RAID1 external journal devices
2051 Description: lconf's call to blkid is confused by RAID1 journal devices
2052 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2053 of returning all block devices that match the journal UUID.
2056 Frequency : always, for aggregate stripe size over 4GB
2058 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2059 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2060 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2063 Frequency : always if "lfs find" used on a local file/directory
2065 Description: "lfs find" segfaults if used on a local file/directory
2066 Details : The case where a directory component was not specified wasn't
2067 handled correctly. Handle this properly.
2070 Frequency : always on ppc64
2072 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2073 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2074 perform endianness conversion.
2077 Frequency : rarely (truncate to non-zero file size after write under load)
2078 Bugzilla : 10730, 10687
2079 Description: Files padded with zeros to next 4K multiple
2080 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2081 it is possible that files that are truncated to a non-zero size
2082 immediately after being written are filled with zero bytes beyond
2083 the truncated size. No file data is lost.
2085 Severity : enhancement
2087 Description: Allow recovery/failover for liblustre clients.
2088 Details : liblustre clients were unaware of failover configurations until
2091 Severity : enhancement
2093 Description: user file locks should fail when not mounting with flock option
2094 Details : Set up an error-returning stub in ll_file_operations.lock field
2095 to prevent incorrect behaviour when client is mounted without
2096 flock option. Also, set up properly f_op->flock field for
2100 Frequency : always on ia64
2102 Description: "lfs df" loops on printing out MDS statfs information
2103 Details : The obd_ioctl_data was not initialized and in some systems
2104 this caused a failure during the ioctl that did not return
2105 an error. Initialize the struct and return an error on failure.
2108 Frequency : SLES 9 only
2110 Description: Error of copying files with lustre special EAs as root
2111 Details : Client side always return success for setxattr call for lustre
2112 special xattr (currently only "trusted.lov").
2115 Frequency : rarely on clusters with both ia64+i386 clients
2117 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2118 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2119 clients are doing concurrent writes to the same file it is
2120 possible that the ia64 clients may LASSERT because the OST
2121 extent locks are not PAGE_SIZE aligned. Ensure that grown
2122 locks are always aligned on the request boundary.
2125 Frequency : specific use, occasional
2127 Description: Overwriting in use executable truncates on-disk binary image
2128 Details : If one node attempts to overwrite an executable in use by
2129 another node, we now correctly return ETXTBSY instead of
2130 truncating the file.
2132 Severity : enhancement
2134 Description: Async OSC create to avoid the blocking unnecessarily.
2135 Details : If a OST has no remain object, system will block on the creating
2136 when need to create a new object on this OST. Now, ways use
2137 pre-created objects when available, instead of blocking on an
2138 empty osc while others are not empty. If we must block, we block
2139 for the shortest possible period of time.
2144 Description: chmod on Lustre root is propagated to other clients
2145 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2146 invalid by the follow_mount time.
2149 Frequency : liblustre clients only
2151 Description: Race in 'instant cancel' lock handling could lead to such locks
2152 never to be granted in case of SMP MDS
2153 Details : Do not destroy not yet granted but cbpending locks in
2157 Frequency : replay/resend of open
2159 Description: non null lock assetion failure in mds_intent_policy
2160 Details : Trying to replay/resend lockless open requests resulted in
2161 mds_open() returning 0 with no lock. Now it sets a flag if
2162 a lock is going to be returned.
2164 Severity : enhancement
2166 Description: Checksum enhancements
2167 Details : New checksum enhancements allow for resending RPCs that failed
2170 Severity : enhancement
2172 Description: Tunables on number of dirty pages in cacche
2173 Details : Allow to set limit on number of dirty pages cached.
2178 Description: client crash on unmount - lock still has references
2179 Details : In some error handling cases it was possible to leak a lock
2180 reference on a client while accessing a file. This was not
2181 harmful to the client during operation, but would cause the
2182 client to crash when the filesystem is unmounted.
2185 Frequency : specific case, rare
2187 Description: ETXTBSY on mds though file not in use by client
2188 Details : ETXTBSY is no longer incorrectly returned when attempting to
2189 chmod or chown a directory that the user previously tried to
2190 execute or a currently-executing binary.
2193 Frequency : extremely rare except on liblustre-based clients
2195 Description: Lustre space not freed when files are deleted
2196 Details : Clean up open-unlinked files after client eviction. Previously
2197 the unlink was skipped and the files remained as orphans.
2202 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2203 Details : In some cases it was possible to send a blocking callback to
2204 a client doing a glimpse, even though that client didn't get
2205 a lock granted. When the glimpse lock is cancelled on the OST
2206 the freed lock is left on the waiting list and corrupted the list.
2209 Frequency : all core dumps
2211 Description: Broke core dumps to lustre
2212 Details : Negative dentry may be unhashed if parent does not have UPDATE
2213 lock, but some callers, e.g. do_coredump, expect dentry to be
2214 hashed after successful create, hash it in ll_create_it.
2216 ------------------------------------------------------------------------------
2218 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2220 * Support for kernels:
2221 2.6.9-42.0.2.EL (RHEL 4)
2222 2.6.5-7.276 (SLES 9)
2223 2.4.21-40.EL (RHEL 3)
2224 2.6.12.6 vanilla (kernel.org)
2228 Frequency : always on RHEL 3
2230 Description: Number of open files grows over time
2231 Details : The number of open files grows over time, whether or not
2232 Lustre is started. This was due to a filp leak introduced
2233 by one of our kernel patches.
2235 ------------------------------------------------------------------------------
2237 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2239 * Support for kernels:
2240 2.6.9-42.EL (RHEL 4)
2241 2.6.5-7.276 (SLES 9)
2242 2.4.21-40.EL (RHEL 3)
2243 2.6.12.6 vanilla (kernel.org)
2248 Bugzilla : 5719, 9635, 9792, 9684
2249 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2250 Details : If a server is under heavy load and cannot reply to new
2251 connection requests before the client resends the (re)connect,
2252 the connection handling code can behave badly if two service
2253 threads are concurrently handing separate (re)connections from
2254 the same client. Add better locking to the connection handling
2255 code, and ensure that only a single connection will be processed
2256 for a given client UUID, even if the lock is dropped.
2258 Severity : enhancement
2260 Description: add TCP zero-copy support to kernel
2261 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2262 sends if the hardware supports scatter-gather and checksumming.
2263 This allows socklnd to do client-write and server-read more
2264 efficiently and reduce CPU utilization from skbuf copying.
2267 Frequency : only if NFS exporting from client
2269 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2270 Details : If setting attributes on a file created under NFS that had
2271 never been opened it would be possible to oops the client
2272 if the file had no objects.
2275 Frequency : always for liblustre
2277 Description: liblustre client does MDS+OSTs setattr RPC for each write
2278 Details : When doing a write from a liblustre client, the client
2279 incorrectly issued an RPC to the MDS and each OST the file was
2280 striped over in order to update the timestamps. When writing
2281 with small chunks and many clients this could overwhelm the MDS
2282 with RPCs. In all cases it would slow down the write because
2283 these RPCs are unnecessary.
2285 Severity : enhancement
2287 Description: allow number of MDS service threads to be changed at module load
2288 Details : It is now possible to change the number of MDS service threads
2289 running. Adding "options mds mds_num_threads={N}" to the MDS's
2290 /etc/modprobe.conf will set the number of threads for the next
2291 time Lustre is restarted (assuming the "mds" module is also
2292 reloaded at that time). The default number of threads will
2293 stay the same, 32 for most systems.
2298 Description: OST crash if filesystem is unformatted or corrupt
2299 Details : If an OST is started on a device that has never been formatted
2300 or if the filesystem is corrupt and cannot even mount then the
2301 error handling cleanup routines would dereference a NULL pointer.
2306 Description: NULL pointer deref in llap_from_page.
2307 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2308 incorrect) mapping if the page was truncated/reclaimed while it was
2309 searched for. Check for this condition and skip such pages when
2310 doing readahead. Introduce extra check to llap_from_page() to
2311 verify page->mapping->host is non-NULL (so page is not anonymous).
2314 Frequency : Sometimes when using sys_sendfile
2316 Description: "page not covered by a lock" warnings from ll_readpage
2317 Details : sendfile called ll_readpage without right page locks present.
2318 Now we introduced ll_file_sendfile that does necessary locking
2319 around call to generic_file_sendfile() much like we do in
2323 Frequency : with certain MDS communication failures at client mount time
2325 Description: NULL pointer deref after failed client mount
2326 Details : a client connection request may delayed by the network layer
2327 and not be sent until after the PTLRPC layer has timed out the
2328 request. If the client fails the mount immediately it will try
2329 to clean up before the network times out the request. Add a
2330 reference from the request import to the obd device and delay
2331 the cleanup until the network drops the request.
2334 Frequency : occasionally during client (re)connect
2336 Description: assertion failure during client (re)connect
2337 Details : processing a client connection request may be delayed by the
2338 client or server longer than the client connect timeout. This
2339 causes the client to resend the connection request. If the
2340 original connection request is replied in this interval, the
2341 client may trip an assertion failure in ptlrpc_connect_interpret()
2342 which thought it would be the only running connect process.
2345 Frequency : only with obd_echo servers and clients that are rebooted
2347 Description: kernel BUG accessing uninitialized data structure
2348 Details : When running an obd_echo server it did not start the ping_evictor
2349 thread, and when a client was evicted an uninitialized data
2350 structure was accessed. Start the ping_evictor in the RPC
2351 service startup instead of the OBD startup.
2353 Severity : enhancement
2354 Bugzilla : 10193 (patchless)
2355 Description: Remove dependency on various unexported kernel interfaces.
2356 Details : No longer need reparent_to_init, exit_mm, exit_files,
2357 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2360 Frequency : rare (only users of deprecated and unsupported LDAP config)
2362 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2363 Details : LDAP apparently contains 'lustreName' attributes instead of
2364 'name'. A simple remapping of the name is sufficient.
2367 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2369 Description: waiting_locks_callback trips kernel BUG if client is evicted
2370 Details : Running with the dump_on_timeout debug flag turned on makes
2371 it possible that the waiting_locks_callback() can try to dump
2372 the Lustre kernel debug logs from an interrupt handler. Defer
2373 this log dumping to the expired_lock_main() thread.
2375 Severity : enhancement
2377 Description: Support NFS exporting on 2.6 kernels.
2378 Details : Implement non-rawops metadata methods for NFS server to use without
2379 changing NFS server code.
2382 Frequency : very rare (synthetic metadata workload only)
2384 Description: two racing renames might cause an MDS thread to deadlock
2385 Details : Running the "racer" program may cause one MDS thread to rename
2386 a file from being the source of a rename to being the target of
2387 a rename at exactly the same time that another thread is doing
2388 so, and the second thread has already enqueued these locks after
2389 doing a lookup of the target and is trying to relock them in
2390 order. Ensure that we don't try to re-lock the same resource.
2393 Frequency : only very large systems with liblustre clients
2395 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2396 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2397 all OSTs in parallel. This allows the eviction of stale liblustre
2398 clients to proceed much faster than if they were done in series,
2399 and also offers similar improvements for other set_info RPCs.
2404 Description: excessive CPU usage during initial read phase on client
2405 Details : During the initial read phase on a client, it would agressively
2406 retry readahead on the file, consuming too much CPU and impacting
2407 performance (since 1.4.5.8). Improve the readahead algorithm
2408 to avoid this, and also improve some other common cases (read
2409 of small files in particular, where "small" is files smaller than
2410 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2415 Description: MDS crash when receiving packet with unknown intent.
2416 Details : Do not LBUG in unknown intent case, just return -EFAULT
2418 Severity : enhancement
2419 Bugzilla : 9293, 9385
2420 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2421 Details : Do not serialize getattr (non-intent version) and statfs.
2424 Frequency : occasional, when OST network is overloaded/intermittent
2426 Description: client evicted by OST after bulk IO timeout
2427 Details : If a client sends a bulk IO request (read or write) the OST
2428 may evict the client if it is unresposive to its data GET/PUT
2429 request. This is incorrect if the network is overloaded (takes
2430 too long to transfer the RPC data) or dropped the OST GET/PUT
2431 request. There is no need to evict the client at all, since
2432 the pinger and/or lock callbacks will handle this, and the
2433 client can restart the bulk request.
2436 Frequency : Always when mmapping file with no objects
2438 Description: client crashes when mmapping file with no objects
2439 Details : Check that we actually have objects in a file before doing any
2440 operations on objects in ll_vm_open, ll_vm_close and
2446 Description: Request leak when working with deleted CWD
2447 Details : Introduce advanced request refcount tracking for requests
2448 referenced from lustre intent.
2450 Severity : Enhancement
2452 Description: Cache open file handles on client.
2453 Details : MDS now will return special lock along with openhandle, if
2454 requested and client is allowed to hold openhandle, even if unused,
2455 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2456 closing files for every read/write openration.
2458 Severity : Enhancement
2460 Description: Cache open negative dentries on client when possible.
2461 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2462 negative dentries on lock revocation.
2467 Description: Remounting a client read-only wasn't possible with a zconf mount
2468 Details : It wasn't possible to remount a client read-only with llmount.
2470 Severity : enhancement
2471 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2472 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2473 MPICH is included to add Lustre-specific ADIO interfaces.
2474 This is based closely on the UFS ADIO layer and only differs
2475 in file creation, in order to allow the OST striping to be set.
2476 This is user-contributed code and not supported by CFS.
2481 Description: extended inode attributes (immutable, append-only) work improperly
2482 when 2.4 and 2.6 kernels are used on client/server or vice versa
2483 Details : Introduce kernel-independent values for these flags.
2485 Severity : enhancement
2488 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2489 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2490 granularity. Now, Fractional MB granularity is be supported,
2491 this is very useful for low memory system.
2493 Severity : enhancement
2495 Description: Getattr by fid
2496 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2497 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2498 have locks and avoids getattr rpc requests for it on every stat.
2501 Frequency : Always, for filesystems larger than 2TB
2503 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2504 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2505 filesystems larger than 2TB to be mounted with mballoc enabled.
2508 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2510 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2511 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2512 filesystems larger than 2TB is mounted then the kernel will
2513 silently corrupt the start of the filesystem. CONFIG_LBD is
2514 enabled for all CFS-supported kernels, but the possibility of
2515 this happening with a modified kernel config exists.
2517 Severity : enhancement
2519 Description: add client O_DIRECT support for 2.6 kernels
2520 Details : It is now possible to do O_DIRECT reads and writes to files
2521 in the Lustre client mountpoint on 2.6 kernel clients.
2523 Severity : enhancement
2525 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2526 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2527 parallel, not waiting for response from every OST before sending
2528 a rpc to the next OST.
2533 Description: setattr vs write race when updating file timestamps
2534 Details : Client processes that update a file timestamp into the past
2535 right after writing to the file (e.g. tar) it is possible that
2536 the updated file modification time can be reset to the current
2537 time due to a race between processing the setattr and write RPC.
2539 Severity : enhancement
2541 Description: Bring 'lfs find' closer in line with regular Linux find.
2542 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2543 -print0 options and obtains all the needed info through the lustre
2546 Severity : enhancement
2548 Description: support up to 1024 configured devices on one node
2549 Details : change obd_dev array from statically allocated to dynamically
2550 allocated structs as they are first used to reduce memory usage
2555 Description: Flush dirty partially truncated pages during truncate
2556 Details : Immediatelly flush partially truncated pages in filter_setattr,
2557 this way we completely avoid having any pages in page cache on OST
2558 and can retire ugly workarounds during writes to flush such pages.
2563 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2564 Details : protect inode from truncation within vfs_unlink() context
2565 just take a reference before calling vfs_unlink() and release it
2566 when parent's i_sem is free.
2571 Description: last_id value checked outside lock on OST caused LASSERT failure
2572 Details : If there were multiple MDS->OST object precreate requests in
2573 flight, it was possible that the OST's last object id was checked
2574 outside a lock and incorrectly tripped an assertion. Move checks
2575 inside locks, and discard old precreate requests.
2578 Frequency : always, if extents are used on OSTs
2580 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2581 Details : OSTs using the extents format would not zero the high 16 bits of
2582 the index physical block number. This is not a problem for any
2583 OST filesystems smaller than 16TB, and no kernels support ext3
2584 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2585 new/modified files) and can be fixed for existing filesystems
2586 with e2fsprogs-1.39-cfs1.
2591 Description: import connection selection may be incorrect if timer wraps
2592 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2593 import connections to be ignored if the 32-bit jiffies counter
2594 wraps. Use a 64-bit jiffies counter.
2597 Frequency : very large clusters immediately after boot
2599 Description: LNET request buffers exhausted under heavy short-term load
2600 Details : If a large number of client requests are generated on a service
2601 that has previously never seen so many requests it is possible
2602 that the request buffer growth cannot keep up with the spike in
2603 demand. Instead of dropping incoming requests, they are held in
2604 the LND until the RPC service can accept more requests.
2607 Frequency : Sometimes during replay
2609 Description: Assertion failure in ll_local_open after replay.
2610 Details : If replay happened on an open request reply before we were able
2611 to set replay handler, reply will become not swabbed tripping the
2612 assertion in ll_local_open. Now we set the handler right after
2613 recognising of open request
2616 Frequency : very rare
2618 Description: kernel reports "badness in vsnprintf"
2619 Details : Reading from the "recovery_status" /proc file in small chunks
2620 may cause a negative length in lprocfs_obd_rd_recovery_status()
2621 call to vsnprintf() (which is otherwise harmless). Exit early
2622 if there is no more space in the output buffer.
2624 Severity : enhancement
2626 Description: clear OBD RPC statistics by writing to them
2627 Details : It is now possible to clear the OBD RPC statistics by writing
2628 to the "stats" file.
2633 Description: Client mtime is not the same on different clients after utimes
2634 Details : In some cases, the client was using the utimes() syscall on
2635 a file cached on another node. The clients now validate the
2636 ctime from the MDS + OSTs to determine which one is right.
2641 Description: Inability to activate failout mode
2642 Details : lconf script incorrectly assumed that in pythong string's numeric
2643 value is used in comparisons.
2646 Frequency : always with multiple stripes per file
2648 Description: Inefficient object allocation for mutli-stripe files
2649 Details : When selecting which OSTs to stripe files over, for files with
2650 a stripe count that divides evenly into the number of OSTs,
2651 the MDS is always picking the same starting OST for each file.
2652 Return the OST selection heuristic to the original design.
2657 Description: mount failures may take full timeout to return an error
2658 Details : Under some heavy load conditions it is possible that a
2659 failed mount can wait for the full obd_timeout interval,
2660 possibly several minutes, before reporting an error.
2661 Instead return an error as soon as the status is known.
2663 Frequency : quota enabled and large files being deleted
2665 Description: releasing more than 4GB of quota at once hangs OST
2666 Details : If a user deletes more than 4GB of files on a single OST it
2667 will cause the OST to spin in an infinite loop. Release
2668 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2673 Description: statfs data retrieved from /proc may be stale or zero
2674 Details : When reading per-device statfs data from /proc, in the
2675 {kbytes,files}_{total,free,avail} files, it may appear
2676 as zero or be out of date.
2679 Frequency : systems with MD RAID1 external journal devices
2681 Description: lconf's call to blkid is confused by RAID1 journal devices
2682 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2683 of returning all block devices that match the journal UUID.
2686 Frequency : always, for aggregate stripe size over 4GB
2688 Description: assertion fails when trying to use 4GB stripe size
2689 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2690 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2693 Frequency : always on ppc64
2695 Description: the first write on an ext3 filesystem with mballoc got stuck
2696 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2697 perform endianness conversion.
2699 ------------------------------------------------------------------------------
2701 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2703 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2704 INTEROPERATE with older versions automatically. Please read the
2705 user documentation before upgrading any part of a live system.
2706 * WARNING: Lustre networking configuration changes are required with
2707 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2710 * Support for kernels:
2711 2.6.9-22.0.2.EL (RHEL 4)
2712 2.6.5-7.244 (SLES 9)
2713 2.6.12.6 vanilla (kernel.org)
2716 Severity : enhancement
2717 Bugzilla : 7981/8208
2718 Description: Introduced Lustre Networking (LNET)
2719 Details : LNET is new networking infrastructure for Lustre, it includes
2720 a reorganized network configuration mode (see the user
2721 documentation for full details) as well as support for routing
2722 between different network fabrics. Lustre Networking Devices
2723 (LNDs) for the supported network fabrics have also been
2724 created for this new infrastructure.
2726 Severity : enhancement
2727 Description: Introduced Access control lists
2728 Details : clients can set ACLs on files and directories in order to have
2729 more fine-grained permissions than the standard Unix UGO+RWX.
2730 The MDS must be started with the "-o acl" mount option.
2732 Severity : enhancement
2733 Description: Introduced filesystem quotas
2734 Details : Administrators may now establish per-user quotas on the
2737 Severity : enhancement
2739 Description: Configuration change for the XT3
2740 The PTLLND is now used to run Lustre over Portals on the XT3
2741 The configure option(s) --with-cray-portals are no longer used.
2742 Rather --with-portals=<path-to-portals-includes> is used to
2743 enable building on the XT3. In addition to enable XT3 specific
2744 features the option --enable-cray-xt3 must be used.
2749 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2750 Details : A race between allocating a new llcd and re-getting the llcd_lock
2751 allowed another thread to grab newly-allocated llcd.
2753 Severity : enhancement
2755 Description: 2.6 OST async journal commit and locking fix to improve performance
2756 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2757 2.6 kernels are now async as they already were in 2.4 kernels so
2758 that they can commit concurrently with the network bulk transfer.
2759 For block-allocated files the filter allocation semaphore is held
2760 to avoid filesystem fragmentation during allocation. BKL lock
2761 removed for 2.6 xattr operations where it is no longer needed.
2766 Description: lconf incorrectly determined whether two IP networks could talk
2767 Details : In some more complicated routing and multiple-network
2768 configurations, lconf will avoid trying to make a network
2769 connection to a disjoint part of the IP space. It was doing the
2770 math incorrectly for one set of cases.
2775 Description: Fix for potential infinite loop processing records in an llog.
2776 Details : If an llog record is corrupted/zeroed, it is possible to loop
2777 forever in llog_process(). Validate the llog record length
2778 and skip the remainder of the block on error.
2781 Frequency : occasional (liblustre only)
2783 Description: liblustre could not open files whose last component is a symlink
2784 Details : sysio_path_walk() would incorrectly pass the open intent to
2785 intermediate path components.
2788 Frequency : rare (liblustre only with non-standard tuning)
2789 Bugzilla : 7201 (7350)
2790 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2791 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2792 after it has been released, only do this for referenced locks
2794 Severity : enhancement
2796 Description: specifying an (invalid) directory default stripe_size of -1
2797 would reset the directory default striping
2798 Details : stripe_size -1 was used internally to signal directory stripe
2799 removal, now use "all default" to signal dir stripe removal
2800 as a directory striping of "all default" is not useful
2803 Frequency : common for large clusters running liblustre clients
2805 Description: doing an ls when liblustre clients are running is slow
2806 Details : sending a glimpse AST to a liblustre client waits for every AST
2807 to time out, as liblustre clients will not respond. Since they
2808 cannot cache data we refresh the OST lock LVB from disk instead.
2810 Severity : enhancement
2812 Description: doing an ls at the same time as file IO can be slow
2813 Details : enqueue and other "small" requests can be blocked behind many
2814 large IO requests. Create a new OST IO portal for non-IO
2815 requests so they can be processed faster.
2818 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2820 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2821 created file with mode 000
2822 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2823 allow an ftruncate() as a non-root user to a file with mode 000.
2824 Lustre now respects this flag to disable mode checks when
2825 truncating a file owned by the user
2828 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2831 Description: Revoking locks from clients that went dead or catatonic might take
2833 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2834 cancellation of such locks instant on servers without waiting for
2835 any reply from clients. Clients drops these locks when cancel
2836 notification from server is received without replying.
2839 Frequency : liblustre-only, when liblustre client dies or becomes busy
2841 Description: Doing ls on Linux clients can take a long time with active
2843 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2844 granting such locks to it in the first place if possible. Locks
2845 are taken by proxy on the OST during the read or write and
2846 dropped immediately afterward. Add connect flags handling, do
2847 not grant locks to liblustre clients for glimpse ASTs.
2849 Severity : enhancement
2851 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2852 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2853 sequential reads and behaves badly with applications doing random
2854 reads. Improve it by reading ahead at least read region, and
2855 avoiding excessive large RPC for small reads.
2857 Severity : enhancement
2859 Description: Creating more than 1000 files for a single job may cause a load
2860 imbalance on the OSTs if there are also a large number of OSTs.
2861 Details : qos_prep_create() uses an OST index reseed value that is an
2862 even multiple of the number of available OSTs so that if the
2863 reseed happens in the middle of the object allocation it will
2864 still utilize the OSTs as uniformly as possible.
2869 Description: OST or MDS may oops in ping_evictor_main()
2870 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2871 but doesn't restart at beginning of obd_exports_timed list
2874 Severity : enhancement
2876 Description: improve by-nid export eviction on the MDS and OST
2877 Details : allow multiple exports with the same NID to be evicted at one
2878 time without re-searching the exports list.
2881 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2883 Description: MDS may oops in groups_free()
2884 Details : in rare race conditions a newly allocated group_info struct is
2885 freed again, and this can be NULL. The 2.4 compatibility code
2886 for groups_free() checked for a NULL pointer, but 2.6 did not.
2889 Frequency : common for liblustre clients doing little filesystem IO
2890 Bugzilla : 9352, 7313
2891 Description: server may evict liblustre clients accessing contended locks
2892 Details : if a client is granted a lock or receives a completion AST
2893 with a blocking AST already set it would not reply to the AST
2894 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2897 Frequency : lfs setstripe, only systems with more than 160 OSTs
2899 Description: unable to set striping with a starting offset beyond OST 160
2900 Details : llapi_create_file() incorrectly limited the starting stripe
2901 index to the maximum single-file stripe count.
2904 Frequency : LDAP users only
2906 Description: lconf did not handle in-kernel recovery with LDAP properly
2907 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2909 Severity : enhancement
2911 Description: bind OST threads to NUMA nodes to improve performance
2912 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2913 node and do their allocations there to localize memory access
2915 Severity : enhancement
2917 Description: llmount can determine client NID directly from Myrinet (GM)
2918 Details : the client NID code from gmnalnid was moved directly into
2919 llmount, removing the need to use this or specifying the
2920 client NID explicitly when mounting GM clients with zeroconf
2923 Frequency : if client is started with down MDS
2925 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2926 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2927 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2928 ptlrpc_import_delayed_req() because we didn't check if the
2929 request was interrupted, and we also didn't break out of the
2930 event loop if there was no timeout
2935 Description: data loss during non-page-aligned writes to a single file from
2936 both multiple nodes and multiple threads on one node at same time
2937 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2938 inconsistency led to false short-reads, that were cached and later
2939 used by ->prepare_write() to fill in partially written page,
2940 leading to data loss.
2943 Frequency : always, if lconf --abort_recovery used
2945 Description: lconf --abort_recovery fails with 'Operation not supported'
2946 Details : lconf was attempting to abort recovery on the MDT device and not
2949 Severity : enhancement
2951 Description: remove cleanup logs
2952 Details : replace lconf-generated cleanup logs with lustre internal
2953 cleanup routines. Eliminates the need for client-cleanup and
2956 Severity : enhancement
2958 Description: add support for EAs (user and system) on lustre filesystems
2959 Details : it is now possible to store extended attributes in the Lustre
2960 client filesystem, and with the user_xattr mount option it
2961 is possible to allow users to store EAs on their files also
2963 Severity : enhancement
2965 Description: Add possibility (config option) to show minimal available OST free
2967 Details : When compiled with --enable-mindf configure option, statfs(2)
2968 (and so, df) will return least minimal free space available from
2969 all OSTs as amount of free space on FS, instead of summary of
2970 free spaces of all OSTs.
2972 Severity : enhancement
2974 Description: do not expand extent locks acquired on OST-side
2975 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2976 by server: they are not cached anyway.
2979 Frequency : when mmap is used/binaries executed from Lustre
2981 Description: Unmmap pages before throwing them away from read cache.
2982 Details : llap_shrink cache now attempts to unmap pages before discarding
2983 them (if unmapping failed - do not discard). SLES9 kernel has
2984 extra checks that trigger if this unmapping is not done first.
2989 Description: lconf didn't resolve symlinks before checking to see whether a
2990 given mountpoint was already in use
2993 Frequency : when migrating failover services
2994 Bugzilla : 6395, 9514
2995 Description: When migrating a subset of services from a node (e.g. failback
2996 from a failover service node) the remaining services would
2997 time out and evict clients.
2998 Details : lconf --force (implied by --failover) sets the global obd_timeout
2999 to 5 seconds in order to quickly disconnect, but this caused
3000 other RPCs to time out too quickly. Do not change the global
3001 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3003 Severity : enhancement
3004 Frequency : if MDS is started with down OST
3005 Bugzilla : 9439,5706
3006 Description: Allow startup/shutdown of an MDS without depending on the
3007 availability of the OSTs.
3008 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3009 Add appropriate locking and lov-osc refcounts for safe
3010 cleaning. Add osc abort_inflight calls in case the
3011 synchronize never started.
3014 Frequency : occasional (Cray XT3 only)
3016 Description: root not authorized to access files in CRAY_PORTALS environment
3017 Details : The client process capabilities were not honoured on the MDS in
3018 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3019 been accessed by an authorized user then root was able to access
3020 the file on the local client also. The root user capabilities
3021 are now allowed on the MDS, as this environment has secure UID.
3024 Frequency : occasional
3026 Description: ldiskfs "too long searching" message happens too often
3027 Details : A debugging message (otherwise harmless) prints too often on
3028 the OST console. This has been reduced to only happen when
3029 there are fragmentation problems on the filesystem.
3034 Description: Division by zero in statfs when all OSCs are inactive
3035 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3036 lov_statfs divides by value returned by lov_get_stripecnt().
3040 Bugzilla : 9489, 3273
3041 Description: First write from each client to each OST was only 4kB in size,
3042 to initialize client writeback cache, which caused sub-optimal
3043 RPCs and poor layout on disk for the first writen file.
3044 Details : Clients now request an initial cache grant at (re)connect time
3045 and so that they can start streaming writes to the cache right
3046 away and always do full-sized RPCs if there is enough data.
3047 If the OST is rebooted the client also re-establishes its grant
3048 so that client cached writes will be honoured under the grant.
3053 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3054 Details : Now I/O RPCs go to different portal number and (presumably) fast
3055 lock requests (and glimses) and other RPCs get their own service
3056 threads pool that should be able to service those RPCs
3059 Severity : enhancement
3061 Description: Ability to exchange lustre version between client and servers and
3062 issue warnings at client side if client is too old. Also for
3063 liblustre clients there is ability to refuse connection of too old
3065 Details : New 'version' field is added to connect data structure that is
3066 filled with version info. That info is later checked by server and
3070 Frequency : rare, liblustre only.
3071 Bugzilla : 9296, 9581
3072 Description: Two simultaneous writes from liblustre at offset within same page
3073 might proceed at the same time overwriting eachother with stale
3075 Details : I/O lock withing llu_file_prwv was released too early, before data
3076 actually was hitting the wire. Extended lock-holding time until
3077 server acknowledges receiving data.
3080 Frequency : extremely rare. Never observed in practice.
3082 Description: avoid generating lustre_handle cookie of 0.
3083 Details : class_handle_hash() generates handle cookies by incrementing
3084 global counter, and can hit 0 occasionaly (this is unlikely, but
3085 not impossible, because initial value of cookie counter is
3086 selected randonly). Value of 0 is used as a sentinel meaning
3087 "unassigned handle" --- avoid it. Also coalesce two critical
3088 sections in this function into one.
3090 Severity : enhancement
3092 Description: allow liblustre clients to delegate truncate locking to OST
3093 Details : To avoid overhead of locking, liblustre client instructs OST to
3094 take extent lock in ost_punch() on client's behalf. New connection
3095 flag is added to handle backward compatibility.
3097 Severity : enhancement
3098 Bugzilla : 4928, 7341, 9758
3099 Description: allow number of OST service threads to be specified
3100 Details : a module parameter allows the number of OST service threads
3101 to be specified via "options ost ost_num_threads={N}" in the
3102 OSS's /etc/modules.conf or /etc/modprobe.conf.
3106 Bugzilla : 6146, 9635, 9895
3107 Description: servers crash with bad pointer in target_handle_connect()
3108 Details : In rare cases when a client is reconnecting it was possible that
3109 the connection request was the last reference for that export.
3110 We would temporarily drop the export reference and get a new
3111 one, but this may have been the last reference and the export
3112 was just destroyed. Get new reference before dropping old one.
3114 Severity : enhancement
3115 Frequency : if client is started with failover MDS
3117 Description: Allow multiple MDS hostnames in the mount command
3118 Details : Try to read the configuration from all specified MDS
3119 hostnames during a client mount in case the "primary"
3122 Severity : enhancement
3124 Description: Stop sending data to evicted clients as soon as possible.
3125 Details : Check if the client we are about to send or are sending data to
3126 was evicted already. (Check is done every second of waiting,
3127 for which l_wait_event interface was extended to allow checking
3128 of exit condition at specified intervals).
3131 Frequency : rare, normally only when NFS exporting is done from client
3133 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3135 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3138 Frequency : always, if srand() is called before liblustre initialization
3140 Description: Liblustre uses system PRNG disturbing its usage by user application
3141 Details : Introduce internal to lustre fast and high-quality PRNG for
3142 lustre usage and make liblustre and some other places in generic
3143 lustre code to use it.
3145 Severity : enhancement
3146 Bugzilla : 9477, 9557, 9870
3147 Description: Verify that the MDS configuration logs are updated when xml is
3148 Details : Check if the .xml configuration logs are newer than the config
3149 logs stored on the MDS and report an error if this is the case.
3150 Request --write-conf, or allow starting with --old_conf.
3152 Severity : enhancement
3154 Description: Handle symlinks in the path when checking if Lustre is mounted.
3155 Details : Resolve intermediate symlinks when checking if a client has
3156 mounted a filesystem to avoid duplicate client mounts.
3161 Description: lconf can hit an error exception but still return success.
3162 Details : The lconf command catches the Command error exception at the top
3163 level script context and will exit with the associated exit
3164 status, but doesn't ensure that this exit status is non-zero.
3169 Description: failure of ptlrpc thread startup can cause oops
3170 Details : Starting a ptlrpc service thread can fail if there are a large
3171 number of threads or the server memory is very fragmented.
3172 Handle this without oopsing.
3175 Frequency : always, only if liblustre and non-default acceptor port was used
3177 Description: liblustre cannot connect to servers with non-default acceptor port
3178 Details : tcpnal_set_default_params() was not called and was therefore
3179 ignoring the environment varaible TCPNAL_PORT, as well as other
3180 TCPNAL_ environment variables
3185 Description: two objects could be created on the same OST for a single file
3186 Details : If an OST is down, in some cases it was possible to create two
3187 objects on a single OST for a single file. No problems other
3188 than potential performance impact and spurious error messages.
3192 Bugzilla : 5681, 9562
3193 Description: Client may oops in ll_unhash_aliases
3194 Details : Client dcache may become inconsistent in race condition.
3195 In some cases "getcwd" can fail if the current directory is
3201 Description: Inode refcounting problems in NFS export code
3202 Details : link_raw functions used to call d_instantiate without obtaining
3203 extra inode reference first.
3207 Bugzilla : 9942, 9903
3208 Description: Referencing freed requests leading to crash, memleaks with NFS.
3209 Details : We used to require that call to ll_revalidate_it was always
3210 followed by ll_lookup_it. Also with revalidate_special() it is
3211 possible to call ll_revalidate_it() twice for the same dentry
3212 even if first occurence returned success. This fix changes semantic
3213 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3214 reference on a request referred from the intent.
3215 ll_intent_release() then releases such a request.
3218 Frequency : rare, normally benchmark loads only
3220 Description: unlinked inodes were kept in memory on the client
3221 Details : If a client is repeatedly creating and unlinking files it
3222 can accumulate a lot of stale inodes in the inode slab cache.
3223 If there is no other client load running this can cause the
3224 client node to run out of memory. Instead flush old inodes
3225 from client cache that have the same inode number as a new inode.
3228 Frequency : SLES9 2.6.5 kernel and long filenames only
3229 Bugzilla : 9969, 10379
3230 Description: utime reports stale NFS file handle
3231 Details : SLES9 uses out-of-dentry names in some cases, which confused
3232 the lustre dentry revalidation. Change it to always use the
3236 Frequency : rare, unless heavy write-truncate concurrency is continuous
3237 Bugzilla : 4180, 6984, 7171, 9963, 9331
3238 Description: OST becomes very slow and/or deadlocked during object unlink
3239 Details : filter_destroy() was holding onto the parent directory lock
3240 while truncating+unlinking objects. For very large objects this
3241 may block other threads for a long time and slow overall OST
3242 responsiveness. It may also be possible to get a lock ordering
3243 deadlock in this case, or run out of journal credits because of
3244 the combined truncate+unlink. Solution is to do object truncate
3245 first in one transaction without parent lock, and then do the
3246 final unlink in a new transaction with the parent lock. This
3247 reduces the lock hold time dramatically.
3250 Frequency : rare, 2.4 kernels only
3252 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3253 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3254 before unmounting their filesystem in order to avoid deadlock.
3255 The kernel_locked() function in 2.4 kernels only checks whether
3256 the kernel lock is held, not whether it is this process that is
3257 holding it as 2.6 kernels do.
3262 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3263 Details : The client ptlrpc code may be trying to reconnect to a down
3264 server before a previous connection attempt has timed out.
3265 Increase the reconnect interval to be longer than the connection
3266 timeout interval to avoid sending duplicate connections to
3270 Frequency : echo_client brw_test command
3272 Description: fix echo_client to work with OST preallocated code
3273 Details : OST preallocation code (5137) didn't take echo_client IO path
3274 into account: echo_client calls filter methods outside of any
3275 OST thread and, hence, there is no per-thread preallocated
3276 pages and buffers to use. Solution: hijack pga pages for IO. As
3277 a byproduct, this avoids unnecessary data copying.
3281 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3282 Description: Client can oops in mdc_commit_close() after open replay
3283 Details : It was possible for the MDS to return an open request with no
3284 transaction number in mds_finish_transno() if the client was
3285 evicted, but without actually returning an error. Clients
3286 would later try to replay that open and may trip an assertion
3287 Simplify the client close codepath, and always return an error
3288 from the MDS in case the open is not successful.
3291 Frequency : rare, 2.6 OSTs only
3293 Description: OST may deadlock under high load on fragmented files
3294 Details : If there was a heavy load and highly-fragmented OST filesystems
3295 it was possible to have all the OST threads deadlock waiting on
3296 allocation of biovecs, because the biovecs were not released
3297 until the entire RPC IO was completed. Instead, release biovecs
3298 as soon as they are complete to ensure forward IO progress.
3300 Severity : enhancement
3302 Description: Support for specifying external journal device at mount
3303 Details : If an OST or MDS device is formatted with an external journal
3304 device, this device major/minor is stored in the ext3 superblock
3305 and may not be valid for failover. Allow detecting and
3306 specifying the external journal at mount time.
3311 Description: Mounting an MDS with pending unlinked files may cause oops
3312 Details : target_finish_recovery() calls mds_postrecov() which returned
3313 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3314 considers this an error and immediately begins cleaning up the
3315 lov, just after starting the mds_lov process
3317 Severity : enhancement
3319 Description: Implement 'lfs df' to report actual free space on per-OST basis
3320 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3321 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3322 usage of objects; '-h' to report in human readable format.
3324 ------------------------------------------------------------------------------
3326 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3333 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3334 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3335 at startup during orphan unlinking. Instead, skip buddy update
3336 before setup, as it will be regenerated after recovery is complete.
3341 Description: If an OST is inactive, its locks might reference stale inodes.
3342 Details : lov_change_cbdata() must iterate over all namespaces, even if
3343 they are inactive to clear inode references from the lock.
3345 Severity : enhancement
3346 Frequency : occasional, if non-standard max_dirty_mb used
3348 Description: Client will block write RPCs if not enough grant
3349 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3350 then the client will block writes while waiting for another RPC
3351 to complete instead of consuming its dirty limit. With change
3352 we get improved performance when max_dirty_mb is small.
3354 Severity : enhancement
3355 Bugzilla : 3389, 6253
3356 Description: Add support for supplementary groups on the MDS.
3357 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3358 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3359 lookups for user supplementary groups into a cache.
3363 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3364 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3365 for liblustre O_EXCL is enforced by MDS.
3369 Description: readdir never returns NULL in liblustre.
3370 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3371 dentry in struct dirent.
3375 Frequency : liblustre only, depends on application IO pattern
3376 Description: liblustre clients evicted if not contacting servers
3377 Details : Don't put liblustre clients into the ping_evictor list, so
3378 they will not be evicted by the pinger ever.
3380 Severity : enhancement
3382 Description: Add ability to evict clients by NID from MDS.
3383 Details : By echoing "nid:$NID" string into
3384 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3385 $NID would be instantly evicted from this MDS and from all active
3386 OSTs connected to it.
3390 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3391 Details : lookup_it_finish() used to query file size from OSTs that was not
3396 Description: service threads change working directory to that of init
3397 Details : Starting lustre service threads may pin the working directory
3398 of the parent thread, making that filesystem busy. Threads
3399 now change to the working directory of init to avoid this.
3403 Frequency : during shutdown only
3404 Description: shutdown with a failed MDS or OST can cause unmount to hang
3405 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3410 Frequency : 2.6 only
3411 Description: chmod/chown may include an extra supplementary group
3412 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3413 supplementary group and if none is specified this is used.
3416 Bugzilla : 5479 (6816)
3418 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3419 Details : If lookup is in progress on a file that is unlinked we might try
3420 to revalidate the inode and fail in revalidate after lookup is
3421 complete and ll_file_open() enqueues the open again but
3422 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3425 Frequency : always, if lconf --abort_recovery used
3427 Description: lconf --abort_recovery fails with 'Operation not supported'
3428 Details : lconf was attempting to abort recovery on the MDT device and not
3431 ------------------------------------------------------------------------------
3433 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3438 Frequency : rare (only unsupported configurations with a node running as an
3440 Bugzilla : 6514, 5137
3441 Description: Mounting a Lustre file system on a node running as an OST could
3443 Details : OSTs now preallocates memory needed to write out data at
3444 startup, instead of when needed, to avoid having to
3445 allocate memory in possibly low memory situations.
3446 Specifically, if the file system is mounted on on OST,
3447 memory pressure could force it to try to write out data,
3448 which it needed to allocate memory to do. Due to the low
3449 memory, it would be unable to do so and the node would
3450 become unresponsive.
3452 Severity : enhancement
3454 Description: Addition of lconf --service command line option
3455 Details : lconf now accepts a '--service <arg>' option, which is
3456 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3458 Severity : enhancement
3460 Description: Failover mode is now the default for OSTs.
3461 Details : By default, OSTs will now run in failover mode. To return to
3462 the old behaviour, add '--failout' to the lmc line for OSTs.
3464 Severity : enhancement
3466 Description: Health checks are now provided for MDS and OSTs
3467 Details : Additional detailed health check information on MSD and OSTs
3468 is now provided through the procfs health_check value.
3471 Frequency : occasional, depends on IO load
3473 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3474 after numerous create/delete cycles
3475 Details : The ext3 inode allocation policy would not allocate new inodes
3476 very well on the OSTs because there are no new directories
3477 being created. Instead we look for groups with free space if
3478 the parent directories are nearly full.
3483 Description: Network or server problems during mount may cause partially
3484 mounted clients instead of returning an error.
3485 Details : The config llog parsing code may overwrite the error return
3486 code during mount error handling, returning success instead
3492 Description: MDS can fail to allocate large reply buffers
3493 Details : After long uptimes the MDS can fail to allocate large reply
3494 buffers (e.g. zconf client mount config records) due to memory
3495 fragmentation or consumption by the buffer cache. Preallocate
3496 some large reply buffers so that these replies can be sent even
3497 under memory pressure.
3501 Frequency : rare (liblustre)
3502 Description: fsx running with liblustre complained that using truncate() to
3503 extend the file doesn't work. This patch corrects that issue.
3504 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3505 ATTR_SIZE and lsm use in llu_setattr_raw.
3509 Frequency : rare, only 2.6 kernels
3510 Description: Unusual file access patterns on the MDS may result in inode
3511 data being lost in very rare circumstances.
3512 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3513 code in the 2.6 kernel caused the inode and/or EA data not to
3514 be read from disk, causing single-file corruption.
3518 Frequency : rare, only 2.6 filesystems using extents
3519 Description: Heavy concurrent write and delete load may cause data corruption.
3520 Details : It was possible under high-load situations to have an extent
3521 metadata block in the block device cache from a just-unlinked
3522 file overwrite a newly-allocated data block. We now unmap any
3523 metadata buffers that alias just-allocated data blocks.
3527 Frequency : filesystems with default stripe_count larger than 77
3528 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3529 Details : lconf specifies an inode size of 4096 bytes when the default
3530 stripe_count is larger than 77. This conflicts with the default
3531 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3532 this case to avoid pinning too much memory for large EAs.
3534 ------------------------------------------------------------------------------
3536 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3541 Frequency : rare (extremely heavy IO load with hundreds of clients)
3543 Description: Client is evicted, gets IO error writing to file
3544 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3545 caused clients to be evicted for AST timeouts. The fixes in
3546 bug 5192 mean we no longer need to have such short AST timeouts
3547 so ldlm_timeout has been increased.
3550 Frequency : occasional during --force or --failover shutdown under load
3551 Bugzilla : 5949, 4834
3552 Description: Server oops/LBUG if stopped with --force or --failover under load
3553 Details : a collection of import/export refcount and cleanup ordering
3554 issues fixed for safer force cleanup
3557 Frequency : only filesystems larger than 120 OSTs
3558 Bugzilla : 5990, 6223
3559 Description: lfs getstripe would oops on a very large filesystem
3560 Details : lov_getconfig used kfree on vmalloc'd memory
3563 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3564 Bugzilla : 6242, 6243
3565 Description: reading from files that had been truncated to a non-zero size
3566 but never opened returned no data
3567 Details : ll_file_read() reads zeros from no-object files to EOF
3572 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3573 Details : There's little harm in aborting MDS/OSS recovery and letting it
3574 try again, so I removed the LASSERT and return an error instead.
3576 Severity : enhancement
3578 Description: New debugging infrastructure for tracking down data corruption
3579 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3580 (b) cover more of the client-side code path, and (c) try to narrow
3581 down where problems occurred
3585 Bugzilla : 3819, 4364, 4397, 6313
3586 Description: Racing close and eviction MDS could cause assertion in mds_close
3587 Details : It was possible to get multiple mfd references during close and
3588 client eviction, leading to one thread referencing a freed mfd.
3590 Severity: : enhancement
3591 Bugzilla : 3262, 6359
3592 Description: Attempts to reconnect to servers are now more aggressive.
3593 Details : This builds on the enhanced upcall-less recovery that was added
3594 in 1.4.2. When trying to reconnect to servers, clients will
3595 now try each server in the failover group every 10 seconds. By
3596 default, clients would previously try one server every 25 seconds.
3601 Description: After recovery, certain operations trigger a failed
3602 assertion on a client.
3603 Details : Failing over an mds, using lconf -d --failover, while a
3604 client was doing a readdir() call would cause the client to
3605 LBUG after recovery completed and the readdir() was resent.
3607 Severity : enhancement
3609 Description: Default groups are now added by lconf
3610 Details : You can now run lconf --group <servicename> without having to
3611 manually add groups with lmc.
3614 Frequency : occasional
3616 Description: Nodes with an elan id of 0 trigger a failed assertion
3619 Frequency : always when accessing e.g. tty/console device nodes
3621 Description: tty and some other devices nodes cannot be used on lustre
3622 Details : file's private_data field is used by device data and lustre
3623 values in there got lost. New field was added to struct file to
3624 store fs-specific private data.
3627 Frequency : when exporting Lustre via NFS
3629 Description: NFSD failed occasionally when looking up a path component
3630 Details : NFSD is looking up ".." which was broken in ext3 directories
3631 that had grown large enough to become hashed.
3634 Frequency : Clusters with multiple interfaces not on the same subnet
3636 Description: Nodes will repeatedly try to reconnect to an interface which it
3637 cannot reach and report an error to the log.
3638 Details : Extra peer list entries will be created by lconf with some peers
3639 unreachable. lconf now validates the peer before adding it.
3642 Frequency : Only if a default stripe is set on the filesystem root.
3644 Description: Setting a default stripe on the filesystem root prevented the
3645 filesystem from being remounted.
3646 Details : The client was sending extra request flags in the root getattr
3647 request and did not allocate a reply buffer for the dir EA.
3650 Frequency : occasional, higher if lots of files are accessed by one client
3651 Bugzilla : 6159, 6097
3652 Description: Client trips assertion regarding lsm mismatch/magic
3653 Details : While revalidating inodes the VFS looks up inodes with ifind()
3654 and in rare cases can find an inode that is being freed.
3655 The ll_test_inode() code will free the lsm during ifind()
3656 when it finds an existing inode and then the VFS later attaches
3657 this free lsm to a new inode.
3661 Bugzilla : 6422, 7030
3662 Description: MDS deadlock between mkdir and client eviction
3663 Details : Creating a new file via mkdir or mknod (starting a transaction
3664 and getting the ns lock) can deadlock with client eviction
3665 (gets ns lock and trying to finish a synchronous transaction).
3668 Frequency : occasional
3669 Description: While starting a server, the fsfilt_ext3 module could not be
3671 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3672 kernels. Previously, lconf would still use the ext3 name
3673 when trying to load modules. Now, it will correctly use
3674 ext3 on 2.4 and ldiskfs on 2.6.
3676 Severity : enhancement
3677 Description: The default stripe count has been changed to 1
3678 Details : The interpretation of the default stripe count (0, to lfs
3679 or lmc) has been changed to mean striping across a single
3680 OST, rather than all available. For general usage we have
3681 found a stripe count of 1 or 2 works best.
3683 Severity : enhancement
3684 Description: Add support for compiling against Cray portals.
3685 Details : Conditional compiling for some areas that are different
3689 Frequency : occasional
3690 Bugzilla : 6409, 6834
3691 Description: Creating files with an explicit stripe count may lead to
3692 a failed assertion on the MDS
3693 Details : If some OSTs are full or unavailable, creating files may
3694 trigger a failed assertion on the MDS. Now, Lustre will
3695 try to use other servers or return an error to the
3699 Frequency : occasional
3701 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3702 caused lock timeout during readahead (since 1.4.2).
3703 Details : Processes doing readahead might match a lock that hasn't been
3704 granted yet if there are overlapping and conflicting lock
3705 requests. The readahead process waits on ungranted lock
3706 (original lock is CBPENDING), while OST waits for that process
3707 to cancel CBPENDING read lock and eventually evicts client.
3709 Severity : enhancement
3711 Description: Initial enabling of flock support for clients
3712 Details : Implements fcntl advisory locking and file status functions.
3713 This feature is provided as an optional mount flag (default
3714 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3715 locking are implemented yet, and those that are are not guaranteed
3716 to be completely correct in production environments.
3717 mount -t lustre -o [flock|noflock] ...
3720 Frequency : occasional
3722 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3723 assertion in the ext3 JBD (journaling) layer.
3724 Details : The b_committed_data struct is protected by the big kernel lock
3725 in 2.4 kernels, serializing journal_commit_transaction() and
3726 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3727 access to this struct. In 2.6 kernels there is finer grained
3728 locking to improve SMP performance of the JBD layer.
3732 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3737 Description: merge and cleanup kernel patches.
3740 -----------------------------------------------------------------------------
3742 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3744 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3745 versions of Lustre. Please update all servers and clients to
3746 version 1.4.2 or later at the same time. You must also run
3747 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3748 to update the configuration logs.
3750 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3751 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3752 - swab llog records of type '0' so we get proper header size/idx (5861)
3753 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3754 - fix rename of one directory over another leaking an inode (5953)
3755 - avoid SetPageDirty on 2.6 (5981)
3756 - don't re-add just-being-destroyed locks to the waiting list (5653)
3757 - when creating new directories, inherit the parent's custom
3758 striping settings if present parent (3048)
3759 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3760 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3761 - don't hold client locks on temporary worklist from l_lru (5666)
3762 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3763 - automatically evict dead clients (5921)
3764 - Update file size properly in create+truncate+fstat case (6196)
3765 - Do not unhash mountpoint dentries, do not allow removal of
3767 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3768 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3769 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3770 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3771 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3772 - Correctly handle reads of files with no objects (6243)
3773 - lctl recover will also mark a device active if deactivate used (5933)
3775 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3776 - allow --write-conf on an MDS with different nettype than client (5619)
3777 - don't write config llogs to MDS for mounts not from that MDS (5617)
3778 - lconf should create multiple TCP connections from a client (5201)
3779 - init scripts are now turned off by default; run chkconfig --on
3780 lustre and chkconfig --on lustrefs to use them
3781 - upcalls are no longer needed for clients to recover to failover
3783 - add --abort-recovery option to lconf to abort recovery on device
3785 - add support for an arbitrary number of OSTs (3026)
3786 - Quota support protocol changes.
3787 - forward compatibility changes to wire structs (6007)
3788 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3789 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3790 - improved "lustre status" script
3791 - initialize blocksize for non-regular files (6062)
3792 - added --disable-server and --disable-client configure options (5782)
3793 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3794 - Vanilla 2.4.29 support
3795 - increase maximum number of obd devices to 520 (6242)
3796 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3797 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3798 SLES 9 2.6.5 kernels
3799 - limit stripes per file to 160 (the maximum EA size) (6093)
3801 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3804 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3805 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3806 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3807 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3808 - small ext3 extents cleanups and fixes (5733)
3809 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3810 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3811 - avoid needless client->OST connect, fix handle mismatch (5317)
3812 - fix DLM error path that led to out-of-sync client, long delays (5779)
3813 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3814 - fix several locking issues related to i_size (5492,5624,5654,5672)
3815 - don't move pending lock onto export if it is already evicted (5683)
3816 - fix kernel oops when creating .foo in unlinked directory (5548)
3817 - fix deadlock in obdfilter statistics vs. object create (5811)
3818 - use time_{before,after} to avoid timer jiffies wrap (5882)
3819 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3820 - Do not leak request if server was not able to process it (5154)
3821 - If mds_open unable to find parent dir, make that negative lookup(5154)
3822 - don't create new directories with extent-mapping (5909, 5936)
3824 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3825 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3826 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3827 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3828 - support for RHEL4 kernel on i686 (5773)
3829 - provide error messages when incompatible logs are encountered (5898)
3831 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3832 * version 1.4.0.10 (1.4.1 release candidate 1)
3834 - don't keep a lock reference when lock is not granted (4238)
3835 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3836 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3837 - drop import inflight refcount on signal_completed_replay error (5255)
3838 - unlock page after async write error during send (3677)
3839 - handle missing objects in filter_preprw_read properly (5265)
3840 - no transno return for symlink open, don't save no-trasno open (3440)
3841 - don't try to complete elan receive that already failed (4012)
3842 - free RPC server reply state on error (5406)
3843 - clean up thread from ptlrpc_start_thread() on error (5160)
3844 - readahead could read extra page into cache that wasn't ejected (5388)
3845 - prevent races in class_attach/setup/cleanup/detach (5260)
3846 - don't dereference de->d_inode after l_dput of de (5458)
3847 - use "int" for stripe value returned from lock_to_stripe (5544)
3848 - mballoc allocation and error-checking fixes in 2.6 (5504)
3849 - block device patches to fix I/O request sizes in 2.6 (5482)
3850 - look up hostnames for IB nals (5602)
3851 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3852 - don't start multiple acceptors for the same port (5277)
3853 - fix incorrect LASSERT in mds_getattr_name (5635)
3854 - export a proc file for general "ping" checking (5628)
3855 - fix "lfs check" to not block when the MDS is down (5628)
3857 - service request history (4965)
3858 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3859 - create an "evict_client" /proc entry on OSTs, like the MDS has
3860 - fix mount usage message, return errors per mount(8) (5168)
3861 - change grep [] to grep "[]" in tests so they work in more UMLs
3862 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3863 - remove ancient LOV_MAGIC_V0 EA support (5047)
3864 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3865 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3866 - allow readahead allocations to fail when low on memory (5383)
3867 - mmap locking landed again, after considerable improvement (2828)
3868 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3870 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3873 - send OST transaction number in read/write reply to free req (4966)
3874 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3875 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3876 - the watchdog thread now runs as interruptible (5246)
3877 - flock/lockf fixes (but it's still disabled, pending 5135)
3878 - don't use EXT3 constants in llite code (5094)
3879 - memory shortage at startup could cause assertion (5176)
3881 - reorganization of lov code
3882 - single portals codebase
3884 - add extents/mballoc support (5025)
3885 - direct I/O reads in the obdfilter (4048)
3886 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3888 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3891 - send OST transaction number in read/write reply to free req (4966)
3892 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3893 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3894 - the watchdog thread now runs as interruptible (5246)
3895 - handle missing objects in filter_preprw_read properly (5265)
3896 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3897 - drop import inflight refcount on signal_completed_replay error (5255)
3898 - unlock page after async write error during send (3677)
3899 - return original error code on reconstructed replies (3761)
3900 - no transno return for symlink open, don't save no-trasno open (3440)
3902 - add pid to ldlm debugging output (4922)
3903 - bump the watchdog timeouts -- we can't handle 30sec yet
3904 - extra debugging for orphan dentry/inode bug (5259)
3906 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3909 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3910 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3911 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3912 - fix NULL dereference in /proc/sys/portals/routes (4827)
3913 - allow failed mdc_close() operations to be interrupted (4561)
3914 - stop precreate on OST before MDS would time out on it (4778)
3915 - don't send partial-page writes before EOF from client (4410)
3916 - discard client grant for sub-page writes on large-page clients (4520)
3917 - don't free dentries not owned by NFS code, check generation (4806)
3918 - fix lsm leak if mds_create_objects() fails (4801)
3919 - limit debug_daemon file size, always print CERROR messages (4789)
3920 - use transno after validating reply (3892)
3921 - process timed out requests if import state changes (3754)
3922 - update mtime on OST during writes, return in glimpse (4829)
3923 - add mkfsoptions to LDAP (4679)
3924 - use ->max_readahead method instead of zapping global ra (5039)
3925 - don't interrupt __l_wait_event() during strace
3927 - add software watchdogs to catch hung threads quickly (4941)
3928 - make lustrefs init script start after nfs is mounted
3929 - fix CWARN/ERROR duplication (4930)
3930 - return async write errors to application if possible (2248)
3931 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3932 - print NAL number in %x format (4645)
3933 - update barely-supported suse-2.4.21-171 series (4842)
3934 - support for sles 9 %post scripts
3935 - support for building 2.6 kernel-source packages
3936 - support for sles km_* packages
3938 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3941 - ignore -ENOENT errors in osc_destroy (3639)
3942 - notify osc create thread that OSC is being cleaned up (4600)
3943 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3944 - reconstruct ost_handle() like mds_handle() (4657)
3945 - create a new thread to do import eviction to avoid deadlock (3969)
3946 - let lconf resolve symlinked-to devices (4629)
3947 - don't unlink "objects" from directory with default EA (4554)
3948 - hold socknal file ref over connect in case target is down (4394)
3949 - allow more than 32000 subdirectories in a single directory (3244)
3950 - fix blocks count for O_DIRECT writes (3751)
3951 - OST returns ENOSPC from object create when no space left (4539)
3952 - don't send truncate RPC if file size isn't changing (4410)
3953 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3954 - bind to privileged port in socknal and tcpnal (3689)
3956 - rate limit CERROR/CWARN console message to avoid overload (4519)
3957 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3958 - basic mmap support (3918)
3959 - kernel patch series update from b1_4 (4711)
3961 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3964 - avoid crash during MDS cleanup with OST shut down (2775)
3965 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3966 - don't use bad inodes on the MDS (3744)
3967 - dynamic object preallocation to improve recovery speed (4236)
3968 - don't hold spinlock over lock dumping or change debug flags (4401)
3969 - don't zero obd_dev when it is force cleaned (3651)
3970 - print grants to console if they go negative (4431)
3971 - "lctl deactivate" will stop automatic recovery attempts (3406)
3972 - look for existing locks in ldlm_handle_enqueue() (3764)
3973 - don't resolve lock handle twice in recovery avoiding race (4401)
3974 - revalidate should check working dir is a directory (4134)
3976 - don't always mark "slow" obdfilter messages as errors (4418)
3978 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3981 - don't close LustreDB during write_conf until it is done (3860)
3982 - fix typo in lconf for_each_profile (3821)
3983 - allow dumping logs from multiple threads at one time (3820)
3984 - don't allow multiple threads in OSC recovery (3812)
3985 - fix debug_size parameters (3864)
3986 - fix mds_postrecov to initialize import for llog ctxt (3121)
3987 - replace config semaphore with spinlock (3306)
3988 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3989 - don't allow enqueue to complete on a destroyed export (3822)
3990 - down write_lock before checking llog header bitmap (3825)
3991 - recover from lock replay timeout (3764)
3992 - up llog sem before sending rpc (3652)
3993 - reduce ns lock hold times when setting kms (3267)
3994 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3995 - fix NULL deref and obd_dev leak on setup error (3312)
3996 - replace some LBUG about llog ops with error handling (3841)
3997 - don't match INVALID dentries from d_lookup and spin (3784)
3998 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3999 - fix invalid assertion in ptlrpc_set_wait (3880)
4001 - add libwrap support for the TCP acceptor (3996)
4002 - add /proc/sys/portals/routes for non-root route listing (3994)
4003 - allow setting MDS UUID in .xml (2580)
4004 - print the stack of a process that LBUGs (4228)
4006 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4009 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4010 - make sure to unset replay flag from failed open requests (3440)
4011 - if default stripe count is 0, use OST count for inode size (3636)
4012 - update parent mtime/ctime on client for create/unlink (2611)
4013 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4014 - free recovery state on server during a forced cleanup (3571)
4015 - unregister_reply for resent reqs (3063)
4016 - loop back devices mounting and status check on 2.6 (3563)
4017 - fix resource-creation race that can provoke i_size == 0 (3513)
4018 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4019 - more debugging for page-accounting assertion (3746)
4020 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4021 - avoid deadlock after precreation failure (3758)
4022 - fix race and lock order deadlock in orphan handling (3450, 3750)
4023 - add validity checks when grabbing inodes from l_ast_data (3599)
4025 - add /proc/.../recovery_status to obdfilter (3428)
4026 - lightweight CDEBUG infrastructure, debug daemon (3668)
4027 - change default OSC RPC parameters to be better on small clusters
4028 - turn off OST read cache for files smaller than 32MB
4029 - install man pages and include them in rpms (3100)
4030 - add new init script for (un)mounting lustre filesystems (2593)
4031 - run chkconfig in %post for init scripts (3701)
4032 - drop scimac NAL (unmaintained)
4034 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4037 - clean kiobufs before and after use (3485)
4038 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4039 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4040 - change init script to fail more clearly if not run as root (1528)
4041 - allow clients to reconnect during replay (1742)
4042 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4043 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4044 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4045 - fix cond_resched (3554)
4046 - only evict unfinished clients after recovery (3515)
4047 - allow bulk resend, prevent data loss (3570)
4048 - dynamic ptlrpc request buffer allocation (2102)
4049 - don't allow unlinking open directory if it isn't empty (2904)
4050 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4051 - remove extraneous obd dereference causing LASSERT failure (3334)
4052 - don't use get_cycles() when creating temp. files on the mds (3156)
4053 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4054 - handle EEXIST for set-stripe, set proper directory name (3336)
4056 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4057 - fix ksocknal_fmb_callback() error messages (2918)
4059 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4062 - don't copy lvb into (possibly NULL) reply on error (2983)
4063 - don't deref dentry after dput, don't free lvb on error (2922)
4064 - use the kms to determine writeback rpc length (2947)
4065 - increment oti_logcookies when osc is inactive (2948)
4066 - update client's i_blocks count via lvb messages (2543)
4067 - handle intent open/close of special files properly (1557)
4068 - mount MDS with errors=remount-ro, like obdfilter (2009)
4069 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4070 - don't use cancelling-locks' kms values (2947)
4071 - use highest lock extent for kms, not last one (2925)
4072 - don't dereference ERR_PTR() dentry in error handling path (3107)
4073 - fix thread race in portals_debug_dumplog() (3122)
4074 - create lprocfs device entries at setup instead of at attach (1519)
4075 - common AST error handler, don't evict client on completion race (3145)
4076 - zero nameidata in detach_mnt in 2.6 (3118)
4077 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4078 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4079 - initialize RPC timeout timer earlier for 2.6 (3219)
4080 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4081 - print nal/nid for unknown nid (3258)
4082 - additional checks for oscc recovery before doing precreate (3284)
4083 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4084 - don't crash in mdc_close for bad permissions on open (3285)
4085 - zero i_rdev for non-device files (3147)
4086 - clear page->private before handing to FS, better assertion (3119)
4087 - tune the read pipeline (3236)
4088 - fix incorrect decref of invalidated dentry (2350)
4089 - provide read-ahead stats and refine rpc in flight stats (3328)
4090 - don't hold journal transaction open across create RPC (3313)
4091 - update atime on MDS at close time (3265)
4092 - close LDAP connection when recovering to avoid server load (3315)
4093 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4094 - don't leak open file on MDS after open resend (3325)
4095 - serialize filter_precreate and filter_destroy_precreated (3329)
4096 - loop device shouldn't call sync_dev() for nul device (3092)
4097 - clear page cache after eviction (2766)
4098 - resynchronize MDS->OST in background (2824)
4099 - refuse to mount the same filesystem twice on same mountpoint (3394)
4100 - allow llmount to create routes for mounting behind routers (3320)
4101 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4102 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4103 - fix rare problem with rename on htree directories (3417)
4105 - allow default OST striping configuration per directory (1414)
4106 - fix compilation for qswnal for 2.6 kernels (3125)
4107 - increase maximum number of MDS request buffers for large systems
4108 - change liblustreapi to be useful for external progs like lfsck (3098)
4109 - increase local configuration timeout for slow disks (3353)
4110 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4112 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4115 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4116 - don't overwrite extent policy data in reply if lock was blocked (2901)
4117 - drop filter export grants atomically with removal from device (2663)
4118 - del obd_self_export from work_list in class_disconnect_exports (2908)
4119 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4120 - swab reply message in mdc_close, other PPC fixes (2464)
4121 - fix destroying of named logs (2325)
4122 - overwrite old logs when running lconf --write_conf (2264)
4123 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4124 - fix race in target_handle_connect (2898)
4125 - mds_reint_create() should take same inode create lock (2926)
4126 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4127 - don't close files for self_export to avoid uninitialized obd (2936)
4128 - allow MDS with the same name as client node (2939)
4129 - hold dentry reference for closed log files for unlink (2325)
4130 - reserve space for all logs during transactions (2059)
4131 - don't evict page beyond end of stripe extent (2925)
4132 - don't oops on a deleted current working directory (2399)
4133 - handle hard links to targets without a parent properly (2517)
4134 - don't dereference NULL lock when racing during eviction (2867)
4135 - don't grow lock extents when lots of conflicting locks (2919)
4137 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4140 - account for cache space usage on clients to avoid data loss (974)
4141 - lfsck support in lustre kernel code (2349)
4142 - reduce journal credits needed for BRW writes (2370)
4143 - orphan handling to avoid losing space on client/server crashes
4144 - ptlrpcd can be blocked, stopping ALL progress (2477)
4145 - use lock value blocks to assist in proper KMS, faster stat (1021)
4146 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4147 - recovery for initial connections (2355)
4148 - fixes for mds_cleanup_orphans (1934)
4149 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4150 - block all file creations until orphan recovery completes (1901)
4151 - client remove rq_connection from request struct (2423)
4152 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4153 - recovery timer race (2670)
4154 - mdc_close recovey bug (2532)
4155 - ptlrpc cleanup bug (2710)
4156 - mds timeout on local locks (2588)
4157 - namespace lock held during RPCs (2431)
4158 - handle interrupted sync write properly (2503)
4159 - don't try to handle a message that hasn't been replied to (2699)
4160 - client assert failure during cleanup after abort recovery (2701)
4161 - leak mdc device after failed mount (2712)
4162 - ptlrpc_check_set allows timedout requests to complete (2714)
4163 - wait for inflight reqs when ptlrpcd finishes (2710)
4164 - make sure unregistered services are removed from the srv_list
4165 - reset bulk XID's when resending them (caught by 1138 test)
4166 - unregister_bulk after timeout
4167 - fix lconf error (2694)
4168 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4169 - readahead locks pages, leaves pending causing memory pressure (2673)
4170 - increase OST request buffers to 4096 on large machines (2729)
4171 - fix up permission of existing directories in simple_mkdir (2661)
4172 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4173 - don't assign transno to errored transactions (2742)
4174 - don't delete objects on OST if given a bogus objid from MDS (2751)
4175 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4176 - if rq_no_resend, then timeout request after recovery (2432)
4177 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4178 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4179 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4180 - MDS recovery shouldn't delete objects that it has given out (2730)
4181 - if enqueue arrives after completion, don't clobber LVB (2819)
4182 - don't unlock pages twice when trigger_group_io returns error (2814)
4183 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4184 - don't write pages to disk if there was an error (1450)
4185 - don't ping imports that have recovery disabled (2676)
4186 - take buffered bytes into account when balancing socknal conn (2817)
4187 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4188 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4189 - remove little-endian swabbing of llog records (1987)
4190 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4191 - retry reposting request buffers if they fail (1191)
4192 - grow extent at grant time to avoid granting a revoked lock (2809)
4193 - lock revoke doesn't evict page if covered by a second lock (2765)
4194 - disable VM readahead to avoid reading outside lock extents (2805)
4196 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4197 - updated kernel patches for hp-2.4.20 kernel (2681)
4199 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4202 - fix truncated write corruption (2366)
4203 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4205 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4206 - ensure len(uuid) < 37 in lmc (1171)
4207 - fix ia64 OOPS in llog_test (2255)
4208 - zero end of page at obdfilter for partial page writes (2648)
4209 - don't leave stale dentries around after renames (bug 2428)
4210 - fix timeouts when evicting a client with a single lock held (2642)
4211 - set deadline for the initial HELLO message to drain (2634)
4212 - print out dotted-quad IP addresses in the socknal (2302)
4214 - additional debugging for MDS client eviction problem (2443)
4215 - fix mkfsoptions support for osts (2603, 2604)
4217 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4220 - add series for the vanilla 2.6.0 kernel
4221 - add series for the vanilla 2.4.24 kernel
4222 - add series for a cray x86/64 UL kernel drop
4223 - fix xattr patches for the vanilla 2.4.19 series
4225 - generate true UUIDs in lmc (1171)
4226 - have portals stack dumping break in UML (2466)
4227 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4228 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4229 - wake callback waiting threads on client eviction (2460)
4230 - Add --ptldebug and --subsystem to lmc (1719)
4231 - update assertion to allow safe interrupt allocation
4232 - set rq_no_resend for cancel requests (2432)
4233 - recalculate ptlrpcd timeout after resend (2494)
4234 - call vfs_rmdir when removing pending directories (2368)
4235 - fix renaming a file to itself (2429)
4236 - lmc creates a default one-stripe lov (2454)
4237 - expand procfs space to handle large clusters (2326)
4238 - increase UML stack to avoid overflow
4239 - update lconf's list of debug and subsystem masks
4240 - fix lfs find --obd (2510)
4241 - /proc tunable for disabling filter read caching (2591)
4242 - stop rpm packages from altering slapd.conf (2301)
4243 - disable nagle in the socknal under 0conf (2578)
4244 - choose mds inode size based on stripe count (2572)
4245 - fix kernel-source rpm problems (2516)
4247 - add --disable-doc to avoid pdf generation (2421)
4248 - update documentation, tests, type-os, comments
4249 - avoid format warnings on ia64
4250 - remove the TOE NAL
4251 - tiny code cleanups by removing unused fields
4253 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4256 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4257 - remove the most blatant lies from BUILDING file (2371)
4258 - change default debug level to reasonable production setting
4259 - reduce client side cache size to reduce cache flush time
4260 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4261 - make TCP zerocopy and pinger support enabled by default (2476)
4262 - sync writes completed after process exits caused crashes (2319)
4263 - maintain correct mount count on the MDS (2356)
4264 - backout 1557, because 2316 wasn't really fixed
4265 - better file I/O statistics gathering in /proc
4266 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4267 - another kernel patch to fix zero-copy TCP function export
4268 - don't take duplicate lock when processing re-sent getattr (2420)
4269 - lctl uses obd_self_export instead of creating new conn (2353)
4270 - MDS/OST recovery case which requires object creation asserted (2425)
4271 - move lfs from /usr/sbin to /usr/bin in packages
4272 - fix race between mds_client_add and mds_client_free (2417)
4273 - use kmalloc instead of slabs in portals (2430)
4274 - don't create duplicate records when a failover MDS is present (2442)
4275 - remove unnecessary mount age check (2332)
4276 - don't remove directory inodes from locks prematurely (2451)
4277 - don't break if MDS service name is the same as hostname (2103)
4278 - fix races in client write RPC generation when cache full (2482)
4280 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4283 - remove now-unused request->rq_obd (278)
4284 - if an allocation fails, print out how much memory we've used (1933)
4285 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4286 - add the "configurable stack size" patch to most series files (1256)
4287 - ability to write large log records, for 100+ OST configs (2306)
4288 - fix NULL deref when filter_prep fails (2314)
4289 - fix operator precedence error in filter_sync
4290 - dynamic allocation of socknal TX descriptors (2315)
4291 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4292 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4293 - ensure that CWARN messages in Portals always get to the syslog
4294 - __init/__exit are not for prototype decls (ldlm_init/exit)
4295 - x86-64 compile warning fixes
4296 - fix gateway LMC keyword conflict (2318)
4297 - fix MDS lock inversions in getattr/reint paths (1844)
4298 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4299 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4300 - DLM race condition prevented some lock evictions (2328)
4301 - ENOMEM detection and retry on socknal sends (2230)
4302 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4303 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4304 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4305 - better error messages when a client is rejected during recovery (1505)
4306 - avoid cancelling locks which were never granted, after failure (2330)
4307 - fix i_sem/journal inversion in mds_client_add (2333)
4308 - fix truncate/getattr lock cycle deadlock (2334)
4309 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4310 - fix two rare exit paths which could leak an l_lock() ref (2321)
4311 - fencepost error in MDS/OST orphan recovery (2226)
4312 - make log record alignment 8 bytes (1988)
4313 - lstripe now fails when requested offset > ost_count (2237)
4314 - ensure that all kernel series have a complete list.h (1607)
4315 - fix crashes in special-file operations (2316)
4316 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4318 - allow configurable automake binary, for testing new versions
4319 - small update to the lfs documentation
4321 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4323 * fix negative export reference count in fsfilt_sync (2312)
4325 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4326 * release candidate 0.9.1
4328 - orphans are moved into the PENDING directory for possible recovery
4329 - replayed opens now open by fid for orphan/rename safety (1042)
4330 - last close of an orphan inode generates a transno (683)
4331 - chdir() and mount() now pin the directory entry (1020)
4332 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4333 - discard very old requests without processing them (1502)
4334 - remove some common, well-understood CERRORs (1505)
4335 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4336 - clear "grant" flags in OST replies until OST grant code lands (1644)
4337 - fix read performance by not clobbering i_blksize on client (1598)
4338 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4339 - make LRU size a /proc tunable, clears locks when reduced (707)
4340 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4341 - server locks take references on exports now (1558)
4342 - build fixes for 2.4.20-rh trees (1663)
4343 - return an error from lov_create if all OSCs are inactive (1751)
4344 - fix import levels when a reconnect happens without a timeout (1597)
4345 - exit early from mds_open if we get a lookup error (1749)
4346 - partial page read at EOF wouldn't wait for disk before sending (1642)
4347 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4348 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4349 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4350 - write/truncate lock inversion (1639)
4351 - Don't auto-load obdclass, portals modules during cleanup (1495)
4352 - fix timestamps from jumping to "now" (1763)
4353 - extra journal assertions (1648)
4354 - add an extra multiunlink test (1771)
4355 - fix read_record/write_record API (1776)
4356 - fix leak of offset_extent, possible incorrect i_size later (1772)
4357 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4358 - Debugging for the kqswnal_get_idle_tx problems (1820)
4359 - Allow recovery to be attempted multiple times (1536)
4360 - Write out MDS last_rcvd file after it is first created (1600)
4361 - Fix tx_descriptor leak in failed transmit situations (1827)
4362 - ext3 journaling fixes for assertion failure after IO error (1871)
4363 - class_export_put() on freed export after completion AST error (1896)
4364 - Fix revalidate looping in VFS (1322)
4365 - Don't access a freed export during MDS_REINT timeout (1521)
4366 - Add open-unlink recovery support on the MDS (1673,1764)
4367 - Return an error if no MDS data was read from last_rcvd (1946)
4368 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4369 - Don't setup a disk device that doesn't match exported UUID (317)
4370 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4371 - avoid committing NULL handle in force close
4372 - local.sh is now a one-stripe LOV configuration
4373 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4374 - fix ext3 htree duplicate directory entry corruption (1516)
4375 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4376 - update to new LOV EA format (2097)
4377 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4379 2003-06-15 Phil Schwan <phil@clusterfs.com>
4382 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4383 - per-import/export recovery handling (958, 931, 959)
4384 - multiple last-rcvd slots, for serving multiple FSes (949)
4385 - connections are again shared between multiple imp/exports (963, 964)
4386 - "umount -f" would hang if any requests needed to be sent (393, 978)
4387 - avoid pinning large req buffer by copying for queued messages (989)
4388 - add "uuid" to "lctl device" command to help upcalls (991)
4389 - "open" RPCs with transnos would confuse recovery counters (1037)
4390 - do proper endian conversion of all wire messages (288, 340, 891)
4391 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4392 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4393 - fix ptlrpc_connection leak in target_handle_connect (1174)
4394 - fix import refcounting bug in OST and MDS cleanup (1134)
4395 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4396 - fix very unlikely obd_types race condition (501)
4397 - remove osc_open hack for echo_client (1187)
4398 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4399 - a failure in read_inode2 leads to deadlock (1139)
4400 - cancel ack-locks as soon as transaction is committed (1072)
4401 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4402 - make sure to commitrw after any preprw to avoid deadlock (1162)
4403 - failing to execute a file in a lustre FS would lock inode (1203)
4404 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4405 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4406 - obd_uuid2tgt would walk off the end of the list (1255)
4407 - on IA64 the osc would give portals incorrect bulk size (1258)
4408 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4409 - fix lock inversion caused by new llite matching code (1282)
4410 - limit the number of dirty pages on a client to 10MB (1286)
4411 - timed out locks were not being corrected cancelled (1289)
4412 - fix O_DIRECT above 4GB on IA-32 (1292)
4413 * major user-visible changes
4414 - fail out/fail over policy now controlled by the upcall (993)
4416 - add OBD_PING to check server availability and failure (954)
4417 - lustre messages are now sent in sending host order (288, 340, 891)
4418 - add eadatalen to MDS getattr reply (340)
4419 - OST read replies may contain second buffer, with per-page status (593)
4421 2003-03-11 Phil Schwan <phil@clusterfs.com>
4424 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4425 - failed open()s could cause deadlock; fixed (867, 869)
4426 - stop cancelling OST locks when files are closed (481)
4427 - overlapping XID spaces caused network corruption (851, 853)
4428 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4429 - setattr_raw added, to do single-RPC, server-side setattrs
4430 - lmc/lconf syntax change for OST UUIDs
4431 - fix crashy race condition between ptlrpc_free_req and osc_close
4432 - don't use request in mdc_enqueue if we hit a timeout (889)
4433 - don't set the inode i_size for regular files from the MDS (896)
4434 - handle out of order completion AST (842)
4435 - don't LBUG if a lock request times out after receiving AST (913)
4436 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4437 - if a bad lock AST arrives, send an error instead of dropping entirely
4438 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4439 - fix leak in bulk IO when only partially completed (899, 900, 926)
4440 - fix O_DIRECT for ia64 (55)
4441 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4442 - C-z after timeout could hang a process forever; fixed (977)
4444 - client-side I/O cache (678, 924, 929, 941, 970)
4446 - READPAGE and SETATTRs which don't take server-side locks get
4449 2003-02-11 Phil Schwan <phil@clusterfs.com>
4452 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4453 - Fix fsfilt_extN_readpage() to read a full page of directory
4454 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4455 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4456 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4457 - Most intent operations were removed, in favour of a new RPC mode
4458 that does a single RPC to the server and bypasses most of the VFS
4459 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4460 - Aggressively cancel local locks on DLM servers
4461 - mds_reint_unlink sends EA to the client if it's the last nlink.
4462 client uses that EA to unlink OST objects.
4463 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4464 - recursive symlinks were fixed (439)
4465 - fixed NULL deref in DEBUG_REQ
4466 - filter_update_lastobjid no longer calls sync, which annoyed extN
4467 - fixed multi-client small-writes to a single file problem (445)
4468 - fixed mtime updates during file writes (607)
4469 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4470 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4471 - fixed Linux OST size reporting problem (444, 656)
4472 - OST now updates object mtime with writes or setattr (607, 619)
4473 - client verifies file size before zeroing page past EOF (445)
4474 - OST now writes last allocated objid to disk with allocation (108)
4475 - LOV on echo now works (409)
4477 - mds_reint_unlink sends a new buffer, with the EA included. this
4478 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4479 set if a regular file was being unlinked, and it was the last link
4480 - use PtlGet from the target for bulk writes (315)
4481 - OST now updates object mtime with writes or setattr (607, 619)
4482 - LDLM now has a grant-time callback to revalidate locked items, if
4484 - Many MDS operations were reorganized to combat race conditions
4486 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4487 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4489 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4492 - Fully reactivate OST imports after reconnection (512, others)
4493 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4494 - More graceful error handling for truncating on dead OST (515)
4495 - Don't error out unless we're actually accessing dead stripes (474)
4496 - Fix garbage sizes when stripes are missing (410)
4497 - LRU counters were broken, causing constant lock purge (433, 432)
4498 - garbage on read from stripes with failed OSTs (441)
4499 - mark OSCs as active before reconnecting during recovery (438)
4500 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4501 - lfind did not preserve OST order in output (443)
4502 - symlinks cause hung clients, incorrect data (439)
4503 - stop dereferencing request after dropping refcount (457)
4504 - don't LASSERT(spin_is_locked) on non-SMP (455)
4505 - fixes for many rename() bugs
4506 - fstat didn't correctly synchronize attributes (399)
4507 - server must handle lock cancellation during blocking AST prep (487)
4508 - bulk descriptors were free()d too soon (511)
4509 - fix paths in lconf, which would load incorrect modules (451, 507)
4510 - fix confusing lconf 'host not found' error message (386)
4511 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4512 - fix race condition in mdc_blocking_ast() for inode access (526)
4513 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4514 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4515 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4516 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4517 - fix lstripe in several ways (kernel side) (527)
4518 - fix request leak in ldlm_cli_enqueue (262)
4519 - incorrect OSC was marked inactive after OST failure
4520 - call mds_fs_cleanup before unmounting filesystem (524)
4521 - fix races between taking ns_lock and ldlm_lock_change_resource
4522 - fix races updating LOV export open file list
4523 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4524 - fix recovery NULL deref in ldlm_cli_cancel_unused
4525 - fix some DLM races by using new hash table for lock handles (419)
4526 - permit the client to specify desired inodes, at replay
4527 - duplicate requests when we queue them for replay reintegration
4528 - fix last_rcvd offset calculation
4529 - sync after each recovered transaction, so we always make progress
4530 - never, not always, ERESTART requests without transnos
4531 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4533 - skip replay if the MDS says that the client is already connected
4534 - don't check for a recovery-enabled export to match lctl's UUID
4535 - don't INC_USE_COUNT for phantom exports
4536 - don't crash when cleaning up phantom exports (567)
4537 - don't double-finish or set replay data for errored mdc_open requests
4538 - abort requests when they time out, so we don't get old replies
4539 - send/receive replies for AST messages again
4540 - if the client says that it doesn't have the lock, cancel it on the
4542 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4543 mark it as destroyed, it will all work out when decref is called
4544 - fix module use counts (22, 581)
4546 - ASTs now expect a reply (server cancels lock on error reply)
4548 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4551 - fix many simultaneous client startup (392)
4552 - fix dentry->d_it clobbering
4553 - credentials weren't being shipped for readdir/getattr operations
4554 - remove invalid assertions triggered during some concurrent MD
4556 - proper Lustre versions added (336, 389)
4557 - fix memory leak for create error case (398)
4558 - fix LOV locking bug that would get cli/srv out of sync
4559 - fix echo client over LOV (409)
4560 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4561 - fix double-O_EXCL intent crash (424)
4562 - avoid sending multiple lock CANCELs (352)
4564 - MDS can do multi-client recovery (modulo bugs in new code)
4566 - many updates, edits, cleanups
4568 2002-11-18 Phil Schwan <phil@clusterfs.com>
4571 - fix null d_it dereference (346)
4572 - fix full OST/dbench hang (333)
4573 - fix permission problem with file removal (286)
4574 - fix removal of OSCs from LOV when they fail
4575 - fix NULL deref during bulk timeout (214)
4576 - fix problems related to multiple filesystems on one MDS (241)
4577 - fixed serious subtle metadata locking bugs
4578 - free locks on clients when inodes are removed due to memory
4580 - fix inode pointer in lock data (285)
4581 - partial support for multiple MDS on a single host (241)
4582 - data locks weren't cancelled at clear_inode time (290, 311)
4583 - intent locks could lead to unbounded lock growth (205)
4584 - added a maximum lock count, an LRU list, and a flusher
4585 - fix multiple rename (365)
4586 - properly abstracted the echo client
4587 - OSC locked 1 byte too many; fixed
4588 - rewrote brw callback code:
4589 - fixed recovery bugs related to LOVs (306)
4590 - fixed too-many-pages-in-one-write crash (191)
4591 - fixed (again) crash in sync_io_timeout (214)
4592 - probably fixed callback-related race (385)
4594 - Add capability to MDS protocol
4595 - LDLM cancellations and callbacks on different portals
4597 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4600 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4601 - defer open object destruction to close time (601981, 138)
4602 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4603 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4604 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4605 - don't call obd_setattr in ll_file_release for destroyed objects
4606 * protocol change to lustre_msg: move |version| and add |flags|
4607 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4608 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4609 * added replay of create, unlink, link and rename operations during
4610 MDS failover; recovery should be much more robust now
4611 * remove failed OSCs from LOVs (only lov_create uses this so far)
4612 * the lustre-HOWTO was brought (more) up to date (582544)
4614 2002-10-23 Phil Schwan <phil@clusterfs.com>
4617 - in-use dentries weren't being reused properly (617851)
4618 - prevent multiple LDLM setup (599178)
4619 - fix LOV size calculations for truncate (617853)
4620 - fix client handling of MDS intent errors (POSIX)
4621 - fix permission bug in lovstripe.c test (624321)
4622 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4623 - truncate past end of file could corrupt data
4624 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4625 - a race in recovery could return ETIMEDOUT to apps (623947)
4626 - building outside the source directory was fixed
4627 * the lustre-HOWTO was brought (more) up to date (582544)
4628 * major progress was made on recovery functionality
4630 2002-10-10 Phil Schwan <phil@clusterfs.com>
4633 - recovery deadlock fix
4634 - rm -rf causes LBUG fix (617817)
4635 - file open by multiple tasks fix (618962)
4636 - directory permissions bugs (602707 and 620007)
4637 - journal_stop fixed with locking (611313)
4638 - O_APPEND failures resolved (618273, perhaps 614459)
4639 - lconf PATH fix (619770)
4640 - IA64 build fix (621450)
4641 - RPC buffer sizes scale with amount of memory
4643 2002-10-01 Phil Schwan <phil@clusterfs.com>
4646 - locks would be cancelled without throwing away data pages,
4647 resulting in inconsistent data (605627)
4648 - inode attributes were not always being refreshed (605627, 612449)
4649 - lconf now continues to cleanup after lctl reports an error
4650 - MDS now enforces user permissions (602707)
4651 - lprocfs cleanup fixed, but not yet enabled (614157)
4652 - fixed infinite server hang, should a client not respond to an AST
4653 - avoid going into recovery if user calls readlink() with a buffer
4654 that's too small (613941)
4655 - AST RPCs no longer require replies (614867) -- this may be changed
4656 - don't crash server if client sends an IOV that's too big (611336)
4657 - fixed lock conversion deadlock (611892)
4658 - fixed the following of symlinks (614622)
4659 * recovery: the server can remove locks from a client that dies, other
4660 clients can make progress
4661 * more extN patch fixes
4662 * compile-time configurable ptlrpc buffer allocations
4664 - collaborative read cache document
4665 - Lustre Lite Performance CDR document-in-progress
4667 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4670 - fix typo in patch-2.4.18
4672 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4675 - clear ptlrpc request each time in handle_incoming_request()
4676 - unlink of files now destroys the object on the OST
4678 2002-09-19 Peter Braam <braam@clusterfs.com>
4680 * add hard link support
4681 * change obdfile creation method
4682 * kernel patch changed
4684 2002-09-19 Peter Braam <braam@clusterfs.com>
4687 - stack overflow bug in extN fixed
4689 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4691 * documentation updates
4692 - add man pages for config tools
4693 - update tests/README to describe testing with new config tools
4694 - finish metadata API descriptions
4695 * bug fixes and cleanups
4696 - statfs workaround for 16TB limit
4697 - LOV stripe allocation improved, can stripe on subset of OSTs
4698 - LOV file size/IO offset was wrong for files > 4GB in size
4699 - object EA data was being dropped, caused files to be unreadable
4700 - memory overflow with non-LOV OST caused memory corruption
4701 - fixed regression tests to work with new config tools, obdfilter
4702 - fixed bug when directory size became larger than 1 block
4703 - fixed bug (for single client case) when PWD was deleted
4704 - invalidate local directory pages when doing intent-based ops
4705 - avoid LDLM oops when lock callback contained bad data
4707 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4709 * documentation updates
4710 * bug fixes and cleanups
4711 - configuration tools
4714 - 64-bit compile warnings
4715 - 64-bit internal statfs data
4717 * test_brw on persistent OST devices
4719 * lprocfs (disabled)
4721 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4723 * documentation updates
4724 * bug fixes and cleanups
4725 * configuration tools
4727 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4730 * many small fixes to 0_5_4
4731 * io/network handling
4732 * thinkos in MDS operations
4734 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4737 * crucial basic fixes to 0.5.3
4738 * IOR, Iozone work over Elan
4741 2002-08-07 Phil Schwan <phil@clusterfs.com>
4742 * version 0_5_3, our first alpha
4743 * we use the new Portals iovs
4744 * documentation updates
4745 * bug fixes and cleanups
4746 * small changes in the DLM wire protocol
4748 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4749 * version 0_5_1 with some initial stability,
4750 * locking on MD and file I/O.
4751 * documentation updates
4752 * several bug fixes since 0.5.0
4753 * small changes in wire protocol
4755 2002-07-18 Phil Schwan <phil@clusterfs.com>
4757 * delivered as Lustre Light Alpha
4758 * fixed a crash after handling invalid MDS requests
4759 * fixed directory pages for architectures with non-4k pages sizes
4761 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4762 * release version v0_4_4
4763 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4765 * Fixes a number of interruption problems with OST operations.
4766 * Update documentation for portals header changes
4767 * Move all wire protocol structs/defines to lustre_idl.h
4768 * Fixes symlink length bug.
4769 * Add tcpdump to repository.
4771 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4772 * release version v0_4_3
4773 * Fixes statfs for inodes on extN.
4774 * Fixes bug in runtests which would delete /etc/hosts.
4775 * Use 64-bit object IDs wherever possible (not into VFS though)
4776 Remove ost_get_info, which is unused by lustre, and out of date.
4778 2002-07-03 Peter Braam <braam@clusterfs.com>
4779 * release version v0_4_2 Fixes a lookup error (type not passed)
4780 * move forward to head of Portals
4781 * move forward to latest Lustre kernel
4783 2002-06-25 Peter Braam <braam@clusterfs.com>
4784 * release version v0_4_1. Hopefully stable on single node use.