1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
21 Frequency : rare, on shutdown ost
22 Description: don't hit live lock with umount ost.
23 Description: shrink_dcache_parent can be in long loop with destroy dentries,
24 use shrink_dcache_sb instead.
28 Description: don't panic with use echo client
29 Details : echo client pass NULL as client nid pointer and this produce null
34 Description: Stack overflow during MDS log replay
35 Details : ease stack pressure by using a thread dealing llog_process.
39 Description: fix build on ppc32
40 Details : compile code with -m64 flag produce wrong object file for ppc32.
44 Description: add message levels for liblustreapi
48 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
49 Details : Move the 'good_osts' check before the 'total_bavail' check. This
50 will result in an -EAGAIN and in the exit call path we call
51 alloc_rr() which will with increasing aggressiveness attempt to
52 aquire precreated objects on the minimum number of required OSCs.
56 Description: Use old size assignment to avoid deadlock
57 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
58 the scheduling while holding a spinlock. We do not need locking
59 for size in ll_update_inode() because size is only updated from
60 the MDS for directories or files without objects, so there is no
61 other place to do the update, and concurrent access to such inodes
62 are protected by the inode lock.
66 Description: resolve "_IOWR redefined" build error on SLES10
70 Description: dump the memory debugging after all modules are unloaded to
71 suppress false negative in conf_sanity test 39
73 Severity : enhancement
75 Description: build kernel-ib packages for OFED 1.3 in our release cycle
80 Description: fix SLES kernel versioning
81 Details : the kernel version for our SLES 10 kernel did not include a "-"
82 before the "smp" at the end. while this was not a problem in
83 general, it did mean that software trying to use the kernel
84 version to try to detect a vendor specific kernel would fail.
85 this was most evident by the OFED build scripts.
89 Description: Don't update lov_desc members until making sure they are valid
90 Details : When updating lov_desc members via proc fs, need fix their
91 validities before doing the real update.
95 Description: don't put request into delay list while invalidate in flight.
96 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
97 import in flight. this produce timeout for invalidate and sometimes
102 Frequency : on ppc only
103 Description: not convert ost objects for directory because it's not exist.
104 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
108 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
110 Severity : enhancement
112 Description: Update to support 2.6.22.14 vanilla kernel.
116 Frequency : rare, on recovery
117 Description: read procfs can produce deadlock in some situation
118 Details : Holding lprocfs lock which send rpc can produce block for destroy
119 obd objects and this also block reconnect with -EALREADY. This isn't
120 fix all lprocfs bugs - but make it rare.
122 Severity : enhancement
124 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
127 Frequency : frequent on X2 node
129 Description: mdc_set_open_replay_data LBUG
130 Details : Set replay data for requests that are eligible for replay.
134 Description: lustre_mgs: operation 101 on unconnected MGS
135 Details : When MGC is disconnected from MGS long enough, MGS will evict the
136 MGC, and late on MGC cannot successfully connect to MGS and a lot
137 of the error messages complaining that MGS is not connected.
140 Frequency : on start mds
142 Description: Implement get_info(last_id) in obdfilter.
145 Frequency : occasional
147 Description: Correctly check stale fid, not start epoch if ost not support SOM
148 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
149 overwrited with new generated fid, but mds can anwer with one of these
150 two fids and both is not stale. setattr incorectly start epoch and
151 assume will be called done_writeting, but without SOM done_writing
155 Frequency : rare, depends on device drivers and load
157 Description: MDS or OSS nodes crash due to stack overflow
158 Details : Code changes in 1.8.0 increased the stack usage of some functions.
159 In some cases, in conjunction with device drivers that use a lot
160 of stack the MDS (or possibly OSS) service threads could overflow
161 the stack. One change which was identified to consume additional
162 stack has been reworked to avoid the extra stack usage.
165 Frequency : occasional
167 Description: Do not fail import if osc_interpret_create gets -EAGAIN
168 Details : If osc_interpret_create got -EAGAIN it immediately exits and
169 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
170 oscc_has_objects and see OSC has no objests and call
171 oscc_internal_create to resend create request.
173 Severity : enhancement
175 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
177 Severity : enhancement
179 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
182 Frequency : very rare
184 Description: Fix replay if there is an un-replied request and open
185 Details : In some cases, older replay request will revert the
186 mcd->mcd_last_xid on MDS which is used to record the client's
189 Severity : enhancement
191 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
193 Severity : enhancement
195 Description: Add rhel5 support to HEAD.
197 Serverity : enhancement
199 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
204 Description: Don't allow skipping OSTs if index has been specified.
205 Details : Don't allow skipping OSTs if index has been specified, make locking
206 in internal create lots better.
210 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
211 Details : access to bitfield in structure is always rounded to long
212 and this produce problem with not atomic change any bit.
216 Description: Lustre make rpms failed.
217 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
218 builds Lustre rpms from tarball.
221 Frequency : rare on shutdown ost
223 Description: If llog cancel was not send before clean_exports phase, this can
224 produce deadlock in llog code.
225 Details : If llog thread has last reference to obd and call class_import_put
226 this produce deadlock because llog_cleanup_commit_master wait when
227 last llog_commit_thread exited, but this never success because was
228 called from llog_commit_thread.
232 Description: allow userland application know is lost one of stripes.
233 Details : fill lvb_blocks with error code on ost and return it to
234 application if error flag found.
238 Description: NULL lov_tgts causing MDS oops
239 Details : more safe checks for NULL lov_tgts for avoid oops.
241 Severity : enhancement
243 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
247 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
248 Details : If an OST is down the MDS will hang indefinitely in
249 obd_statfs_rqset() waiting for the statfs data. While for
250 MDS QOS usage of statfs, it should not stuck in waiting.
252 Severity : enhancement
254 Description: remote_acl support
255 Details : Support ACL-based permission check for remote user.
256 Support setfacl/getfacl for remote user with the utils
257 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
258 the system "{s,g}etfacl" utils.
260 Severity : enhancement
262 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
264 Severity : enhancement
266 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
270 Description: make mgs_setparam() handle fsname containing dash
271 Details : fsname containing a dash does not work with lctl conf_param
273 Severity : enhancement
275 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
277 Severity : enhancement
279 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
283 Frequency : rare, at shutdown
284 Description: access already free / zero obd_namespace.
285 Details : if client_disconnect_export was called without force flag set,
286 and exist connect request in flight, this can produce access to
287 NULL pointer (or already free pointer) when connect_interpret
288 store ocd flags in obd_namespace.
292 Frequency : only at startup
293 Description: not alloc memory with spinlock held.
294 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
295 if any spinlock held.
297 Severity : enhancement
299 Description: make lustre randomly fail allocating memory
300 Details : Make lustre randomly failed allocating memory for testing purpose.
302 Severity : enhancement
304 Description: lost problems with lov objid file
305 Details : Fixes some scability and access to not inited memory problems
306 in work with lov objdid file.
311 Description: lfs find does not continue on file error
312 Details : Continue other files processing when a file/dir is absent.
316 Description: Inconsistent usage of lustre_pack_reply()
317 Details : Standardize the usage of lustre_pack_reply() such that it
318 always generate a CERROR on failure.
321 Frequency : occasional
323 Description: MDS hang or stay in waiting lock
324 Details : If client receive lock with CBPENDING flag ldlm need send lock
325 cancel as separate rpc, to avoid situation when cancel request
326 can't processed due all i/o threads stay in wait lock.
330 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
334 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
338 Description: Update to SLES9 kernel-2.6.5-7.287.3.
342 Description: Update to RHEL4 latest kernel.
344 Severity : enhancement
346 Description: Build SLES10 patchless client fails
347 Details : The configure was broken by run ./configure with
348 --with-linux-obj=.... argument for patchless client. When the
349 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
350 can't search header adequately. Use absolute path such as
351 -I($LINUX)/include instead.
355 Description: interrupt oig_wait produce painc on resend.
356 Details : brw_redo_request can be used for resend requests from ptlrpcd and
357 private set, and this produce situation when rq_ptlrpcd_data not
358 copyed to new allocated request and triggered LBUG on assert
359 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
360 ptlrpcd set if request is changed and can be safety changed to use
365 Description: LASSERT_{REQ,REP}SWAB macros are buggy
366 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
367 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
368 these macros and replace them with their difinitions instead.
372 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
373 Details : Update which_patch & target file for SLES10 latest kernel.
376 Description: Remove client patches from SLES 10 kernel.
377 Details : This causes SLES 10 clients to behave as patchless clients
378 even on a Lustre-patched (server) kernel.
380 Severity : enhancement
382 Description: self-adjustable client's lru lists
383 Details : use adaptive algorithm for managing client cached locks lru
384 lists according to current server load, other client's work
385 pattern, memory activities, etc. Both, server and client
386 side namespaces provide number of proc tunables for controlling
389 Severity : enhancement
391 Description: light-weight GSS support
392 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
393 and reduce performance overhead.
395 Severity : enhancement
397 Description: Linux keyring support
398 Details : Support using service of Linux keyring for Lustre GSS internal
399 context refresh/cache mechanism.
403 Description: Fix errors in lfs documentation
404 Details : Fixes man pages
408 Description: don't use GFP_* in generic Lustre code.
409 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
413 Description: obdclass is limited by single OBD_ALLOC(idarray)
414 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
418 Description: warnings with build patchless client with vanila 2.6.19 and up
419 Details : change old ctl_table style and replace ctl_table/ctl_table_header
420 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
424 Frequency : Only for SLES
425 Description: sanity_quota fail test_1
426 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
427 file, which makes TSTID[2] inunique.
431 Frequency : Only for SLES9
432 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
433 Details : Update target/ChangeLog/which_patch .
437 Frequency : Only for test_10 in sanity-quota.sh
438 Description: a bug in quota test code
439 Details : checking a test flag in a wrong place
443 Frequency : Only for test_13 in sanity-quota.sh
444 Description: a bug in quota test script
445 Details : shouldn't check overfull allocated quota
449 Description: jbd statistics
450 Details : Port older jbd statistics patch for sles10
452 Severity : enhancement
454 Description: Add printing inode info into message about error in writepage.
459 Description: reply_lock_interpret crash due to race with it and lock cancel.
460 Details : Do not replay locks that are being cancelled. Do not reference
461 locks by their address during replay, just by their handle.
465 Frequency : When flocks are used.
466 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
467 Details : Flock locks might destroy just granted lock if it could be merged
468 with another existing flock, this is done in completion handler,
469 so teach ldlm_cli_enquque_fini that this is a valid case for
475 Description: Oops in read and write path when failing to allocate lock.
476 Details : Check if lock allocation failed and return error back.
480 Description: lstripe command fails for valid OST index
481 Details : The stripe offset is compared to lov->desc.ld_tgt_count
482 instead of lov->desc.ld_active_tgt_count.
486 Description: sanity.sh failed test 103
487 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
492 Description: df doesn't work properly if diskfs blocksize != 4K
493 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
497 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
498 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
502 Description: obd_health_check_timeout too short
503 Details : set obd_health_check_timeout as 1.5x of obd_timeout
507 Description: llapi_file_create() does not allow some changes
508 Details : add llapi_file_open() that allows specifying the mode and
509 open flags, and also returns an open file handle.
513 Description: merge and cleanup kernel patches.
514 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
518 Description: Add journal checksum support.(Kernel part)
519 Details : The journal checksum feature adds two new flags i.e
520 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
521 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
522 indicates that the commit block contains the checksum for
523 the blocks described by the descriptor blocks. Now commit
524 record can be sent to disk without waiting for descriptor
525 blocks to be written to disk. This behavior is controlled
526 using JBD2_FEATURE_ASYNC_COMMIT flag.
530 Description: OSS needs mutliple precreate threads
531 Details : Add ability to start more than one create thread per OSS.
535 Description: Sanity.sh test_65k failed: setstripe should have succeeded
536 Details : execute lfs setstripe on client
540 Description: mds_obd_create error creating tmp object
541 Details : When the user sets quota on root, llog will be affected and can't
542 create files and write files.
545 Frequency : Always on ia64 patchless client, and possibly others.
547 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
548 Details : This allows the patchless client to be loaded on architectures
553 Description: RedHat Update kernel for RHEL5
554 Details : Add the kernel config file for RHEL5.
558 Description: RedHat Update kernel for RHEL5
559 Details : Modify the kernel config file more closer RHEL5.
563 Description: Build failure against Centos5 (RHEL5)
564 Details : Define PAGE_SIZE when it isn't present.
568 Description: test_7 on sanity-quota.sh can't be used on separate machines
569 Details : add facet to handle it
573 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
574 Details : with NFS, the anon dentry's parent was set to itself in
575 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
576 corresponding dentry other than use rec->ur_name.
578 Severity : enhancement
580 Description: lfs setstripe enhancement
581 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
585 Description: enable data checksumming by default
586 Details : enable checksum by default, allow --disable-checksum
587 configure option and "-o nochecksum" mount option. Checksums
588 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
592 Description: lustre support for RHEL5
593 Details : Add support for RHEL5.
597 Description: Client eviction due to failover config
598 Details : after a connection loss, the lustre client should attempt to
599 reconnect to the last active server first before trying the
600 other potential connections.
604 Description: when mds and osts use different quota unit(32bit and 64bit),
605 quota will be released repeatly.
606 Details : void sending multiple quota reqs to mds, which will keep the
607 status between the reqs.
611 Description: osts not allocated evenly to files
612 Details : change the condition to increase offset_idx
615 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
617 Description: Data corruption for OSTs that are formatted larger than 2TB
619 Details : When generating the bio request for lustre file writes the
620 sector number would overflow a temporary variable before being
621 used for the IO. The data reads correctly from Lustre (which
622 will overflow in a similar manner) but other file data or
623 filesystem metadata may be corrupted in some cases.
627 Description: Tune the kernel for good SCSI performance.
628 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
629 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
634 Description: rewrite ext2-derived code in obdclass/uuid.c
635 Details : rewrite inherited code (uuid parsing code from ext2 utils)
636 from scratch preserving functionality.
640 Description: "lfs find -obd UUID" prints directories
641 Details : "lfs find -obd UUID" will return all directory names instead
642 of just file names. It is incorrect because the directories
643 do not reside on the OSTs.
647 Description: use i_size_read and i_size_write in 2.6 port
648 Details : replace inode->i_size access with i_size_read/write()
651 Frequency : when removing large files
653 Description: scheduling issue during removal of large Lustre files
654 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
655 It causes scheduling issues when removing large files (17TB in the
659 Frequency : only with liblustre clients on XT3
661 Description: evictions taking too long
662 Details : allow llrd to evict clients directly on OSTs
665 Frequency : only on ppc
667 Description: /proc/fs/lustre/devices broken on ppc
668 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
669 We should make sure the type of 'index' is loff_t and then cast
670 explicitly as needed below. Do not assign an explicitly cast
674 Frequency : only for rhel5
676 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
677 Details : Modify the target file & which_kernel.
679 Severity : enhancement
681 Description: omit set fsid for export NFS
682 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
683 as FS_REQUIRES_DEV for avoid problems with generate fsid.
687 Frequency : Always, for kernels after 2.6.16
688 Description: Fix warning idr_remove called for id=.. which is not allocated.
689 Details : Last kernels save old s_dev before kill super and not allow
690 to restore from callback - restore it before call kill_anon_super.
694 Description: buffer overruns could theoretically occur
695 Details : llapi_semantic_traverse() modifies the "path" argument by
696 appending values to the end of the origin string, and a
697 overrun may occur. Adding buffer overrun check in liblustreapi.
701 Description: Fix error on 'ls .' at the top of the Lustre mount.
702 Details : Don't revalidate dentry if it is a root dentry.
706 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
707 Details : Modify vm-tunables-rhel4.patch.
711 Description: Kernel config for 2.6.18-vanilla.
712 Details : Modify targets/2.6-vanilla.target.in.
713 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
714 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
715 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
716 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
720 Description: improve handling recoverable errors
721 Details : if request processig with error which can be recoverable on server
722 request should be resend, otherwise page released from cache and
727 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
728 Details : Modify target file & which_patch.
729 A flaw was found in the IA32 system call emulation provided
730 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
731 value could be stored in the %RAX register, which could trigger an
732 out-of-bounds system call table access. An untrusted local user
733 could exploit this flaw to run code in the kernel
734 (ie a root privilege escalation). (CVE-2007-4573).
738 Description: O_DIRECT bypasses client statistics.
739 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
740 read_ahead_stats were updated. copied the stats section from
741 osc_send_oap_rpc() into async_internal().
745 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
749 Description: change order of libsysio includes
750 Details : '#include sysio.h' should always come before '#include xtio.h'
754 Description: handle "serious error: objid * already exists" more gracefully
755 Details : If LAST_ID value on disk is smaller than the objects existing in
756 the O/0/d* directories, it indicates disk corruption and causes an
757 LBUG(). If the object is 0-length, then we should use the existing
758 object. This will help to avoid a full fsck in most cases.
760 Severity : enhancement
762 Description: adapt the lustre_config script to support the upgrade case
763 Details : Add "-u" option for lustre_config script to support upgrading 1.4
764 server targets to 1.6 in parallel.
768 Description: To avoid grant space > avaible space when the disk is almost
769 full. Without this patch you might see the error "grant XXXX >
770 available" or some LBUG about grant, when the disk is almost
772 Details : In filter_check_grant, for non_grant cache write, we should
773 check the left space by if (*left > ungranted + bytes), instead
774 of (*left > ungranted), because only we are sure the left space
775 is enough for another "bytes", then the ungrant space should be
776 increase. In client, we should update cl_avail_grant only there
777 is OBD_MD_FLGRANT in the reply.
781 Description: Update RHEL 4 kernel to fix local root privilege escalation.
782 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
783 described in CVE-2007-4573. This problem could allow untrusted
784 local users to gain root access.
787 Frequency : when using O_DIRECT and quotas
789 Description: Incorrect file ownership on O_DIRECT output files
790 Details : block usage reported by 'lfs quota' does not take into account
791 files that have been written with O_DIRECT.
796 Description: touch file failed when fs is not full
797 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
798 otherwise we can get ENOSP while fs is not full.
802 Description: parallel lock callbacks
803 Details : Instead of sending blocking and completion callbacks as separated
804 requests, adding them to a set and sending in parallel.
807 Frequency : only for Cray XT3
808 Bugzilla : 12829/13455
809 Description: Changing primary group doesn't change the group lustre assigns to
811 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
812 overridden with the primary group provided by the group upcall,
813 whereas the supplied fsgid can be trusted if it is in the list of
814 supplementary groups returned by the group upcall.
816 Severity : enhancement
818 Description: Allow masking D_WARNING, D_ERROR messages from console
819 Details : Console messages can now be disabled via lnet.printk.
823 Description: User code with malformed file open parameter crashes client node
824 Details : Before packing join_file req, all the related reference should be
825 checked carefully in case some malformed flags cause fake
826 join_file req on client.
830 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
831 Details : ldlm_completion_ast() assumes that a lock is granted when the req
832 mode is equal to the granted mode. However, it should also check
833 that LDLM_FL_CP_REQD is not set.
837 Description: Heavy nfs access might result in deadlocks
838 Details : After ELC code landed, it is now improper to enqueue any mds
839 locks under och_sem, because enqueue might want to decide to
840 cancel open locks for same inode we are holding och_sem for.
844 Description: Client eviction while running blogbench
845 Details : A lot of unlink operations with concurrent I/O can lead to a
846 deadlock causing evictions. To address the problem, the number of
847 oustanding OST_DESTROY requests is now throttled to
848 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
849 callbacks are processed in priority.
853 Description: enable ACLs on MDS by default
854 Details : ACLs must be enabled on MDS by default.
857 Frequency : PPC/PPC64 only
859 Description: conflicts between asm-ppc64/types.h and lustre_types.h
860 Details : fix duplicated definitions between asm-ppc64/types.h and
861 lustre_types.h on PPC.
864 Frequency : PPC/PPC64 only
866 Description: asm-ppc/segment.h does not exist
867 Details : fix compile issue on PPC.
871 Description: better handle error messages in extents code
874 Frequency : RHEL4 only
876 Description: mkfs is very slow on IA64/RHEL4
877 Details : A performance regression has been discovered in the MPT Fusion
878 driver between versions 3.02.73rh and 3.02.99.00rh. As a
879 consequence, we have downgraded the MPT Fusion driver in the RHEL4
880 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
882 Severity : enhancement
884 Description: SNMP support enhancement
885 Details : Adding total number of sampled request for an MDS node in snmp
888 Severity : enhancement
890 Description: Optimize ldlm waiting list processing for PR extent locks
891 Details : When processing waiting list for read extent lock and meeting read
892 lock that is same or wider to it that is not contended, skip
893 processing rest of the list and immediatelly return current
894 status of conflictness, since we are guaranteed there are no
895 conflicting locks in the rest of the list.
899 Description: Time out and refuse to reconnect
900 Details : When the failover node is the primary node, it is possible
901 to have two identical connections in imp_conn_list. We must
902 compare not conn's pointers but NIDs, otherwise we can defeat
903 connection throttling.
907 Description: port llog fixes from b1_6 into HEAD
908 Details : Port llog reference couting and some llog cleanups from b1_6
909 (bug 10800) into HEAD, for protect from panic and access to already
910 free llog structures.
914 Description: Detect stride IO mode in read-ahead
915 Details : When a client does stride read, read-ahead should detect that and
916 read-ahead pages according to the detected stride pattern.
920 Description: data checksumming impacts single node performance
921 Details : add support for several checksum algorithm. Currently, only CRC32
922 and Adler-32 are supported. The checksum type can be changed on
923 the fly via /proc/fs/lustre/osc/*/checksum_type.
927 Description: use adler32 for page checksums
928 Details : when available, use the Adler-32 algorithm instead of CRC32 for
933 Description: build for x2 fails
934 Details : fix compile issue on Cray systems.
938 Description: Properly match for duplicate locks
939 Details : Due to different lock order from skiplists code, we need to
940 traverse entire list for now
943 Frequency : only on PPC/SLES10
945 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
946 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
947 be defined. Add a hack in mkfs_lustre.c to work around this compile
952 Description: LASSERT on MDS when client holding flock lock dies
953 Details : ldlm pool logic depends on number of granted locks equal to
954 number of released locks which is not true for flock locks, so
955 just exclude such locks from consideration.
959 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
960 Details : Use fsfilt_sync() for both the whole filesystem sync and
961 individual file sync to eliminate dangerous inode locking
962 with I_LOCK that can lead to a deadlock.
966 Description: performance in 1.6.3
967 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
968 sure that 1MB requests are not fragmented by the block layer.
970 Severity : enhancement
972 Description: organize the server-side client stats on per-nid basis
973 Details : Change the structure of stats under obdfilter and mds to
981 The "uuid"s file would list the uuids of _active_ exports.
982 And the clear entry is to clear all stats and stale nids.
984 Severity : enhancement
986 Description: eliminate client locks in face of contention
987 Details : file contention detection and lockless i/o implementation
990 --------------------------------------------------------------------------------
992 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
994 * Support for kernels:
995 2.6.5-7.283 (SLES 9),
996 2.6.9-55.EL (RHEL 4),
997 2.6.16.46-0.14 (SLES 10),
998 2.6.18.8 vanilla (kernel.org)
999 * Client support for unpatched kernels:
1000 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1001 2.6.16 - 2.6.22 vanilla (kernel.org)
1002 * Due to recently discovered recovery problems, we do not recommend
1003 using patchless RHEL 4 clients with this or any earlier release.
1004 * Recommended e2fsprogs version: 1.39.cfs8
1005 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1006 * Starting with this release, the ldiskfs backing filesystem required
1007 by Lustre is now in its own package, lustre-ldiskfs. This package
1008 should be installed. It is versioned separately from Lustre and
1009 may be released separately in future.
1013 Description: block reactivating mgc import until all deactivates complete
1014 Details : Fix race when failing back MDT/MGS to itself (testing)
1017 Frequency : at statup only
1019 Description: mds_lov_synchronize race leads to various problems
1020 Details : simultaneous MDT->OST connections at startup can cause the
1021 sync to abort, leaving the OSC in a bad state.
1023 Severity : enhancement
1025 Description: add optional extra BUILD_VERSION info
1026 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1027 to override the lustre version.
1030 Frequency : 2.6.18 servers only
1032 Description: ll_kern_mount() doesn't release the module reference
1033 Details : The ldiskfs module reference count never drops down to 0
1034 because ll_kern_mount() doesn't release the module reference.
1039 Description: server LBUG when using old ost_num_threads parameter
1040 Details : Accept the old ost_num_threads parameter but warn that it
1041 is deprecated, and fix an off-by-one error that caused an LBUG.
1046 Description: Transient SCSI error results in persistent IO issue
1047 Details : iobuf->dr_error is not reinitialized to 0 between two
1051 Frequency : sometimes when underlying device returns I/O errors
1053 Description: OSTs not going read-only during write failures
1054 Details : OSTs are not remounted read-only when the journal commit threads
1055 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1056 instead of the ext3 wrappers.
1060 Description: poor connect scaling with increasing client count
1061 Details : Don't run filter_grant_sanity_check for more than 100 exports
1062 to improve scaling for large numbers of clients.
1065 Frequency : SLES10 only
1067 Description: sanity-quota.sh quotacheck failed: rc = -22
1068 Details : Quotas cannot be enabled on SLES10.
1071 Frequency : liblustre clients only
1073 Description: getdirentries does not give error when run on compute nodes
1074 Details : getdirentries does not fail when the size specified as an argument
1075 is too small to contain at least one entry
1077 Severity : enhancement
1079 Description: Add LNET router traceability for debug purposes
1080 Details : If a checksum failure occurs with a router as part of the
1081 IO path, the NID of the last router that forwarded the bulk data
1082 is printed so it can be identified.
1087 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1088 Details : Due to a race condition, liblustre clients were occasionally
1089 evicted incorrectly.
1091 Severity : enhancement
1093 Description: lfs setstripe use optional parameters instead of postional
1096 Severity : enhancement
1098 Description: Nanosecond timestamp support for ldiskfs
1099 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1100 resolution timestamps. There is not yet support for this at
1101 the Lustre filesystem level.
1104 Frequency : during server recovery
1106 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1107 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1111 Frequency : nfs export on patchless client
1113 Description: connectathon hang when test nfs export over patchless client
1114 Details : Disconnected dentry cannot be found with lookup, so we do not need
1115 to unhash it or make it invalid
1118 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1119 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1120 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1125 Description: open req refcounting wrong on reconnect
1126 Details : If reconnect happened between getting open reply from server and
1127 call to mdc_set_replay_data in ll_file_open, we will schedule
1128 replay for unreferenced request that we are about to free.
1129 Subsequent close will crash in variety of ways.
1130 Check that request is still eligible for replay in
1131 mdc_set_replay_data().
1136 Description: disable writes to filesystem when reading health_check file
1137 Details : the default for reading the health_check proc file has changed
1138 to NOT do a journal transaction and write to disk, because this
1139 can cause reads of the /proc file to hang and block HA state
1140 checking on a healthy but otherwise heavily loaded system. It
1141 is possible to return to the previous behaviour during configure
1142 with --enable-health-write.
1144 Severity : enhancement
1146 Description: 64-bit inode version
1147 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1148 to the inode. This will be required for version-based recovery.
1153 Description: MDS fails to start if a duplicate client export is detected
1154 Details : in some rare cases it was possible for a client to connect to
1155 an MDS multiple times. Upon recovery the MDS would detect this
1156 and fail during startup. Handle this more gracefully.
1158 Severity : enhancement
1160 Description: Add -o localflock option to simulate old noflock
1162 Details : This will achieve local-only flock/fcntl locks
1168 Description: log_commit_thread vs filter_destroy race leads to crash
1169 Details : Take import reference before releasing llog record semaphore
1174 Description: Wrong request locking in request set processing
1175 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1176 imp_delayed_list, in this place should be used imp_lock.
1179 Frequency : when reconnection
1181 Description: Grant Leak when osc reconnect to OST
1182 Details : When osc reconnect ost, OST(filter) should check whether it
1183 should grant more space to client by comparing fed_grant and
1184 cl_avail_grant, and return the granted space to client instead
1185 of "new granted" space, because client will call osc_init_grant
1186 to update the client grant space info.
1189 Frequency : when client reconnect to OST
1191 Description: Grant Leak when osc do resend and replay bulk write
1192 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1193 bulk write request, because the grant info while be sync between
1194 OSC and OST when reconnect, and we should ignore the grant info
1195 these of resend/replay write req.
1200 Description: Grant space more than avaiable left space sometimes.
1201 Details : When then OST is about to be full, if two bulk writing from
1202 different clients came to OST. Accord the avaliable space of the
1203 OST, the first req should be permitted, and the second one
1204 should be denied by ENOSPC. But if the seconde arrived before
1205 the first one is commited. The OST might wrongly permit second
1206 writing, which will cause grant space > avaiable space.
1209 Frequency : when client is evicted
1211 Description: Grant might be wrongly erased when osc is evicted by OST
1212 Details : when the import is evicted by server, it will fork another
1213 thread ptlrpc_invalidate_import_thread to invalidate the
1214 import, where the grant will be set to 0. While the original
1215 thread will update the grant it got when connecting. So if
1216 the former happened latter, the grant will be wrongly errased
1217 because of this race.
1222 Description: Checking Stale with correct fid
1223 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1224 to check whether it is stale, because sometimes, we want the
1225 enqueue happened anyway, and op_data.fid2 will not be initialized.
1227 Severity : enhancement
1229 Description: update patchless client
1230 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1233 Frequency : only with 2.4 kernel
1235 Description: random memory corruption
1236 Details : size of struct ll_inode_info is to big for union inode.u and this
1237 can be cause of random memory corruption.
1242 Description: Memory leak in recovery
1243 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1244 It should also check obd_fail before fsfilt_start, otherwise if
1245 fsfilt_start return -EROFS,(failover mds during mds recovery).
1246 then the req will return with repmsg->transno = 0 and rc = EROFS.
1247 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1248 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1249 be freed no matter whether fsfilt_commit success or not.
1252 Frequency : only with huge count clients
1254 Description: Prevents from taking the superblock lock in llap_from_page for
1256 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1257 for llap_from_page call in ll_removepage prevents from taking the
1258 superblock lock for a soon died page.
1263 Description: Not check open intent error before release open handle
1264 Details : in some rare cases, the open intent error is not checked before
1265 release open handle, which may cause
1266 ASSERTION(open_req->rq_transno != 0), because it tries to release
1267 the failed open handle.
1272 Description: Set cat log bitmap only after create log success.
1273 Details : in some rare cases, the cat log bitmap is set too early. and it
1274 should be set only after create log success.
1278 Description: Accessing a block bevice can re-enable I/O when Lustre is
1279 tearing down a device.
1280 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1285 Description: service threads may hog cpus when there are a lot of requests
1286 Details : Insert cond_resched to give other threads a chance to use some CPU
1291 Description: the cat log was not initialized in recovery
1292 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1293 unlink log on mds will not be initialized until mds post
1294 recovery. And also in mds post recovery, the unlink log will
1295 initialization will be done asynchronausly, so there will be race
1296 between add unlink log and unlink log initialization.
1300 Description: brw_stats were being printed incorrectly
1301 Details : brw_stats were being printed as log2 but all of them were not
1302 recorded as log2. Also remove some code duplication arising from
1303 filter_tally_{read,write}.
1307 Frequency : rare, only in recovery.
1308 Description: ASSERTION(req->rq_type != LI_POISON) failed
1309 Details : imp_lock should be held while iterating over imp_sending_list for
1310 prevent destroy request after get timeout in ptlrpc_queue_wait.
1314 Description: replay-single.sh test 52 fails
1315 Details : A lock's skiplist need to be cleanup when it being unlinked
1316 from its resource list.
1320 Description: Short directio read returns full requested size rather than
1322 Details : Direct I/O operations should return actual amount of bytes
1323 transferred rather than requested size.
1325 Severity : enhancement
1327 Description: metadata RPC reduction (e.g. for rm performance)
1328 Details : decrease the amount of synchronous RPC between clients and servers
1329 by canceling conflicing lock before the operation on the client
1330 and packing thier handles into the main operation RPC to server.
1332 Severity : enhancement
1334 Description: Async OSC create to avoid the blocking unnecessarily.
1335 Details : If a OST has no remain object, system will block on the creating
1336 when need to create a new object on this OST. Now, ways use
1337 pre-created objects when available, instead of blocking on an
1338 empty osc while others are not empty. If we must block, we block
1339 for the shortest possible period of time.
1341 Severity : enhancement
1343 Description: refine locking for avoid write wrong info into lov_objid file
1344 Details : fix possible races with add new target and write/update data in
1347 --------------------------------------------------------------------------------
1349 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1354 Frequency : on some architectures
1356 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1357 Details : Uninitialized flags sometimes cause configuration commands to
1361 Frequency : patchless clients only
1363 Description: missing __iget() symbol export
1364 Details : The __iget() symbol export is missing. To avoid the need for
1365 this on patchless clients the deathrow inode reaper is turned
1366 off, and we depend on the VM to clean up old inodes. This
1367 dependency was during via the fix for bug 12181.
1369 --------------------------------------------------------------------------------
1371 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1373 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1374 INTEROPERATE with older versions automatically. In many cases a
1375 special upgrade step is needed. Please read the
1376 user documentation before upgrading any part of a live system.
1377 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1378 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1379 * WARNING: Lustre configuration and startup changes are required with
1380 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1386 Severity : enhancement
1388 Description: MountConf
1389 Details : Lustre configuration is now managed via mkfs and mount
1390 commands instead of lmc and lconf. New obd types (MGS, MGC)
1391 are added for dynamic configuration management. See
1392 https://mail.clusterfs.com/wikis/lustre/MountConf for
1395 Severity : enhancement
1397 Description: dynamic OST addition
1398 Details : OSTs can now be added to a live filesystem
1400 Severity : enhancement
1402 Description: startup order invariance
1403 Details : MDTs and OSTs can be started in any order. Clients only
1404 require the MDT to complete startup.
1406 Severity : enhancement
1408 Description: parallel, asynchronous orphan cleanup
1409 Details : orphan cleanup is now performed in separate threads for each
1410 OST, allowing parallel non-blocking operation.
1412 Severity : enhancement
1414 Description: optimized stripe assignment
1415 Details : stripe assignments are now made based on ost space available,
1416 ost previous usage, and OSS previous usage, in order to try
1417 to optimize storage space and networking resources.
1419 Severity : enhancement
1421 Description: Permanently set tunables
1422 Details : All writable /proc/fs/lustre tunables can now be permanently
1423 set on a per-server basis, at mkfs time or on a live system.
1425 Severity : enhancement
1427 Description: Lustre message v2
1428 Details : Add lustre message format v2.
1430 Severity : enhancement
1432 Description: client OST exclusion list
1433 Details : Clients can be started with a list of OSTs that should be
1434 declared "inactive" for known non-responsive OSTs.
1438 Description: SPEC SFS validation failure on NFS v2 over lustre.
1439 Details : Changes the blocksize for regular files to be 2x RPC size,
1440 and not depend on stripe size.
1442 Severity : enhancement
1444 Description: Multiple MD RPCs in flight.
1445 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1446 To avoid overly-overloading MDS, introduce a limit on number of
1447 MDS RPCs in flight for a single client and add /proc controls
1448 to adjust this limit.
1450 Severity : enhancement
1452 Description: client read/write statistics
1453 Details : Add client read/write call usage stats for performance
1454 analysis of user processes.
1455 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1456 file access. extents_stats shows chunk size distribution.
1457 extents_stats_per_process show chunk size distribution per
1460 Severity : enhancement
1462 Description: mds statistics
1463 Details : Add detailed mds operations statistics in
1464 /proc/fs/lustre/mds/*/stats.
1468 Description: Failure of copying files with lustre special EAs.
1469 Details : Client side always return success for setxattr call for lustre
1470 special xattr (currently only "trusted.lov").
1475 Description: Refcount LNET uuids
1476 Details : The global LNET uuid list grew linearly with every startup;
1477 refcount repeated list entries instead of always adding to
1480 Severity : enhancement
1482 Description: Dynamic service threads
1483 Details : Within a small range, start extra service threads
1484 automatically when the request queue builds up.
1487 Frequency : mixed-endian client/server environments
1489 Description: mixed-endian crashes
1490 Details : The new msg_v2 system had some failures in mixed-endian
1493 Severity : enhancement
1495 Description: Easy OST removal
1496 Details : OSTs can be permanently deactivated with e.g. 'lctl
1497 conf_param lustre-OST0001.osc.active=0'
1499 Severity : enhancement
1501 Description: MGS proc entries
1502 Details : Added basic proc entries for the MGS showing what filesystems
1505 Severity : enhancement
1507 Description: provide MGS failover
1508 Details : Added config lock reacquisition after MGS server failover.
1510 Severity : enhancement
1512 Description: add Linux 2.4 support
1513 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1517 Description: a large application tries to do I/O to the same resource and dies
1518 in the middle of it.
1519 Details : Check the req->rq_arrival time after the call to
1520 ost_brw_lock_get(), but before we do anything about
1521 processing it & sending the BULK transfer request. This
1522 should help move old stale pending locks off the queue as
1523 quickly as obd_timeout.
1526 Frequency : when an incorrect nid is specified during startup
1528 Description: ptlrpc connect to non-existant node causes kernel crash
1529 Details : LNET can't be re-entered from an event callback, which
1530 happened when we expire a message after the export has been
1531 cleaned up. Instead, hand the zombie cleanup off to another
1534 Severity : enhancement
1536 Description: plain/inodebits lock performance improvement
1537 Details : Grouping plain/inodebits in granted list by their request modes
1538 and bits policy, thus improving the performance of search through
1542 Frequency : only if OST filesystem is corrupted
1544 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1545 Details : for a short time RPCs with bulk IO are in the replay list,
1546 but replay of bulk IOs is unimplemented. If the OST filesystem
1547 is corrupted due to disk cache incoherency and then replay is
1548 started it is possible to trip an assertion. Avoid putting
1549 committed RPCs into the replay list at all to avoid this issue.
1552 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1554 Description: System hang on startup
1555 Details : This bug allowed the liblustre (e.g. catamount) client to
1556 return to the app before handling all startup RPCs. This
1557 could leave the node unresponsive to lustre network traffic
1558 and manifested as a server ptllnd timeout.
1560 Severity : enhancement
1562 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1563 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1564 boolean to control whether to print peer debug info when a
1565 client's RPC times out.
1568 Frequency : only for kernels with patches from Lustre below 1.4.3
1570 Description: Remove old rdonly API
1571 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1574 Frequency : only for devices with external journals
1576 Description: Set external device read-only also
1577 Details : During a commanded failover stop, we set the disk device
1578 read-only while the server shuts down. We now also set any
1579 external journal device read-only at the same time.
1582 Frequency : when upgrading from 1.4 while trying to change parameters
1584 Description: The wrong (new) MDC name was used when setting parameters for
1585 upgraded MDT's. Also allows changing of OSC (and MDC)
1586 parameters if --writeconf is specified at tunefs upgrade time.
1589 Frequency : when setting specific ost indicies
1591 Description: QOS code breaks on skipped indicies
1592 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1593 created with --index need not be sequential.
1595 Severity : enhancement
1597 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1598 Details : The uninit_groups feature works in conjunction with the kernel
1599 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1600 up the pass1 processing of e2fsck. This is a read-only feature
1601 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1602 mount filesystems that have had this feature enabled.
1604 Severity : enhancement
1606 Description: Improve multi-block allocation algorithm to avoid fragmentation
1607 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1608 allocation locality and avoid filesystem fragmentation.
1610 ------------------------------------------------------------------------------
1612 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1614 * Support for kernels:
1615 2.4.21-47.0.1.EL (RHEL 3)
1616 2.6.5-7.283 (SLES 9)
1617 2.6.9-42.0.10.EL (RHEL 4)
1618 2.6.12.6 vanilla (kernel.org)
1619 2.6.16.27-0.9 (SLES 10)
1620 * Recommended e2fsprogs version: 1.39.cfs5
1622 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1626 Frequency : occasional, depends on client load and configuration
1627 Bugzilla : 12181, 12203
1628 Description: data loss for recently-modified files
1630 Details : In some cases it is possible that recently written or created
1631 files may not be written to disk in a timely manner (this should
1632 normally be within 30s unless client IO load is very high).
1633 The problem appears as zero-length files or files that are a
1634 multiple of 1MB in size after a client crash or client eviction
1635 that are missing data at the end of the file.
1637 This problem is more likely to be hit on clients where files are
1638 repeatedly created and unlinked in the same directory, clients
1639 have a large amount of RAM, have many CPUs, the filesystem has
1640 many OSTs, the clients are rebooted frequently, and/or the files
1641 are not accessed by other nodes after being written.
1643 The presence of the problem can be detected by looking at
1644 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1645 smaller than the second (nr_unused) then dirty files will not
1646 be flushed automatically to disk. "sync; sleep 10" should be
1647 run several times on the node before unmounting it to update
1648 Lustre (this is also safe to run on nodes without this problem).
1650 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1651 that can cause this same problem, so customers using that kernel
1652 also need to update the kernel in addition to Lustre. In order
1653 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1655 It is normal that files written just before a client crash (less
1656 than 30s) may not yet have been flushed to disk, even for local
1660 Frequency : frequent on thin XT3 nodes
1662 Description: UUID collision on thin XT3 Linux nodes
1663 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1664 randomly, since we relied on an insufficiently-seeded PRNG.
1669 Description: OSS hangs after "All ost request buffers busy"
1670 Details : A deadlock between quota and journal operations caused OSS
1671 hangs after printing "All ost request buffers busy."
1674 Frequency : always on liblustre builds
1676 Description: Cleanup compiler warnings on liblustre
1679 Frequency : always on liblustre builds on XT3
1681 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1685 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1686 > 32000 subdirectories
1687 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1688 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1689 subdirectory count crosses 32000. This will aid e2fsck to
1690 correctly handle more than 32000 subdirectories.
1693 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1695 Description: System hang on startup
1696 Details : This bug allowed the liblustre (e.g. catamount) client to
1697 return to the app before handling all startup RPCs. This
1698 could leave the node unresponsive to lustre network traffic
1699 and manifested as a server ptllnd timeout.
1701 Severity : enhancement
1703 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1704 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1705 boolean to control whether to print peer debug info when a
1706 client's RPC times out.
1711 Description: make O_SYNC working on 2.6 kernels
1712 Details : 2.6 kernels use different method for mark pages for write,
1713 so need add a code to lustre for O_SYNC work.
1718 Description: Failure to close file and release space on NFS
1719 Details : Put inode details into lock acquired in ll_intent_file_open.
1720 Use mdc_intent_lock in ll_intent_open to properly
1721 detect all kind of errors unhandled by mdc_enqueue.
1726 Description: proc file read during shutdown sometimes raced obd removal,
1728 Details : Add lock to prevent obd access after proc file removal.
1731 Frequency : Only for files larger than 4GB on 32-bit clients.
1733 Description: improperly doing page alignment of locks
1734 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1735 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1740 Description: RPCs being resent when they shouldn't be
1741 Details : Some RPCs that should not be resent are being resent. This
1742 can cause inconsistencies in the RPC state machine. Do not
1743 resend such requests.
1746 Frequency : rare, only with NFS export
1748 Description: Crash on NFS re-export node
1749 Details : under very unusual load conditions an assertion is hit in
1750 ll_intent_file_open()
1753 Frequency : only if OST filesystem is corrupted
1755 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1756 Details : for a short time RPCs with bulk IO are in the replay list,
1757 but replay of bulk IOs is unimplemented. If the OST filesystem
1758 is corrupted due to disk cache incoherency and then replay is
1759 started it is possible to trip an assertion. Avoid putting
1760 committed RPCs into the replay list at all to avoid this issue.
1765 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1766 Details : Large single O_DIRECT read and write calls can fail to allocate
1767 a sufficiently large buffer to process the request. In case of
1768 allocation failure the allocation is retried with a smaller
1769 buffer and broken into smaller requests.
1771 Severity : enhancement
1773 Description: Add -o localflock option to simulate old noflock behaviour.
1774 Details : This will achieve local-only flock/fcntl locks coherentness.
1779 Description: versioning check is incomplete
1780 Details : Checking the version difference of client vs. server, report
1781 error if the gap is too big.
1786 Description: add support PG_writeback bit
1787 Details : add support for PG_writeback bit for Lustre, for more carefull
1788 work with page cache in 2.6 kernel. This also fix some deadlocks
1789 and remove hack for work O_SYNC with 2.6 kernel.
1791 Severity : enhancement
1793 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1794 Details : The uninit_groups feature works in conjunction with the kernel
1795 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1796 up the pass1 processing of e2fsck. This is a read-only feature
1797 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1798 mount filesystems that have had this feature enabled.
1800 Severity : enhancement
1802 Description: Improve multi-block allocation algorithm to avoid fragmentation
1803 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1804 allocation locality and avoid filesystem fragmentation.
1806 ------------------------------------------------------------------------------
1808 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1810 * Support for kernels:
1811 2.6.9-42.0.3.EL (RHEL 4)
1812 2.6.5-7.276 (SLES 9)
1813 2.4.21-47.0.1.EL (RHEL 3)
1814 2.6.12.6 vanilla (kernel.org)
1815 2.6.16.21-0.8 (SLES10)
1816 * Recommended e2fsprogs version: 1.39.cfs2-0
1818 * The backwards-compatible /proc/sys/portals symlink has been removed
1819 in this release. Before upgrading, please ensure that you change
1820 any configuration scripts or /etc/sysctl.conf files that access
1821 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1822 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1823 in advance of the upgrade on any system running Lustre 1.4.6 or
1824 newer, since /proc/sys/lnet was added in that version.
1825 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1829 Frequency : only when quota is used
1831 Description: avoid scanning export list for quota master
1832 Details : Change the algorithms to avoid scanning export list in order
1833 to improve the efficiency.
1836 Frequency : MDS failover only, very rarely
1838 Description: "went back in time" messages on mds failover
1839 Details : The greatest transno may be lost when the current operation
1840 finishes with an error (transno==0) and the client's last_rcvd
1841 record is over-written. Save the greatest transno in the
1842 mds_last_transno for this case.
1845 Frequency : always for specific kernels and striping counts
1847 Description: client may get "Matching packet too big" without ACL support
1848 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1849 when trying to access files in certain configurations. The
1850 clients should in fact be denied when mounting because they do
1851 not understand ACLs.
1854 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1856 Description: many clients connecting with IO in progress causes connect timeouts
1857 Details : Avoid synchronous journal commits to avoid delays caused by many
1858 clients connecting/disconnecting when bulk IO is in progress.
1859 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1860 OST_IO_PORTAL to avoid delays behind potentially many pending
1864 Frequency : occasionally with multiple writers to a single file
1866 Description: shared writes to file may result in wrong size reported by stat()
1867 Details : Allow growing of kms when extent lock is cancelled
1870 Frequency : always with random mmap IO to multi-striped file
1872 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1873 Details : Only if the hole is at the end of OST object so that kms is too
1874 small. Fix is to increase kms accordingly in ll_nopage.
1877 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1879 Description: writes to a missing object would leak memory on the OST
1880 Details : If there is an inconsistency between the MDS and OST filesystems,
1881 such that the MDS references an object that doesn't exist, writes
1882 to that object will leak memory due to incorrect cleanup in the
1883 error handling path, eventually running out of memory on the OST.
1888 Description: Creating too long symlink causes lustre errors
1889 Details : Check symlink and name lengths before sending requests to MDS.
1892 Frequency : only if flock is enabled (not on by default)
1894 Description: posix locks not released on fd closure on 2.6.9+
1895 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1896 kernels, this caused such locks not to be released on fd close and
1897 then assertions on fs unmount about still used locks.
1900 Frequency : MDS failover only, very rarely
1902 Description: clients may get ASSERTION(granted_lock != NULL)
1903 Details : When request was taking a long time, and a client was resending
1904 a getattr by name lock request. The were multiple lock requests
1905 with the same client lock handle and
1906 mds_getattr_name->fixup_handle_for_resent_request found one of the
1907 lock handles but later failed with ASSERTION(granted_lock != NULL).
1912 Description: handle->h_buffer_credits > 0, assertion failure
1913 Details : h_buffer_credits is zero after truncate, causing assertion
1914 failure. This patch extends the transaction or creates a new
1918 Frequency : NFS re-export or patchless client
1919 Bugzilla : 11179, 10796
1920 Description: Crash on NFS re-export node (__d_move)
1921 Details : We do not want to hash the dentry if we don't have a lock.
1922 But if this dentry is later used in d_move, we'd hit uninitialised
1923 list head d_hash, so we just do this to init d_hash field but
1924 leave dentry unhashed.
1927 Frequency : NFS re-export or patchless client
1929 Description: NFS exports has problem with symbolic link
1930 Details : lustre client didn't properly install dentry when re-exported
1931 to NFS or running patchless client.
1934 Frequency : NFS re-export or patchless client
1936 Description: Various nfs/patchless fixes.
1937 Details : fixes reuse disconected alias for lookup process - this fixes
1938 warning "find_exported_dentry: npd != pd",
1939 fix permission error with open files at nfs.
1940 fix apply umask when do revalidate.
1943 Frequency : occasional
1945 Description: Crash on NFS re-export node
1946 Details : calling clear_page() on the wrong pointer triggered oops in
1947 generic_mapping_read().
1950 Frequency : rarely, using O_DIRECT IO
1952 Description: unaligned directio crashes client with LASSERT
1953 Details : check for unaligned buffers before trying any requests.
1956 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1958 Description: stale data returned from RAID cache
1959 Details : If only a small amount of IO is done to the RAID device before
1960 reading it again it is possible to get stale data from the RAID
1961 cache instead of reading it from disk.
1964 Frequency : always for sles10 kernel
1966 Description: sles10 support
1967 Details : ll_follow_link: compile fixes and using of nd_set_link
1968 under newer kernels.
1971 Frequency : depends on arch, kernel and compiler version, always on sles10
1974 Description: recursive or deep enough symlinks cause stack overflow
1975 Details : getting rid of large stack-allocated variable in
1979 Frequency : depends on hardware
1981 Description: lustre write performance loss in the SLES10 kernel
1982 Details : the performance loss is caused by using of write barriers in the
1983 ext3 code. The SLES10 kernel turns barrier support on by
1984 default. The fix is to undo that change for ldiskfs.
1986 ------------------------------------------------------------------------------
1988 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1990 * Support for kernels:
1991 2.6.9-42.0.3EL (RHEL 4)
1992 2.6.5-7.276 (SLES 9)
1993 2.4.21-47.0.1.EL (RHEL 3)
1994 2.6.12.6 vanilla (kernel.org)
1998 Frequency : quota enabled and large files being deleted
2000 Description: releasing more than 4GB of quota at once hangs OST
2001 Details : If a user deletes more than 4GB of files on a single OST it
2002 will cause the OST to spin in an infinite loop. Release
2003 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2008 Description: statfs data retrieved from /proc may be stale or zero
2009 Details : When reading per-device statfs data from /proc, in the
2010 {kbytes,files}_{total,free,avail} files, it may appear
2011 as zero or be out of date.
2014 Frequency : systems with MD RAID1 external journal devices
2016 Description: lconf's call to blkid is confused by RAID1 journal devices
2017 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2018 of returning all block devices that match the journal UUID.
2021 Frequency : always, for aggregate stripe size over 4GB
2023 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2024 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2025 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2028 Frequency : always if "lfs find" used on a local file/directory
2030 Description: "lfs find" segfaults if used on a local file/directory
2031 Details : The case where a directory component was not specified wasn't
2032 handled correctly. Handle this properly.
2035 Frequency : always on ppc64
2037 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2038 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2039 perform endianness conversion.
2042 Frequency : rarely (truncate to non-zero file size after write under load)
2043 Bugzilla : 10730, 10687
2044 Description: Files padded with zeros to next 4K multiple
2045 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2046 it is possible that files that are truncated to a non-zero size
2047 immediately after being written are filled with zero bytes beyond
2048 the truncated size. No file data is lost.
2050 Severity : enhancement
2052 Description: Allow recovery/failover for liblustre clients.
2053 Details : liblustre clients were unaware of failover configurations until
2056 Severity : enhancement
2058 Description: user file locks should fail when not mounting with flock option
2059 Details : Set up an error-returning stub in ll_file_operations.lock field
2060 to prevent incorrect behaviour when client is mounted without
2061 flock option. Also, set up properly f_op->flock field for
2065 Frequency : always on ia64
2067 Description: "lfs df" loops on printing out MDS statfs information
2068 Details : The obd_ioctl_data was not initialized and in some systems
2069 this caused a failure during the ioctl that did not return
2070 an error. Initialize the struct and return an error on failure.
2073 Frequency : SLES 9 only
2075 Description: Error of copying files with lustre special EAs as root
2076 Details : Client side always return success for setxattr call for lustre
2077 special xattr (currently only "trusted.lov").
2080 Frequency : rarely on clusters with both ia64+i386 clients
2082 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2083 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2084 clients are doing concurrent writes to the same file it is
2085 possible that the ia64 clients may LASSERT because the OST
2086 extent locks are not PAGE_SIZE aligned. Ensure that grown
2087 locks are always aligned on the request boundary.
2090 Frequency : specific use, occasional
2092 Description: Overwriting in use executable truncates on-disk binary image
2093 Details : If one node attempts to overwrite an executable in use by
2094 another node, we now correctly return ETXTBSY instead of
2095 truncating the file.
2097 Severity : enhancement
2099 Description: Async OSC create to avoid the blocking unnecessarily.
2100 Details : If a OST has no remain object, system will block on the creating
2101 when need to create a new object on this OST. Now, ways use
2102 pre-created objects when available, instead of blocking on an
2103 empty osc while others are not empty. If we must block, we block
2104 for the shortest possible period of time.
2109 Description: chmod on Lustre root is propagated to other clients
2110 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2111 invalid by the follow_mount time.
2114 Frequency : liblustre clients only
2116 Description: Race in 'instant cancel' lock handling could lead to such locks
2117 never to be granted in case of SMP MDS
2118 Details : Do not destroy not yet granted but cbpending locks in
2122 Frequency : replay/resend of open
2124 Description: non null lock assetion failure in mds_intent_policy
2125 Details : Trying to replay/resend lockless open requests resulted in
2126 mds_open() returning 0 with no lock. Now it sets a flag if
2127 a lock is going to be returned.
2129 Severity : enhancement
2131 Description: Checksum enhancements
2132 Details : New checksum enhancements allow for resending RPCs that failed
2135 Severity : enhancement
2137 Description: Tunables on number of dirty pages in cacche
2138 Details : Allow to set limit on number of dirty pages cached.
2143 Description: client crash on unmount - lock still has references
2144 Details : In some error handling cases it was possible to leak a lock
2145 reference on a client while accessing a file. This was not
2146 harmful to the client during operation, but would cause the
2147 client to crash when the filesystem is unmounted.
2150 Frequency : specific case, rare
2152 Description: ETXTBSY on mds though file not in use by client
2153 Details : ETXTBSY is no longer incorrectly returned when attempting to
2154 chmod or chown a directory that the user previously tried to
2155 execute or a currently-executing binary.
2158 Frequency : extremely rare except on liblustre-based clients
2160 Description: Lustre space not freed when files are deleted
2161 Details : Clean up open-unlinked files after client eviction. Previously
2162 the unlink was skipped and the files remained as orphans.
2167 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2168 Details : In some cases it was possible to send a blocking callback to
2169 a client doing a glimpse, even though that client didn't get
2170 a lock granted. When the glimpse lock is cancelled on the OST
2171 the freed lock is left on the waiting list and corrupted the list.
2174 Frequency : all core dumps
2176 Description: Broke core dumps to lustre
2177 Details : Negative dentry may be unhashed if parent does not have UPDATE
2178 lock, but some callers, e.g. do_coredump, expect dentry to be
2179 hashed after successful create, hash it in ll_create_it.
2181 ------------------------------------------------------------------------------
2183 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2185 * Support for kernels:
2186 2.6.9-42.0.2.EL (RHEL 4)
2187 2.6.5-7.276 (SLES 9)
2188 2.4.21-40.EL (RHEL 3)
2189 2.6.12.6 vanilla (kernel.org)
2193 Frequency : always on RHEL 3
2195 Description: Number of open files grows over time
2196 Details : The number of open files grows over time, whether or not
2197 Lustre is started. This was due to a filp leak introduced
2198 by one of our kernel patches.
2200 ------------------------------------------------------------------------------
2202 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2204 * Support for kernels:
2205 2.6.9-42.EL (RHEL 4)
2206 2.6.5-7.276 (SLES 9)
2207 2.4.21-40.EL (RHEL 3)
2208 2.6.12.6 vanilla (kernel.org)
2213 Bugzilla : 5719, 9635, 9792, 9684
2214 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2215 Details : If a server is under heavy load and cannot reply to new
2216 connection requests before the client resends the (re)connect,
2217 the connection handling code can behave badly if two service
2218 threads are concurrently handing separate (re)connections from
2219 the same client. Add better locking to the connection handling
2220 code, and ensure that only a single connection will be processed
2221 for a given client UUID, even if the lock is dropped.
2223 Severity : enhancement
2225 Description: add TCP zero-copy support to kernel
2226 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2227 sends if the hardware supports scatter-gather and checksumming.
2228 This allows socklnd to do client-write and server-read more
2229 efficiently and reduce CPU utilization from skbuf copying.
2232 Frequency : only if NFS exporting from client
2234 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2235 Details : If setting attributes on a file created under NFS that had
2236 never been opened it would be possible to oops the client
2237 if the file had no objects.
2240 Frequency : always for liblustre
2242 Description: liblustre client does MDS+OSTs setattr RPC for each write
2243 Details : When doing a write from a liblustre client, the client
2244 incorrectly issued an RPC to the MDS and each OST the file was
2245 striped over in order to update the timestamps. When writing
2246 with small chunks and many clients this could overwhelm the MDS
2247 with RPCs. In all cases it would slow down the write because
2248 these RPCs are unnecessary.
2250 Severity : enhancement
2252 Description: allow number of MDS service threads to be changed at module load
2253 Details : It is now possible to change the number of MDS service threads
2254 running. Adding "options mds mds_num_threads={N}" to the MDS's
2255 /etc/modprobe.conf will set the number of threads for the next
2256 time Lustre is restarted (assuming the "mds" module is also
2257 reloaded at that time). The default number of threads will
2258 stay the same, 32 for most systems.
2263 Description: OST crash if filesystem is unformatted or corrupt
2264 Details : If an OST is started on a device that has never been formatted
2265 or if the filesystem is corrupt and cannot even mount then the
2266 error handling cleanup routines would dereference a NULL pointer.
2271 Description: NULL pointer deref in llap_from_page.
2272 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2273 incorrect) mapping if the page was truncated/reclaimed while it was
2274 searched for. Check for this condition and skip such pages when
2275 doing readahead. Introduce extra check to llap_from_page() to
2276 verify page->mapping->host is non-NULL (so page is not anonymous).
2279 Frequency : Sometimes when using sys_sendfile
2281 Description: "page not covered by a lock" warnings from ll_readpage
2282 Details : sendfile called ll_readpage without right page locks present.
2283 Now we introduced ll_file_sendfile that does necessary locking
2284 around call to generic_file_sendfile() much like we do in
2288 Frequency : with certain MDS communication failures at client mount time
2290 Description: NULL pointer deref after failed client mount
2291 Details : a client connection request may delayed by the network layer
2292 and not be sent until after the PTLRPC layer has timed out the
2293 request. If the client fails the mount immediately it will try
2294 to clean up before the network times out the request. Add a
2295 reference from the request import to the obd device and delay
2296 the cleanup until the network drops the request.
2299 Frequency : occasionally during client (re)connect
2301 Description: assertion failure during client (re)connect
2302 Details : processing a client connection request may be delayed by the
2303 client or server longer than the client connect timeout. This
2304 causes the client to resend the connection request. If the
2305 original connection request is replied in this interval, the
2306 client may trip an assertion failure in ptlrpc_connect_interpret()
2307 which thought it would be the only running connect process.
2310 Frequency : only with obd_echo servers and clients that are rebooted
2312 Description: kernel BUG accessing uninitialized data structure
2313 Details : When running an obd_echo server it did not start the ping_evictor
2314 thread, and when a client was evicted an uninitialized data
2315 structure was accessed. Start the ping_evictor in the RPC
2316 service startup instead of the OBD startup.
2318 Severity : enhancement
2319 Bugzilla : 10193 (patchless)
2320 Description: Remove dependency on various unexported kernel interfaces.
2321 Details : No longer need reparent_to_init, exit_mm, exit_files,
2322 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2325 Frequency : rare (only users of deprecated and unsupported LDAP config)
2327 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2328 Details : LDAP apparently contains 'lustreName' attributes instead of
2329 'name'. A simple remapping of the name is sufficient.
2332 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2334 Description: waiting_locks_callback trips kernel BUG if client is evicted
2335 Details : Running with the dump_on_timeout debug flag turned on makes
2336 it possible that the waiting_locks_callback() can try to dump
2337 the Lustre kernel debug logs from an interrupt handler. Defer
2338 this log dumping to the expired_lock_main() thread.
2340 Severity : enhancement
2342 Description: Support NFS exporting on 2.6 kernels.
2343 Details : Implement non-rawops metadata methods for NFS server to use without
2344 changing NFS server code.
2347 Frequency : very rare (synthetic metadata workload only)
2349 Description: two racing renames might cause an MDS thread to deadlock
2350 Details : Running the "racer" program may cause one MDS thread to rename
2351 a file from being the source of a rename to being the target of
2352 a rename at exactly the same time that another thread is doing
2353 so, and the second thread has already enqueued these locks after
2354 doing a lookup of the target and is trying to relock them in
2355 order. Ensure that we don't try to re-lock the same resource.
2358 Frequency : only very large systems with liblustre clients
2360 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2361 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2362 all OSTs in parallel. This allows the eviction of stale liblustre
2363 clients to proceed much faster than if they were done in series,
2364 and also offers similar improvements for other set_info RPCs.
2369 Description: excessive CPU usage during initial read phase on client
2370 Details : During the initial read phase on a client, it would agressively
2371 retry readahead on the file, consuming too much CPU and impacting
2372 performance (since 1.4.5.8). Improve the readahead algorithm
2373 to avoid this, and also improve some other common cases (read
2374 of small files in particular, where "small" is files smaller than
2375 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2380 Description: MDS crash when receiving packet with unknown intent.
2381 Details : Do not LBUG in unknown intent case, just return -EFAULT
2383 Severity : enhancement
2384 Bugzilla : 9293, 9385
2385 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2386 Details : Do not serialize getattr (non-intent version) and statfs.
2389 Frequency : occasional, when OST network is overloaded/intermittent
2391 Description: client evicted by OST after bulk IO timeout
2392 Details : If a client sends a bulk IO request (read or write) the OST
2393 may evict the client if it is unresposive to its data GET/PUT
2394 request. This is incorrect if the network is overloaded (takes
2395 too long to transfer the RPC data) or dropped the OST GET/PUT
2396 request. There is no need to evict the client at all, since
2397 the pinger and/or lock callbacks will handle this, and the
2398 client can restart the bulk request.
2401 Frequency : Always when mmapping file with no objects
2403 Description: client crashes when mmapping file with no objects
2404 Details : Check that we actually have objects in a file before doing any
2405 operations on objects in ll_vm_open, ll_vm_close and
2411 Description: Request leak when working with deleted CWD
2412 Details : Introduce advanced request refcount tracking for requests
2413 referenced from lustre intent.
2415 Severity : Enhancement
2417 Description: Cache open file handles on client.
2418 Details : MDS now will return special lock along with openhandle, if
2419 requested and client is allowed to hold openhandle, even if unused,
2420 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2421 closing files for every read/write openration.
2423 Severity : Enhancement
2425 Description: Cache open negative dentries on client when possible.
2426 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2427 negative dentries on lock revocation.
2432 Description: Remounting a client read-only wasn't possible with a zconf mount
2433 Details : It wasn't possible to remount a client read-only with llmount.
2435 Severity : enhancement
2436 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2437 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2438 MPICH is included to add Lustre-specific ADIO interfaces.
2439 This is based closely on the UFS ADIO layer and only differs
2440 in file creation, in order to allow the OST striping to be set.
2441 This is user-contributed code and not supported by CFS.
2446 Description: extended inode attributes (immutable, append-only) work improperly
2447 when 2.4 and 2.6 kernels are used on client/server or vice versa
2448 Details : Introduce kernel-independent values for these flags.
2450 Severity : enhancement
2453 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2454 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2455 granularity. Now, Fractional MB granularity is be supported,
2456 this is very useful for low memory system.
2458 Severity : enhancement
2460 Description: Getattr by fid
2461 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2462 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2463 have locks and avoids getattr rpc requests for it on every stat.
2466 Frequency : Always, for filesystems larger than 2TB
2468 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2469 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2470 filesystems larger than 2TB to be mounted with mballoc enabled.
2473 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2475 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2476 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2477 filesystems larger than 2TB is mounted then the kernel will
2478 silently corrupt the start of the filesystem. CONFIG_LBD is
2479 enabled for all CFS-supported kernels, but the possibility of
2480 this happening with a modified kernel config exists.
2482 Severity : enhancement
2484 Description: add client O_DIRECT support for 2.6 kernels
2485 Details : It is now possible to do O_DIRECT reads and writes to files
2486 in the Lustre client mountpoint on 2.6 kernel clients.
2488 Severity : enhancement
2490 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2491 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2492 parallel, not waiting for response from every OST before sending
2493 a rpc to the next OST.
2498 Description: setattr vs write race when updating file timestamps
2499 Details : Client processes that update a file timestamp into the past
2500 right after writing to the file (e.g. tar) it is possible that
2501 the updated file modification time can be reset to the current
2502 time due to a race between processing the setattr and write RPC.
2504 Severity : enhancement
2506 Description: Bring 'lfs find' closer in line with regular Linux find.
2507 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2508 -print0 options and obtains all the needed info through the lustre
2511 Severity : enhancement
2513 Description: support up to 1024 configured devices on one node
2514 Details : change obd_dev array from statically allocated to dynamically
2515 allocated structs as they are first used to reduce memory usage
2520 Description: Flush dirty partially truncated pages during truncate
2521 Details : Immediatelly flush partially truncated pages in filter_setattr,
2522 this way we completely avoid having any pages in page cache on OST
2523 and can retire ugly workarounds during writes to flush such pages.
2528 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2529 Details : protect inode from truncation within vfs_unlink() context
2530 just take a reference before calling vfs_unlink() and release it
2531 when parent's i_sem is free.
2536 Description: last_id value checked outside lock on OST caused LASSERT failure
2537 Details : If there were multiple MDS->OST object precreate requests in
2538 flight, it was possible that the OST's last object id was checked
2539 outside a lock and incorrectly tripped an assertion. Move checks
2540 inside locks, and discard old precreate requests.
2543 Frequency : always, if extents are used on OSTs
2545 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2546 Details : OSTs using the extents format would not zero the high 16 bits of
2547 the index physical block number. This is not a problem for any
2548 OST filesystems smaller than 16TB, and no kernels support ext3
2549 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2550 new/modified files) and can be fixed for existing filesystems
2551 with e2fsprogs-1.39-cfs1.
2556 Description: import connection selection may be incorrect if timer wraps
2557 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2558 import connections to be ignored if the 32-bit jiffies counter
2559 wraps. Use a 64-bit jiffies counter.
2562 Frequency : very large clusters immediately after boot
2564 Description: LNET request buffers exhausted under heavy short-term load
2565 Details : If a large number of client requests are generated on a service
2566 that has previously never seen so many requests it is possible
2567 that the request buffer growth cannot keep up with the spike in
2568 demand. Instead of dropping incoming requests, they are held in
2569 the LND until the RPC service can accept more requests.
2572 Frequency : Sometimes during replay
2574 Description: Assertion failure in ll_local_open after replay.
2575 Details : If replay happened on an open request reply before we were able
2576 to set replay handler, reply will become not swabbed tripping the
2577 assertion in ll_local_open. Now we set the handler right after
2578 recognising of open request
2581 Frequency : very rare
2583 Description: kernel reports "badness in vsnprintf"
2584 Details : Reading from the "recovery_status" /proc file in small chunks
2585 may cause a negative length in lprocfs_obd_rd_recovery_status()
2586 call to vsnprintf() (which is otherwise harmless). Exit early
2587 if there is no more space in the output buffer.
2589 Severity : enhancement
2591 Description: clear OBD RPC statistics by writing to them
2592 Details : It is now possible to clear the OBD RPC statistics by writing
2593 to the "stats" file.
2598 Description: Client mtime is not the same on different clients after utimes
2599 Details : In some cases, the client was using the utimes() syscall on
2600 a file cached on another node. The clients now validate the
2601 ctime from the MDS + OSTs to determine which one is right.
2606 Description: Inability to activate failout mode
2607 Details : lconf script incorrectly assumed that in pythong string's numeric
2608 value is used in comparisons.
2611 Frequency : always with multiple stripes per file
2613 Description: Inefficient object allocation for mutli-stripe files
2614 Details : When selecting which OSTs to stripe files over, for files with
2615 a stripe count that divides evenly into the number of OSTs,
2616 the MDS is always picking the same starting OST for each file.
2617 Return the OST selection heuristic to the original design.
2622 Description: mount failures may take full timeout to return an error
2623 Details : Under some heavy load conditions it is possible that a
2624 failed mount can wait for the full obd_timeout interval,
2625 possibly several minutes, before reporting an error.
2626 Instead return an error as soon as the status is known.
2628 Frequency : quota enabled and large files being deleted
2630 Description: releasing more than 4GB of quota at once hangs OST
2631 Details : If a user deletes more than 4GB of files on a single OST it
2632 will cause the OST to spin in an infinite loop. Release
2633 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2638 Description: statfs data retrieved from /proc may be stale or zero
2639 Details : When reading per-device statfs data from /proc, in the
2640 {kbytes,files}_{total,free,avail} files, it may appear
2641 as zero or be out of date.
2644 Frequency : systems with MD RAID1 external journal devices
2646 Description: lconf's call to blkid is confused by RAID1 journal devices
2647 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2648 of returning all block devices that match the journal UUID.
2651 Frequency : always, for aggregate stripe size over 4GB
2653 Description: assertion fails when trying to use 4GB stripe size
2654 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2655 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2658 Frequency : always on ppc64
2660 Description: the first write on an ext3 filesystem with mballoc got stuck
2661 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2662 perform endianness conversion.
2664 ------------------------------------------------------------------------------
2666 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2668 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2669 INTEROPERATE with older versions automatically. Please read the
2670 user documentation before upgrading any part of a live system.
2671 * WARNING: Lustre networking configuration changes are required with
2672 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2675 * Support for kernels:
2676 2.6.9-22.0.2.EL (RHEL 4)
2677 2.6.5-7.244 (SLES 9)
2678 2.6.12.6 vanilla (kernel.org)
2681 Severity : enhancement
2682 Bugzilla : 7981/8208
2683 Description: Introduced Lustre Networking (LNET)
2684 Details : LNET is new networking infrastructure for Lustre, it includes
2685 a reorganized network configuration mode (see the user
2686 documentation for full details) as well as support for routing
2687 between different network fabrics. Lustre Networking Devices
2688 (LNDs) for the supported network fabrics have also been
2689 created for this new infrastructure.
2691 Severity : enhancement
2692 Description: Introduced Access control lists
2693 Details : clients can set ACLs on files and directories in order to have
2694 more fine-grained permissions than the standard Unix UGO+RWX.
2695 The MDS must be started with the "-o acl" mount option.
2697 Severity : enhancement
2698 Description: Introduced filesystem quotas
2699 Details : Administrators may now establish per-user quotas on the
2702 Severity : enhancement
2704 Description: Configuration change for the XT3
2705 The PTLLND is now used to run Lustre over Portals on the XT3
2706 The configure option(s) --with-cray-portals are no longer used.
2707 Rather --with-portals=<path-to-portals-includes> is used to
2708 enable building on the XT3. In addition to enable XT3 specific
2709 features the option --enable-cray-xt3 must be used.
2714 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2715 Details : A race between allocating a new llcd and re-getting the llcd_lock
2716 allowed another thread to grab newly-allocated llcd.
2718 Severity : enhancement
2720 Description: 2.6 OST async journal commit and locking fix to improve performance
2721 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2722 2.6 kernels are now async as they already were in 2.4 kernels so
2723 that they can commit concurrently with the network bulk transfer.
2724 For block-allocated files the filter allocation semaphore is held
2725 to avoid filesystem fragmentation during allocation. BKL lock
2726 removed for 2.6 xattr operations where it is no longer needed.
2731 Description: lconf incorrectly determined whether two IP networks could talk
2732 Details : In some more complicated routing and multiple-network
2733 configurations, lconf will avoid trying to make a network
2734 connection to a disjoint part of the IP space. It was doing the
2735 math incorrectly for one set of cases.
2740 Description: Fix for potential infinite loop processing records in an llog.
2741 Details : If an llog record is corrupted/zeroed, it is possible to loop
2742 forever in llog_process(). Validate the llog record length
2743 and skip the remainder of the block on error.
2746 Frequency : occasional (liblustre only)
2748 Description: liblustre could not open files whose last component is a symlink
2749 Details : sysio_path_walk() would incorrectly pass the open intent to
2750 intermediate path components.
2753 Frequency : rare (liblustre only with non-standard tuning)
2754 Bugzilla : 7201 (7350)
2755 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2756 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2757 after it has been released, only do this for referenced locks
2759 Severity : enhancement
2761 Description: specifying an (invalid) directory default stripe_size of -1
2762 would reset the directory default striping
2763 Details : stripe_size -1 was used internally to signal directory stripe
2764 removal, now use "all default" to signal dir stripe removal
2765 as a directory striping of "all default" is not useful
2768 Frequency : common for large clusters running liblustre clients
2770 Description: doing an ls when liblustre clients are running is slow
2771 Details : sending a glimpse AST to a liblustre client waits for every AST
2772 to time out, as liblustre clients will not respond. Since they
2773 cannot cache data we refresh the OST lock LVB from disk instead.
2775 Severity : enhancement
2777 Description: doing an ls at the same time as file IO can be slow
2778 Details : enqueue and other "small" requests can be blocked behind many
2779 large IO requests. Create a new OST IO portal for non-IO
2780 requests so they can be processed faster.
2783 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2785 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2786 created file with mode 000
2787 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2788 allow an ftruncate() as a non-root user to a file with mode 000.
2789 Lustre now respects this flag to disable mode checks when
2790 truncating a file owned by the user
2793 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2796 Description: Revoking locks from clients that went dead or catatonic might take
2798 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2799 cancellation of such locks instant on servers without waiting for
2800 any reply from clients. Clients drops these locks when cancel
2801 notification from server is received without replying.
2804 Frequency : liblustre-only, when liblustre client dies or becomes busy
2806 Description: Doing ls on Linux clients can take a long time with active
2808 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2809 granting such locks to it in the first place if possible. Locks
2810 are taken by proxy on the OST during the read or write and
2811 dropped immediately afterward. Add connect flags handling, do
2812 not grant locks to liblustre clients for glimpse ASTs.
2814 Severity : enhancement
2816 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2817 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2818 sequential reads and behaves badly with applications doing random
2819 reads. Improve it by reading ahead at least read region, and
2820 avoiding excessive large RPC for small reads.
2822 Severity : enhancement
2824 Description: Creating more than 1000 files for a single job may cause a load
2825 imbalance on the OSTs if there are also a large number of OSTs.
2826 Details : qos_prep_create() uses an OST index reseed value that is an
2827 even multiple of the number of available OSTs so that if the
2828 reseed happens in the middle of the object allocation it will
2829 still utilize the OSTs as uniformly as possible.
2834 Description: OST or MDS may oops in ping_evictor_main()
2835 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2836 but doesn't restart at beginning of obd_exports_timed list
2839 Severity : enhancement
2841 Description: improve by-nid export eviction on the MDS and OST
2842 Details : allow multiple exports with the same NID to be evicted at one
2843 time without re-searching the exports list.
2846 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2848 Description: MDS may oops in groups_free()
2849 Details : in rare race conditions a newly allocated group_info struct is
2850 freed again, and this can be NULL. The 2.4 compatibility code
2851 for groups_free() checked for a NULL pointer, but 2.6 did not.
2854 Frequency : common for liblustre clients doing little filesystem IO
2855 Bugzilla : 9352, 7313
2856 Description: server may evict liblustre clients accessing contended locks
2857 Details : if a client is granted a lock or receives a completion AST
2858 with a blocking AST already set it would not reply to the AST
2859 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2862 Frequency : lfs setstripe, only systems with more than 160 OSTs
2864 Description: unable to set striping with a starting offset beyond OST 160
2865 Details : llapi_create_file() incorrectly limited the starting stripe
2866 index to the maximum single-file stripe count.
2869 Frequency : LDAP users only
2871 Description: lconf did not handle in-kernel recovery with LDAP properly
2872 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2874 Severity : enhancement
2876 Description: bind OST threads to NUMA nodes to improve performance
2877 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2878 node and do their allocations there to localize memory access
2880 Severity : enhancement
2882 Description: llmount can determine client NID directly from Myrinet (GM)
2883 Details : the client NID code from gmnalnid was moved directly into
2884 llmount, removing the need to use this or specifying the
2885 client NID explicitly when mounting GM clients with zeroconf
2888 Frequency : if client is started with down MDS
2890 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2891 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2892 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2893 ptlrpc_import_delayed_req() because we didn't check if the
2894 request was interrupted, and we also didn't break out of the
2895 event loop if there was no timeout
2900 Description: data loss during non-page-aligned writes to a single file from
2901 both multiple nodes and multiple threads on one node at same time
2902 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2903 inconsistency led to false short-reads, that were cached and later
2904 used by ->prepare_write() to fill in partially written page,
2905 leading to data loss.
2908 Frequency : always, if lconf --abort_recovery used
2910 Description: lconf --abort_recovery fails with 'Operation not supported'
2911 Details : lconf was attempting to abort recovery on the MDT device and not
2914 Severity : enhancement
2916 Description: remove cleanup logs
2917 Details : replace lconf-generated cleanup logs with lustre internal
2918 cleanup routines. Eliminates the need for client-cleanup and
2921 Severity : enhancement
2923 Description: add support for EAs (user and system) on lustre filesystems
2924 Details : it is now possible to store extended attributes in the Lustre
2925 client filesystem, and with the user_xattr mount option it
2926 is possible to allow users to store EAs on their files also
2928 Severity : enhancement
2930 Description: Add possibility (config option) to show minimal available OST free
2932 Details : When compiled with --enable-mindf configure option, statfs(2)
2933 (and so, df) will return least minimal free space available from
2934 all OSTs as amount of free space on FS, instead of summary of
2935 free spaces of all OSTs.
2937 Severity : enhancement
2939 Description: do not expand extent locks acquired on OST-side
2940 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2941 by server: they are not cached anyway.
2944 Frequency : when mmap is used/binaries executed from Lustre
2946 Description: Unmmap pages before throwing them away from read cache.
2947 Details : llap_shrink cache now attempts to unmap pages before discarding
2948 them (if unmapping failed - do not discard). SLES9 kernel has
2949 extra checks that trigger if this unmapping is not done first.
2954 Description: lconf didn't resolve symlinks before checking to see whether a
2955 given mountpoint was already in use
2958 Frequency : when migrating failover services
2959 Bugzilla : 6395, 9514
2960 Description: When migrating a subset of services from a node (e.g. failback
2961 from a failover service node) the remaining services would
2962 time out and evict clients.
2963 Details : lconf --force (implied by --failover) sets the global obd_timeout
2964 to 5 seconds in order to quickly disconnect, but this caused
2965 other RPCs to time out too quickly. Do not change the global
2966 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2968 Severity : enhancement
2969 Frequency : if MDS is started with down OST
2970 Bugzilla : 9439,5706
2971 Description: Allow startup/shutdown of an MDS without depending on the
2972 availability of the OSTs.
2973 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2974 Add appropriate locking and lov-osc refcounts for safe
2975 cleaning. Add osc abort_inflight calls in case the
2976 synchronize never started.
2979 Frequency : occasional (Cray XT3 only)
2981 Description: root not authorized to access files in CRAY_PORTALS environment
2982 Details : The client process capabilities were not honoured on the MDS in
2983 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2984 been accessed by an authorized user then root was able to access
2985 the file on the local client also. The root user capabilities
2986 are now allowed on the MDS, as this environment has secure UID.
2989 Frequency : occasional
2991 Description: ldiskfs "too long searching" message happens too often
2992 Details : A debugging message (otherwise harmless) prints too often on
2993 the OST console. This has been reduced to only happen when
2994 there are fragmentation problems on the filesystem.
2999 Description: Division by zero in statfs when all OSCs are inactive
3000 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3001 lov_statfs divides by value returned by lov_get_stripecnt().
3005 Bugzilla : 9489, 3273
3006 Description: First write from each client to each OST was only 4kB in size,
3007 to initialize client writeback cache, which caused sub-optimal
3008 RPCs and poor layout on disk for the first writen file.
3009 Details : Clients now request an initial cache grant at (re)connect time
3010 and so that they can start streaming writes to the cache right
3011 away and always do full-sized RPCs if there is enough data.
3012 If the OST is rebooted the client also re-establishes its grant
3013 so that client cached writes will be honoured under the grant.
3018 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3019 Details : Now I/O RPCs go to different portal number and (presumably) fast
3020 lock requests (and glimses) and other RPCs get their own service
3021 threads pool that should be able to service those RPCs
3024 Severity : enhancement
3026 Description: Ability to exchange lustre version between client and servers and
3027 issue warnings at client side if client is too old. Also for
3028 liblustre clients there is ability to refuse connection of too old
3030 Details : New 'version' field is added to connect data structure that is
3031 filled with version info. That info is later checked by server and
3035 Frequency : rare, liblustre only.
3036 Bugzilla : 9296, 9581
3037 Description: Two simultaneous writes from liblustre at offset within same page
3038 might proceed at the same time overwriting eachother with stale
3040 Details : I/O lock withing llu_file_prwv was released too early, before data
3041 actually was hitting the wire. Extended lock-holding time until
3042 server acknowledges receiving data.
3045 Frequency : extremely rare. Never observed in practice.
3047 Description: avoid generating lustre_handle cookie of 0.
3048 Details : class_handle_hash() generates handle cookies by incrementing
3049 global counter, and can hit 0 occasionaly (this is unlikely, but
3050 not impossible, because initial value of cookie counter is
3051 selected randonly). Value of 0 is used as a sentinel meaning
3052 "unassigned handle" --- avoid it. Also coalesce two critical
3053 sections in this function into one.
3055 Severity : enhancement
3057 Description: allow liblustre clients to delegate truncate locking to OST
3058 Details : To avoid overhead of locking, liblustre client instructs OST to
3059 take extent lock in ost_punch() on client's behalf. New connection
3060 flag is added to handle backward compatibility.
3062 Severity : enhancement
3063 Bugzilla : 4928, 7341, 9758
3064 Description: allow number of OST service threads to be specified
3065 Details : a module parameter allows the number of OST service threads
3066 to be specified via "options ost ost_num_threads={N}" in the
3067 OSS's /etc/modules.conf or /etc/modprobe.conf.
3071 Bugzilla : 6146, 9635, 9895
3072 Description: servers crash with bad pointer in target_handle_connect()
3073 Details : In rare cases when a client is reconnecting it was possible that
3074 the connection request was the last reference for that export.
3075 We would temporarily drop the export reference and get a new
3076 one, but this may have been the last reference and the export
3077 was just destroyed. Get new reference before dropping old one.
3079 Severity : enhancement
3080 Frequency : if client is started with failover MDS
3082 Description: Allow multiple MDS hostnames in the mount command
3083 Details : Try to read the configuration from all specified MDS
3084 hostnames during a client mount in case the "primary"
3087 Severity : enhancement
3089 Description: Stop sending data to evicted clients as soon as possible.
3090 Details : Check if the client we are about to send or are sending data to
3091 was evicted already. (Check is done every second of waiting,
3092 for which l_wait_event interface was extended to allow checking
3093 of exit condition at specified intervals).
3096 Frequency : rare, normally only when NFS exporting is done from client
3098 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3100 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3103 Frequency : always, if srand() is called before liblustre initialization
3105 Description: Liblustre uses system PRNG disturbing its usage by user application
3106 Details : Introduce internal to lustre fast and high-quality PRNG for
3107 lustre usage and make liblustre and some other places in generic
3108 lustre code to use it.
3110 Severity : enhancement
3111 Bugzilla : 9477, 9557, 9870
3112 Description: Verify that the MDS configuration logs are updated when xml is
3113 Details : Check if the .xml configuration logs are newer than the config
3114 logs stored on the MDS and report an error if this is the case.
3115 Request --write-conf, or allow starting with --old_conf.
3117 Severity : enhancement
3119 Description: Handle symlinks in the path when checking if Lustre is mounted.
3120 Details : Resolve intermediate symlinks when checking if a client has
3121 mounted a filesystem to avoid duplicate client mounts.
3126 Description: lconf can hit an error exception but still return success.
3127 Details : The lconf command catches the Command error exception at the top
3128 level script context and will exit with the associated exit
3129 status, but doesn't ensure that this exit status is non-zero.
3134 Description: failure of ptlrpc thread startup can cause oops
3135 Details : Starting a ptlrpc service thread can fail if there are a large
3136 number of threads or the server memory is very fragmented.
3137 Handle this without oopsing.
3140 Frequency : always, only if liblustre and non-default acceptor port was used
3142 Description: liblustre cannot connect to servers with non-default acceptor port
3143 Details : tcpnal_set_default_params() was not called and was therefore
3144 ignoring the environment varaible TCPNAL_PORT, as well as other
3145 TCPNAL_ environment variables
3150 Description: two objects could be created on the same OST for a single file
3151 Details : If an OST is down, in some cases it was possible to create two
3152 objects on a single OST for a single file. No problems other
3153 than potential performance impact and spurious error messages.
3157 Bugzilla : 5681, 9562
3158 Description: Client may oops in ll_unhash_aliases
3159 Details : Client dcache may become inconsistent in race condition.
3160 In some cases "getcwd" can fail if the current directory is
3166 Description: Inode refcounting problems in NFS export code
3167 Details : link_raw functions used to call d_instantiate without obtaining
3168 extra inode reference first.
3172 Bugzilla : 9942, 9903
3173 Description: Referencing freed requests leading to crash, memleaks with NFS.
3174 Details : We used to require that call to ll_revalidate_it was always
3175 followed by ll_lookup_it. Also with revalidate_special() it is
3176 possible to call ll_revalidate_it() twice for the same dentry
3177 even if first occurence returned success. This fix changes semantic
3178 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3179 reference on a request referred from the intent.
3180 ll_intent_release() then releases such a request.
3183 Frequency : rare, normally benchmark loads only
3185 Description: unlinked inodes were kept in memory on the client
3186 Details : If a client is repeatedly creating and unlinking files it
3187 can accumulate a lot of stale inodes in the inode slab cache.
3188 If there is no other client load running this can cause the
3189 client node to run out of memory. Instead flush old inodes
3190 from client cache that have the same inode number as a new inode.
3193 Frequency : SLES9 2.6.5 kernel and long filenames only
3194 Bugzilla : 9969, 10379
3195 Description: utime reports stale NFS file handle
3196 Details : SLES9 uses out-of-dentry names in some cases, which confused
3197 the lustre dentry revalidation. Change it to always use the
3201 Frequency : rare, unless heavy write-truncate concurrency is continuous
3202 Bugzilla : 4180, 6984, 7171, 9963, 9331
3203 Description: OST becomes very slow and/or deadlocked during object unlink
3204 Details : filter_destroy() was holding onto the parent directory lock
3205 while truncating+unlinking objects. For very large objects this
3206 may block other threads for a long time and slow overall OST
3207 responsiveness. It may also be possible to get a lock ordering
3208 deadlock in this case, or run out of journal credits because of
3209 the combined truncate+unlink. Solution is to do object truncate
3210 first in one transaction without parent lock, and then do the
3211 final unlink in a new transaction with the parent lock. This
3212 reduces the lock hold time dramatically.
3215 Frequency : rare, 2.4 kernels only
3217 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3218 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3219 before unmounting their filesystem in order to avoid deadlock.
3220 The kernel_locked() function in 2.4 kernels only checks whether
3221 the kernel lock is held, not whether it is this process that is
3222 holding it as 2.6 kernels do.
3227 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3228 Details : The client ptlrpc code may be trying to reconnect to a down
3229 server before a previous connection attempt has timed out.
3230 Increase the reconnect interval to be longer than the connection
3231 timeout interval to avoid sending duplicate connections to
3235 Frequency : echo_client brw_test command
3237 Description: fix echo_client to work with OST preallocated code
3238 Details : OST preallocation code (5137) didn't take echo_client IO path
3239 into account: echo_client calls filter methods outside of any
3240 OST thread and, hence, there is no per-thread preallocated
3241 pages and buffers to use. Solution: hijack pga pages for IO. As
3242 a byproduct, this avoids unnecessary data copying.
3246 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3247 Description: Client can oops in mdc_commit_close() after open replay
3248 Details : It was possible for the MDS to return an open request with no
3249 transaction number in mds_finish_transno() if the client was
3250 evicted, but without actually returning an error. Clients
3251 would later try to replay that open and may trip an assertion
3252 Simplify the client close codepath, and always return an error
3253 from the MDS in case the open is not successful.
3256 Frequency : rare, 2.6 OSTs only
3258 Description: OST may deadlock under high load on fragmented files
3259 Details : If there was a heavy load and highly-fragmented OST filesystems
3260 it was possible to have all the OST threads deadlock waiting on
3261 allocation of biovecs, because the biovecs were not released
3262 until the entire RPC IO was completed. Instead, release biovecs
3263 as soon as they are complete to ensure forward IO progress.
3265 Severity : enhancement
3267 Description: Support for specifying external journal device at mount
3268 Details : If an OST or MDS device is formatted with an external journal
3269 device, this device major/minor is stored in the ext3 superblock
3270 and may not be valid for failover. Allow detecting and
3271 specifying the external journal at mount time.
3276 Description: Mounting an MDS with pending unlinked files may cause oops
3277 Details : target_finish_recovery() calls mds_postrecov() which returned
3278 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3279 considers this an error and immediately begins cleaning up the
3280 lov, just after starting the mds_lov process
3282 Severity : enhancement
3284 Description: Implement 'lfs df' to report actual free space on per-OST basis
3285 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3286 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3287 usage of objects; '-h' to report in human readable format.
3289 ------------------------------------------------------------------------------
3291 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3298 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3299 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3300 at startup during orphan unlinking. Instead, skip buddy update
3301 before setup, as it will be regenerated after recovery is complete.
3306 Description: If an OST is inactive, its locks might reference stale inodes.
3307 Details : lov_change_cbdata() must iterate over all namespaces, even if
3308 they are inactive to clear inode references from the lock.
3310 Severity : enhancement
3311 Frequency : occasional, if non-standard max_dirty_mb used
3313 Description: Client will block write RPCs if not enough grant
3314 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3315 then the client will block writes while waiting for another RPC
3316 to complete instead of consuming its dirty limit. With change
3317 we get improved performance when max_dirty_mb is small.
3319 Severity : enhancement
3320 Bugzilla : 3389, 6253
3321 Description: Add support for supplementary groups on the MDS.
3322 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3323 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3324 lookups for user supplementary groups into a cache.
3328 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3329 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3330 for liblustre O_EXCL is enforced by MDS.
3334 Description: readdir never returns NULL in liblustre.
3335 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3336 dentry in struct dirent.
3340 Frequency : liblustre only, depends on application IO pattern
3341 Description: liblustre clients evicted if not contacting servers
3342 Details : Don't put liblustre clients into the ping_evictor list, so
3343 they will not be evicted by the pinger ever.
3345 Severity : enhancement
3347 Description: Add ability to evict clients by NID from MDS.
3348 Details : By echoing "nid:$NID" string into
3349 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3350 $NID would be instantly evicted from this MDS and from all active
3351 OSTs connected to it.
3355 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3356 Details : lookup_it_finish() used to query file size from OSTs that was not
3361 Description: service threads change working directory to that of init
3362 Details : Starting lustre service threads may pin the working directory
3363 of the parent thread, making that filesystem busy. Threads
3364 now change to the working directory of init to avoid this.
3368 Frequency : during shutdown only
3369 Description: shutdown with a failed MDS or OST can cause unmount to hang
3370 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3375 Frequency : 2.6 only
3376 Description: chmod/chown may include an extra supplementary group
3377 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3378 supplementary group and if none is specified this is used.
3381 Bugzilla : 5479 (6816)
3383 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3384 Details : If lookup is in progress on a file that is unlinked we might try
3385 to revalidate the inode and fail in revalidate after lookup is
3386 complete and ll_file_open() enqueues the open again but
3387 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3390 Frequency : always, if lconf --abort_recovery used
3392 Description: lconf --abort_recovery fails with 'Operation not supported'
3393 Details : lconf was attempting to abort recovery on the MDT device and not
3396 ------------------------------------------------------------------------------
3398 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3403 Frequency : rare (only unsupported configurations with a node running as an
3405 Bugzilla : 6514, 5137
3406 Description: Mounting a Lustre file system on a node running as an OST could
3408 Details : OSTs now preallocates memory needed to write out data at
3409 startup, instead of when needed, to avoid having to
3410 allocate memory in possibly low memory situations.
3411 Specifically, if the file system is mounted on on OST,
3412 memory pressure could force it to try to write out data,
3413 which it needed to allocate memory to do. Due to the low
3414 memory, it would be unable to do so and the node would
3415 become unresponsive.
3417 Severity : enhancement
3419 Description: Addition of lconf --service command line option
3420 Details : lconf now accepts a '--service <arg>' option, which is
3421 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3423 Severity : enhancement
3425 Description: Failover mode is now the default for OSTs.
3426 Details : By default, OSTs will now run in failover mode. To return to
3427 the old behaviour, add '--failout' to the lmc line for OSTs.
3429 Severity : enhancement
3431 Description: Health checks are now provided for MDS and OSTs
3432 Details : Additional detailed health check information on MSD and OSTs
3433 is now provided through the procfs health_check value.
3436 Frequency : occasional, depends on IO load
3438 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3439 after numerous create/delete cycles
3440 Details : The ext3 inode allocation policy would not allocate new inodes
3441 very well on the OSTs because there are no new directories
3442 being created. Instead we look for groups with free space if
3443 the parent directories are nearly full.
3448 Description: Network or server problems during mount may cause partially
3449 mounted clients instead of returning an error.
3450 Details : The config llog parsing code may overwrite the error return
3451 code during mount error handling, returning success instead
3457 Description: MDS can fail to allocate large reply buffers
3458 Details : After long uptimes the MDS can fail to allocate large reply
3459 buffers (e.g. zconf client mount config records) due to memory
3460 fragmentation or consumption by the buffer cache. Preallocate
3461 some large reply buffers so that these replies can be sent even
3462 under memory pressure.
3466 Frequency : rare (liblustre)
3467 Description: fsx running with liblustre complained that using truncate() to
3468 extend the file doesn't work. This patch corrects that issue.
3469 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3470 ATTR_SIZE and lsm use in llu_setattr_raw.
3474 Frequency : rare, only 2.6 kernels
3475 Description: Unusual file access patterns on the MDS may result in inode
3476 data being lost in very rare circumstances.
3477 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3478 code in the 2.6 kernel caused the inode and/or EA data not to
3479 be read from disk, causing single-file corruption.
3483 Frequency : rare, only 2.6 filesystems using extents
3484 Description: Heavy concurrent write and delete load may cause data corruption.
3485 Details : It was possible under high-load situations to have an extent
3486 metadata block in the block device cache from a just-unlinked
3487 file overwrite a newly-allocated data block. We now unmap any
3488 metadata buffers that alias just-allocated data blocks.
3492 Frequency : filesystems with default stripe_count larger than 77
3493 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3494 Details : lconf specifies an inode size of 4096 bytes when the default
3495 stripe_count is larger than 77. This conflicts with the default
3496 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3497 this case to avoid pinning too much memory for large EAs.
3499 ------------------------------------------------------------------------------
3501 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3506 Frequency : rare (extremely heavy IO load with hundreds of clients)
3508 Description: Client is evicted, gets IO error writing to file
3509 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3510 caused clients to be evicted for AST timeouts. The fixes in
3511 bug 5192 mean we no longer need to have such short AST timeouts
3512 so ldlm_timeout has been increased.
3515 Frequency : occasional during --force or --failover shutdown under load
3516 Bugzilla : 5949, 4834
3517 Description: Server oops/LBUG if stopped with --force or --failover under load
3518 Details : a collection of import/export refcount and cleanup ordering
3519 issues fixed for safer force cleanup
3522 Frequency : only filesystems larger than 120 OSTs
3523 Bugzilla : 5990, 6223
3524 Description: lfs getstripe would oops on a very large filesystem
3525 Details : lov_getconfig used kfree on vmalloc'd memory
3528 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3529 Bugzilla : 6242, 6243
3530 Description: reading from files that had been truncated to a non-zero size
3531 but never opened returned no data
3532 Details : ll_file_read() reads zeros from no-object files to EOF
3537 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3538 Details : There's little harm in aborting MDS/OSS recovery and letting it
3539 try again, so I removed the LASSERT and return an error instead.
3541 Severity : enhancement
3543 Description: New debugging infrastructure for tracking down data corruption
3544 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3545 (b) cover more of the client-side code path, and (c) try to narrow
3546 down where problems occurred
3550 Bugzilla : 3819, 4364, 4397, 6313
3551 Description: Racing close and eviction MDS could cause assertion in mds_close
3552 Details : It was possible to get multiple mfd references during close and
3553 client eviction, leading to one thread referencing a freed mfd.
3555 Severity: : enhancement
3556 Bugzilla : 3262, 6359
3557 Description: Attempts to reconnect to servers are now more aggressive.
3558 Details : This builds on the enhanced upcall-less recovery that was added
3559 in 1.4.2. When trying to reconnect to servers, clients will
3560 now try each server in the failover group every 10 seconds. By
3561 default, clients would previously try one server every 25 seconds.
3566 Description: After recovery, certain operations trigger a failed
3567 assertion on a client.
3568 Details : Failing over an mds, using lconf -d --failover, while a
3569 client was doing a readdir() call would cause the client to
3570 LBUG after recovery completed and the readdir() was resent.
3572 Severity : enhancement
3574 Description: Default groups are now added by lconf
3575 Details : You can now run lconf --group <servicename> without having to
3576 manually add groups with lmc.
3579 Frequency : occasional
3581 Description: Nodes with an elan id of 0 trigger a failed assertion
3584 Frequency : always when accessing e.g. tty/console device nodes
3586 Description: tty and some other devices nodes cannot be used on lustre
3587 Details : file's private_data field is used by device data and lustre
3588 values in there got lost. New field was added to struct file to
3589 store fs-specific private data.
3592 Frequency : when exporting Lustre via NFS
3594 Description: NFSD failed occasionally when looking up a path component
3595 Details : NFSD is looking up ".." which was broken in ext3 directories
3596 that had grown large enough to become hashed.
3599 Frequency : Clusters with multiple interfaces not on the same subnet
3601 Description: Nodes will repeatedly try to reconnect to an interface which it
3602 cannot reach and report an error to the log.
3603 Details : Extra peer list entries will be created by lconf with some peers
3604 unreachable. lconf now validates the peer before adding it.
3607 Frequency : Only if a default stripe is set on the filesystem root.
3609 Description: Setting a default stripe on the filesystem root prevented the
3610 filesystem from being remounted.
3611 Details : The client was sending extra request flags in the root getattr
3612 request and did not allocate a reply buffer for the dir EA.
3615 Frequency : occasional, higher if lots of files are accessed by one client
3616 Bugzilla : 6159, 6097
3617 Description: Client trips assertion regarding lsm mismatch/magic
3618 Details : While revalidating inodes the VFS looks up inodes with ifind()
3619 and in rare cases can find an inode that is being freed.
3620 The ll_test_inode() code will free the lsm during ifind()
3621 when it finds an existing inode and then the VFS later attaches
3622 this free lsm to a new inode.
3626 Bugzilla : 6422, 7030
3627 Description: MDS deadlock between mkdir and client eviction
3628 Details : Creating a new file via mkdir or mknod (starting a transaction
3629 and getting the ns lock) can deadlock with client eviction
3630 (gets ns lock and trying to finish a synchronous transaction).
3633 Frequency : occasional
3634 Description: While starting a server, the fsfilt_ext3 module could not be
3636 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3637 kernels. Previously, lconf would still use the ext3 name
3638 when trying to load modules. Now, it will correctly use
3639 ext3 on 2.4 and ldiskfs on 2.6.
3641 Severity : enhancement
3642 Description: The default stripe count has been changed to 1
3643 Details : The interpretation of the default stripe count (0, to lfs
3644 or lmc) has been changed to mean striping across a single
3645 OST, rather than all available. For general usage we have
3646 found a stripe count of 1 or 2 works best.
3648 Severity : enhancement
3649 Description: Add support for compiling against Cray portals.
3650 Details : Conditional compiling for some areas that are different
3654 Frequency : occasional
3655 Bugzilla : 6409, 6834
3656 Description: Creating files with an explicit stripe count may lead to
3657 a failed assertion on the MDS
3658 Details : If some OSTs are full or unavailable, creating files may
3659 trigger a failed assertion on the MDS. Now, Lustre will
3660 try to use other servers or return an error to the
3664 Frequency : occasional
3666 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3667 caused lock timeout during readahead (since 1.4.2).
3668 Details : Processes doing readahead might match a lock that hasn't been
3669 granted yet if there are overlapping and conflicting lock
3670 requests. The readahead process waits on ungranted lock
3671 (original lock is CBPENDING), while OST waits for that process
3672 to cancel CBPENDING read lock and eventually evicts client.
3674 Severity : enhancement
3676 Description: Initial enabling of flock support for clients
3677 Details : Implements fcntl advisory locking and file status functions.
3678 This feature is provided as an optional mount flag (default
3679 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3680 locking are implemented yet, and those that are are not guaranteed
3681 to be completely correct in production environments.
3682 mount -t lustre -o [flock|noflock] ...
3685 Frequency : occasional
3687 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3688 assertion in the ext3 JBD (journaling) layer.
3689 Details : The b_committed_data struct is protected by the big kernel lock
3690 in 2.4 kernels, serializing journal_commit_transaction() and
3691 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3692 access to this struct. In 2.6 kernels there is finer grained
3693 locking to improve SMP performance of the JBD layer.
3697 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3702 Description: merge and cleanup kernel patches.
3705 -----------------------------------------------------------------------------
3707 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3709 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3710 versions of Lustre. Please update all servers and clients to
3711 version 1.4.2 or later at the same time. You must also run
3712 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3713 to update the configuration logs.
3715 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3716 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3717 - swab llog records of type '0' so we get proper header size/idx (5861)
3718 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3719 - fix rename of one directory over another leaking an inode (5953)
3720 - avoid SetPageDirty on 2.6 (5981)
3721 - don't re-add just-being-destroyed locks to the waiting list (5653)
3722 - when creating new directories, inherit the parent's custom
3723 striping settings if present parent (3048)
3724 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3725 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3726 - don't hold client locks on temporary worklist from l_lru (5666)
3727 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3728 - automatically evict dead clients (5921)
3729 - Update file size properly in create+truncate+fstat case (6196)
3730 - Do not unhash mountpoint dentries, do not allow removal of
3732 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3733 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3734 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3735 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3736 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3737 - Correctly handle reads of files with no objects (6243)
3738 - lctl recover will also mark a device active if deactivate used (5933)
3740 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3741 - allow --write-conf on an MDS with different nettype than client (5619)
3742 - don't write config llogs to MDS for mounts not from that MDS (5617)
3743 - lconf should create multiple TCP connections from a client (5201)
3744 - init scripts are now turned off by default; run chkconfig --on
3745 lustre and chkconfig --on lustrefs to use them
3746 - upcalls are no longer needed for clients to recover to failover
3748 - add --abort-recovery option to lconf to abort recovery on device
3750 - add support for an arbitrary number of OSTs (3026)
3751 - Quota support protocol changes.
3752 - forward compatibility changes to wire structs (6007)
3753 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3754 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3755 - improved "lustre status" script
3756 - initialize blocksize for non-regular files (6062)
3757 - added --disable-server and --disable-client configure options (5782)
3758 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3759 - Vanilla 2.4.29 support
3760 - increase maximum number of obd devices to 520 (6242)
3761 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3762 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3763 SLES 9 2.6.5 kernels
3764 - limit stripes per file to 160 (the maximum EA size) (6093)
3766 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3769 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3770 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3771 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3772 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3773 - small ext3 extents cleanups and fixes (5733)
3774 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3775 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3776 - avoid needless client->OST connect, fix handle mismatch (5317)
3777 - fix DLM error path that led to out-of-sync client, long delays (5779)
3778 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3779 - fix several locking issues related to i_size (5492,5624,5654,5672)
3780 - don't move pending lock onto export if it is already evicted (5683)
3781 - fix kernel oops when creating .foo in unlinked directory (5548)
3782 - fix deadlock in obdfilter statistics vs. object create (5811)
3783 - use time_{before,after} to avoid timer jiffies wrap (5882)
3784 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3785 - Do not leak request if server was not able to process it (5154)
3786 - If mds_open unable to find parent dir, make that negative lookup(5154)
3787 - don't create new directories with extent-mapping (5909, 5936)
3789 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3790 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3791 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3792 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3793 - support for RHEL4 kernel on i686 (5773)
3794 - provide error messages when incompatible logs are encountered (5898)
3796 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3797 * version 1.4.0.10 (1.4.1 release candidate 1)
3799 - don't keep a lock reference when lock is not granted (4238)
3800 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3801 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3802 - drop import inflight refcount on signal_completed_replay error (5255)
3803 - unlock page after async write error during send (3677)
3804 - handle missing objects in filter_preprw_read properly (5265)
3805 - no transno return for symlink open, don't save no-trasno open (3440)
3806 - don't try to complete elan receive that already failed (4012)
3807 - free RPC server reply state on error (5406)
3808 - clean up thread from ptlrpc_start_thread() on error (5160)
3809 - readahead could read extra page into cache that wasn't ejected (5388)
3810 - prevent races in class_attach/setup/cleanup/detach (5260)
3811 - don't dereference de->d_inode after l_dput of de (5458)
3812 - use "int" for stripe value returned from lock_to_stripe (5544)
3813 - mballoc allocation and error-checking fixes in 2.6 (5504)
3814 - block device patches to fix I/O request sizes in 2.6 (5482)
3815 - look up hostnames for IB nals (5602)
3816 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3817 - don't start multiple acceptors for the same port (5277)
3818 - fix incorrect LASSERT in mds_getattr_name (5635)
3819 - export a proc file for general "ping" checking (5628)
3820 - fix "lfs check" to not block when the MDS is down (5628)
3822 - service request history (4965)
3823 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3824 - create an "evict_client" /proc entry on OSTs, like the MDS has
3825 - fix mount usage message, return errors per mount(8) (5168)
3826 - change grep [] to grep "[]" in tests so they work in more UMLs
3827 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3828 - remove ancient LOV_MAGIC_V0 EA support (5047)
3829 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3830 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3831 - allow readahead allocations to fail when low on memory (5383)
3832 - mmap locking landed again, after considerable improvement (2828)
3833 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3835 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3838 - send OST transaction number in read/write reply to free req (4966)
3839 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3840 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3841 - the watchdog thread now runs as interruptible (5246)
3842 - flock/lockf fixes (but it's still disabled, pending 5135)
3843 - don't use EXT3 constants in llite code (5094)
3844 - memory shortage at startup could cause assertion (5176)
3846 - reorganization of lov code
3847 - single portals codebase
3849 - add extents/mballoc support (5025)
3850 - direct I/O reads in the obdfilter (4048)
3851 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3853 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3856 - send OST transaction number in read/write reply to free req (4966)
3857 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3858 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3859 - the watchdog thread now runs as interruptible (5246)
3860 - handle missing objects in filter_preprw_read properly (5265)
3861 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3862 - drop import inflight refcount on signal_completed_replay error (5255)
3863 - unlock page after async write error during send (3677)
3864 - return original error code on reconstructed replies (3761)
3865 - no transno return for symlink open, don't save no-trasno open (3440)
3867 - add pid to ldlm debugging output (4922)
3868 - bump the watchdog timeouts -- we can't handle 30sec yet
3869 - extra debugging for orphan dentry/inode bug (5259)
3871 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3874 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3875 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3876 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3877 - fix NULL dereference in /proc/sys/portals/routes (4827)
3878 - allow failed mdc_close() operations to be interrupted (4561)
3879 - stop precreate on OST before MDS would time out on it (4778)
3880 - don't send partial-page writes before EOF from client (4410)
3881 - discard client grant for sub-page writes on large-page clients (4520)
3882 - don't free dentries not owned by NFS code, check generation (4806)
3883 - fix lsm leak if mds_create_objects() fails (4801)
3884 - limit debug_daemon file size, always print CERROR messages (4789)
3885 - use transno after validating reply (3892)
3886 - process timed out requests if import state changes (3754)
3887 - update mtime on OST during writes, return in glimpse (4829)
3888 - add mkfsoptions to LDAP (4679)
3889 - use ->max_readahead method instead of zapping global ra (5039)
3890 - don't interrupt __l_wait_event() during strace
3892 - add software watchdogs to catch hung threads quickly (4941)
3893 - make lustrefs init script start after nfs is mounted
3894 - fix CWARN/ERROR duplication (4930)
3895 - return async write errors to application if possible (2248)
3896 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3897 - print NAL number in %x format (4645)
3898 - update barely-supported suse-2.4.21-171 series (4842)
3899 - support for sles 9 %post scripts
3900 - support for building 2.6 kernel-source packages
3901 - support for sles km_* packages
3903 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3906 - ignore -ENOENT errors in osc_destroy (3639)
3907 - notify osc create thread that OSC is being cleaned up (4600)
3908 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3909 - reconstruct ost_handle() like mds_handle() (4657)
3910 - create a new thread to do import eviction to avoid deadlock (3969)
3911 - let lconf resolve symlinked-to devices (4629)
3912 - don't unlink "objects" from directory with default EA (4554)
3913 - hold socknal file ref over connect in case target is down (4394)
3914 - allow more than 32000 subdirectories in a single directory (3244)
3915 - fix blocks count for O_DIRECT writes (3751)
3916 - OST returns ENOSPC from object create when no space left (4539)
3917 - don't send truncate RPC if file size isn't changing (4410)
3918 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3919 - bind to privileged port in socknal and tcpnal (3689)
3921 - rate limit CERROR/CWARN console message to avoid overload (4519)
3922 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3923 - basic mmap support (3918)
3924 - kernel patch series update from b1_4 (4711)
3926 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3929 - avoid crash during MDS cleanup with OST shut down (2775)
3930 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3931 - don't use bad inodes on the MDS (3744)
3932 - dynamic object preallocation to improve recovery speed (4236)
3933 - don't hold spinlock over lock dumping or change debug flags (4401)
3934 - don't zero obd_dev when it is force cleaned (3651)
3935 - print grants to console if they go negative (4431)
3936 - "lctl deactivate" will stop automatic recovery attempts (3406)
3937 - look for existing locks in ldlm_handle_enqueue() (3764)
3938 - don't resolve lock handle twice in recovery avoiding race (4401)
3939 - revalidate should check working dir is a directory (4134)
3941 - don't always mark "slow" obdfilter messages as errors (4418)
3943 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3946 - don't close LustreDB during write_conf until it is done (3860)
3947 - fix typo in lconf for_each_profile (3821)
3948 - allow dumping logs from multiple threads at one time (3820)
3949 - don't allow multiple threads in OSC recovery (3812)
3950 - fix debug_size parameters (3864)
3951 - fix mds_postrecov to initialize import for llog ctxt (3121)
3952 - replace config semaphore with spinlock (3306)
3953 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3954 - don't allow enqueue to complete on a destroyed export (3822)
3955 - down write_lock before checking llog header bitmap (3825)
3956 - recover from lock replay timeout (3764)
3957 - up llog sem before sending rpc (3652)
3958 - reduce ns lock hold times when setting kms (3267)
3959 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3960 - fix NULL deref and obd_dev leak on setup error (3312)
3961 - replace some LBUG about llog ops with error handling (3841)
3962 - don't match INVALID dentries from d_lookup and spin (3784)
3963 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3964 - fix invalid assertion in ptlrpc_set_wait (3880)
3966 - add libwrap support for the TCP acceptor (3996)
3967 - add /proc/sys/portals/routes for non-root route listing (3994)
3968 - allow setting MDS UUID in .xml (2580)
3969 - print the stack of a process that LBUGs (4228)
3971 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3974 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3975 - make sure to unset replay flag from failed open requests (3440)
3976 - if default stripe count is 0, use OST count for inode size (3636)
3977 - update parent mtime/ctime on client for create/unlink (2611)
3978 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3979 - free recovery state on server during a forced cleanup (3571)
3980 - unregister_reply for resent reqs (3063)
3981 - loop back devices mounting and status check on 2.6 (3563)
3982 - fix resource-creation race that can provoke i_size == 0 (3513)
3983 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3984 - more debugging for page-accounting assertion (3746)
3985 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3986 - avoid deadlock after precreation failure (3758)
3987 - fix race and lock order deadlock in orphan handling (3450, 3750)
3988 - add validity checks when grabbing inodes from l_ast_data (3599)
3990 - add /proc/.../recovery_status to obdfilter (3428)
3991 - lightweight CDEBUG infrastructure, debug daemon (3668)
3992 - change default OSC RPC parameters to be better on small clusters
3993 - turn off OST read cache for files smaller than 32MB
3994 - install man pages and include them in rpms (3100)
3995 - add new init script for (un)mounting lustre filesystems (2593)
3996 - run chkconfig in %post for init scripts (3701)
3997 - drop scimac NAL (unmaintained)
3999 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4002 - clean kiobufs before and after use (3485)
4003 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4004 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4005 - change init script to fail more clearly if not run as root (1528)
4006 - allow clients to reconnect during replay (1742)
4007 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4008 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4009 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4010 - fix cond_resched (3554)
4011 - only evict unfinished clients after recovery (3515)
4012 - allow bulk resend, prevent data loss (3570)
4013 - dynamic ptlrpc request buffer allocation (2102)
4014 - don't allow unlinking open directory if it isn't empty (2904)
4015 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4016 - remove extraneous obd dereference causing LASSERT failure (3334)
4017 - don't use get_cycles() when creating temp. files on the mds (3156)
4018 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4019 - handle EEXIST for set-stripe, set proper directory name (3336)
4021 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4022 - fix ksocknal_fmb_callback() error messages (2918)
4024 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4027 - don't copy lvb into (possibly NULL) reply on error (2983)
4028 - don't deref dentry after dput, don't free lvb on error (2922)
4029 - use the kms to determine writeback rpc length (2947)
4030 - increment oti_logcookies when osc is inactive (2948)
4031 - update client's i_blocks count via lvb messages (2543)
4032 - handle intent open/close of special files properly (1557)
4033 - mount MDS with errors=remount-ro, like obdfilter (2009)
4034 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4035 - don't use cancelling-locks' kms values (2947)
4036 - use highest lock extent for kms, not last one (2925)
4037 - don't dereference ERR_PTR() dentry in error handling path (3107)
4038 - fix thread race in portals_debug_dumplog() (3122)
4039 - create lprocfs device entries at setup instead of at attach (1519)
4040 - common AST error handler, don't evict client on completion race (3145)
4041 - zero nameidata in detach_mnt in 2.6 (3118)
4042 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4043 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4044 - initialize RPC timeout timer earlier for 2.6 (3219)
4045 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4046 - print nal/nid for unknown nid (3258)
4047 - additional checks for oscc recovery before doing precreate (3284)
4048 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4049 - don't crash in mdc_close for bad permissions on open (3285)
4050 - zero i_rdev for non-device files (3147)
4051 - clear page->private before handing to FS, better assertion (3119)
4052 - tune the read pipeline (3236)
4053 - fix incorrect decref of invalidated dentry (2350)
4054 - provide read-ahead stats and refine rpc in flight stats (3328)
4055 - don't hold journal transaction open across create RPC (3313)
4056 - update atime on MDS at close time (3265)
4057 - close LDAP connection when recovering to avoid server load (3315)
4058 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4059 - don't leak open file on MDS after open resend (3325)
4060 - serialize filter_precreate and filter_destroy_precreated (3329)
4061 - loop device shouldn't call sync_dev() for nul device (3092)
4062 - clear page cache after eviction (2766)
4063 - resynchronize MDS->OST in background (2824)
4064 - refuse to mount the same filesystem twice on same mountpoint (3394)
4065 - allow llmount to create routes for mounting behind routers (3320)
4066 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4067 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4068 - fix rare problem with rename on htree directories (3417)
4070 - allow default OST striping configuration per directory (1414)
4071 - fix compilation for qswnal for 2.6 kernels (3125)
4072 - increase maximum number of MDS request buffers for large systems
4073 - change liblustreapi to be useful for external progs like lfsck (3098)
4074 - increase local configuration timeout for slow disks (3353)
4075 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4077 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4080 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4081 - don't overwrite extent policy data in reply if lock was blocked (2901)
4082 - drop filter export grants atomically with removal from device (2663)
4083 - del obd_self_export from work_list in class_disconnect_exports (2908)
4084 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4085 - swab reply message in mdc_close, other PPC fixes (2464)
4086 - fix destroying of named logs (2325)
4087 - overwrite old logs when running lconf --write_conf (2264)
4088 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4089 - fix race in target_handle_connect (2898)
4090 - mds_reint_create() should take same inode create lock (2926)
4091 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4092 - don't close files for self_export to avoid uninitialized obd (2936)
4093 - allow MDS with the same name as client node (2939)
4094 - hold dentry reference for closed log files for unlink (2325)
4095 - reserve space for all logs during transactions (2059)
4096 - don't evict page beyond end of stripe extent (2925)
4097 - don't oops on a deleted current working directory (2399)
4098 - handle hard links to targets without a parent properly (2517)
4099 - don't dereference NULL lock when racing during eviction (2867)
4100 - don't grow lock extents when lots of conflicting locks (2919)
4102 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4105 - account for cache space usage on clients to avoid data loss (974)
4106 - lfsck support in lustre kernel code (2349)
4107 - reduce journal credits needed for BRW writes (2370)
4108 - orphan handling to avoid losing space on client/server crashes
4109 - ptlrpcd can be blocked, stopping ALL progress (2477)
4110 - use lock value blocks to assist in proper KMS, faster stat (1021)
4111 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4112 - recovery for initial connections (2355)
4113 - fixes for mds_cleanup_orphans (1934)
4114 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4115 - block all file creations until orphan recovery completes (1901)
4116 - client remove rq_connection from request struct (2423)
4117 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4118 - recovery timer race (2670)
4119 - mdc_close recovey bug (2532)
4120 - ptlrpc cleanup bug (2710)
4121 - mds timeout on local locks (2588)
4122 - namespace lock held during RPCs (2431)
4123 - handle interrupted sync write properly (2503)
4124 - don't try to handle a message that hasn't been replied to (2699)
4125 - client assert failure during cleanup after abort recovery (2701)
4126 - leak mdc device after failed mount (2712)
4127 - ptlrpc_check_set allows timedout requests to complete (2714)
4128 - wait for inflight reqs when ptlrpcd finishes (2710)
4129 - make sure unregistered services are removed from the srv_list
4130 - reset bulk XID's when resending them (caught by 1138 test)
4131 - unregister_bulk after timeout
4132 - fix lconf error (2694)
4133 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4134 - readahead locks pages, leaves pending causing memory pressure (2673)
4135 - increase OST request buffers to 4096 on large machines (2729)
4136 - fix up permission of existing directories in simple_mkdir (2661)
4137 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4138 - don't assign transno to errored transactions (2742)
4139 - don't delete objects on OST if given a bogus objid from MDS (2751)
4140 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4141 - if rq_no_resend, then timeout request after recovery (2432)
4142 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4143 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4144 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4145 - MDS recovery shouldn't delete objects that it has given out (2730)
4146 - if enqueue arrives after completion, don't clobber LVB (2819)
4147 - don't unlock pages twice when trigger_group_io returns error (2814)
4148 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4149 - don't write pages to disk if there was an error (1450)
4150 - don't ping imports that have recovery disabled (2676)
4151 - take buffered bytes into account when balancing socknal conn (2817)
4152 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4153 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4154 - remove little-endian swabbing of llog records (1987)
4155 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4156 - retry reposting request buffers if they fail (1191)
4157 - grow extent at grant time to avoid granting a revoked lock (2809)
4158 - lock revoke doesn't evict page if covered by a second lock (2765)
4159 - disable VM readahead to avoid reading outside lock extents (2805)
4161 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4162 - updated kernel patches for hp-2.4.20 kernel (2681)
4164 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4167 - fix truncated write corruption (2366)
4168 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4170 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4171 - ensure len(uuid) < 37 in lmc (1171)
4172 - fix ia64 OOPS in llog_test (2255)
4173 - zero end of page at obdfilter for partial page writes (2648)
4174 - don't leave stale dentries around after renames (bug 2428)
4175 - fix timeouts when evicting a client with a single lock held (2642)
4176 - set deadline for the initial HELLO message to drain (2634)
4177 - print out dotted-quad IP addresses in the socknal (2302)
4179 - additional debugging for MDS client eviction problem (2443)
4180 - fix mkfsoptions support for osts (2603, 2604)
4182 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4185 - add series for the vanilla 2.6.0 kernel
4186 - add series for the vanilla 2.4.24 kernel
4187 - add series for a cray x86/64 UL kernel drop
4188 - fix xattr patches for the vanilla 2.4.19 series
4190 - generate true UUIDs in lmc (1171)
4191 - have portals stack dumping break in UML (2466)
4192 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4193 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4194 - wake callback waiting threads on client eviction (2460)
4195 - Add --ptldebug and --subsystem to lmc (1719)
4196 - update assertion to allow safe interrupt allocation
4197 - set rq_no_resend for cancel requests (2432)
4198 - recalculate ptlrpcd timeout after resend (2494)
4199 - call vfs_rmdir when removing pending directories (2368)
4200 - fix renaming a file to itself (2429)
4201 - lmc creates a default one-stripe lov (2454)
4202 - expand procfs space to handle large clusters (2326)
4203 - increase UML stack to avoid overflow
4204 - update lconf's list of debug and subsystem masks
4205 - fix lfs find --obd (2510)
4206 - /proc tunable for disabling filter read caching (2591)
4207 - stop rpm packages from altering slapd.conf (2301)
4208 - disable nagle in the socknal under 0conf (2578)
4209 - choose mds inode size based on stripe count (2572)
4210 - fix kernel-source rpm problems (2516)
4212 - add --disable-doc to avoid pdf generation (2421)
4213 - update documentation, tests, type-os, comments
4214 - avoid format warnings on ia64
4215 - remove the TOE NAL
4216 - tiny code cleanups by removing unused fields
4218 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4221 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4222 - remove the most blatant lies from BUILDING file (2371)
4223 - change default debug level to reasonable production setting
4224 - reduce client side cache size to reduce cache flush time
4225 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4226 - make TCP zerocopy and pinger support enabled by default (2476)
4227 - sync writes completed after process exits caused crashes (2319)
4228 - maintain correct mount count on the MDS (2356)
4229 - backout 1557, because 2316 wasn't really fixed
4230 - better file I/O statistics gathering in /proc
4231 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4232 - another kernel patch to fix zero-copy TCP function export
4233 - don't take duplicate lock when processing re-sent getattr (2420)
4234 - lctl uses obd_self_export instead of creating new conn (2353)
4235 - MDS/OST recovery case which requires object creation asserted (2425)
4236 - move lfs from /usr/sbin to /usr/bin in packages
4237 - fix race between mds_client_add and mds_client_free (2417)
4238 - use kmalloc instead of slabs in portals (2430)
4239 - don't create duplicate records when a failover MDS is present (2442)
4240 - remove unnecessary mount age check (2332)
4241 - don't remove directory inodes from locks prematurely (2451)
4242 - don't break if MDS service name is the same as hostname (2103)
4243 - fix races in client write RPC generation when cache full (2482)
4245 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4248 - remove now-unused request->rq_obd (278)
4249 - if an allocation fails, print out how much memory we've used (1933)
4250 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4251 - add the "configurable stack size" patch to most series files (1256)
4252 - ability to write large log records, for 100+ OST configs (2306)
4253 - fix NULL deref when filter_prep fails (2314)
4254 - fix operator precedence error in filter_sync
4255 - dynamic allocation of socknal TX descriptors (2315)
4256 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4257 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4258 - ensure that CWARN messages in Portals always get to the syslog
4259 - __init/__exit are not for prototype decls (ldlm_init/exit)
4260 - x86-64 compile warning fixes
4261 - fix gateway LMC keyword conflict (2318)
4262 - fix MDS lock inversions in getattr/reint paths (1844)
4263 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4264 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4265 - DLM race condition prevented some lock evictions (2328)
4266 - ENOMEM detection and retry on socknal sends (2230)
4267 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4268 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4269 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4270 - better error messages when a client is rejected during recovery (1505)
4271 - avoid cancelling locks which were never granted, after failure (2330)
4272 - fix i_sem/journal inversion in mds_client_add (2333)
4273 - fix truncate/getattr lock cycle deadlock (2334)
4274 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4275 - fix two rare exit paths which could leak an l_lock() ref (2321)
4276 - fencepost error in MDS/OST orphan recovery (2226)
4277 - make log record alignment 8 bytes (1988)
4278 - lstripe now fails when requested offset > ost_count (2237)
4279 - ensure that all kernel series have a complete list.h (1607)
4280 - fix crashes in special-file operations (2316)
4281 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4283 - allow configurable automake binary, for testing new versions
4284 - small update to the lfs documentation
4286 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4288 * fix negative export reference count in fsfilt_sync (2312)
4290 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4291 * release candidate 0.9.1
4293 - orphans are moved into the PENDING directory for possible recovery
4294 - replayed opens now open by fid for orphan/rename safety (1042)
4295 - last close of an orphan inode generates a transno (683)
4296 - chdir() and mount() now pin the directory entry (1020)
4297 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4298 - discard very old requests without processing them (1502)
4299 - remove some common, well-understood CERRORs (1505)
4300 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4301 - clear "grant" flags in OST replies until OST grant code lands (1644)
4302 - fix read performance by not clobbering i_blksize on client (1598)
4303 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4304 - make LRU size a /proc tunable, clears locks when reduced (707)
4305 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4306 - server locks take references on exports now (1558)
4307 - build fixes for 2.4.20-rh trees (1663)
4308 - return an error from lov_create if all OSCs are inactive (1751)
4309 - fix import levels when a reconnect happens without a timeout (1597)
4310 - exit early from mds_open if we get a lookup error (1749)
4311 - partial page read at EOF wouldn't wait for disk before sending (1642)
4312 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4313 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4314 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4315 - write/truncate lock inversion (1639)
4316 - Don't auto-load obdclass, portals modules during cleanup (1495)
4317 - fix timestamps from jumping to "now" (1763)
4318 - extra journal assertions (1648)
4319 - add an extra multiunlink test (1771)
4320 - fix read_record/write_record API (1776)
4321 - fix leak of offset_extent, possible incorrect i_size later (1772)
4322 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4323 - Debugging for the kqswnal_get_idle_tx problems (1820)
4324 - Allow recovery to be attempted multiple times (1536)
4325 - Write out MDS last_rcvd file after it is first created (1600)
4326 - Fix tx_descriptor leak in failed transmit situations (1827)
4327 - ext3 journaling fixes for assertion failure after IO error (1871)
4328 - class_export_put() on freed export after completion AST error (1896)
4329 - Fix revalidate looping in VFS (1322)
4330 - Don't access a freed export during MDS_REINT timeout (1521)
4331 - Add open-unlink recovery support on the MDS (1673,1764)
4332 - Return an error if no MDS data was read from last_rcvd (1946)
4333 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4334 - Don't setup a disk device that doesn't match exported UUID (317)
4335 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4336 - avoid committing NULL handle in force close
4337 - local.sh is now a one-stripe LOV configuration
4338 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4339 - fix ext3 htree duplicate directory entry corruption (1516)
4340 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4341 - update to new LOV EA format (2097)
4342 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4344 2003-06-15 Phil Schwan <phil@clusterfs.com>
4347 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4348 - per-import/export recovery handling (958, 931, 959)
4349 - multiple last-rcvd slots, for serving multiple FSes (949)
4350 - connections are again shared between multiple imp/exports (963, 964)
4351 - "umount -f" would hang if any requests needed to be sent (393, 978)
4352 - avoid pinning large req buffer by copying for queued messages (989)
4353 - add "uuid" to "lctl device" command to help upcalls (991)
4354 - "open" RPCs with transnos would confuse recovery counters (1037)
4355 - do proper endian conversion of all wire messages (288, 340, 891)
4356 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4357 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4358 - fix ptlrpc_connection leak in target_handle_connect (1174)
4359 - fix import refcounting bug in OST and MDS cleanup (1134)
4360 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4361 - fix very unlikely obd_types race condition (501)
4362 - remove osc_open hack for echo_client (1187)
4363 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4364 - a failure in read_inode2 leads to deadlock (1139)
4365 - cancel ack-locks as soon as transaction is committed (1072)
4366 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4367 - make sure to commitrw after any preprw to avoid deadlock (1162)
4368 - failing to execute a file in a lustre FS would lock inode (1203)
4369 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4370 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4371 - obd_uuid2tgt would walk off the end of the list (1255)
4372 - on IA64 the osc would give portals incorrect bulk size (1258)
4373 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4374 - fix lock inversion caused by new llite matching code (1282)
4375 - limit the number of dirty pages on a client to 10MB (1286)
4376 - timed out locks were not being corrected cancelled (1289)
4377 - fix O_DIRECT above 4GB on IA-32 (1292)
4378 * major user-visible changes
4379 - fail out/fail over policy now controlled by the upcall (993)
4381 - add OBD_PING to check server availability and failure (954)
4382 - lustre messages are now sent in sending host order (288, 340, 891)
4383 - add eadatalen to MDS getattr reply (340)
4384 - OST read replies may contain second buffer, with per-page status (593)
4386 2003-03-11 Phil Schwan <phil@clusterfs.com>
4389 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4390 - failed open()s could cause deadlock; fixed (867, 869)
4391 - stop cancelling OST locks when files are closed (481)
4392 - overlapping XID spaces caused network corruption (851, 853)
4393 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4394 - setattr_raw added, to do single-RPC, server-side setattrs
4395 - lmc/lconf syntax change for OST UUIDs
4396 - fix crashy race condition between ptlrpc_free_req and osc_close
4397 - don't use request in mdc_enqueue if we hit a timeout (889)
4398 - don't set the inode i_size for regular files from the MDS (896)
4399 - handle out of order completion AST (842)
4400 - don't LBUG if a lock request times out after receiving AST (913)
4401 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4402 - if a bad lock AST arrives, send an error instead of dropping entirely
4403 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4404 - fix leak in bulk IO when only partially completed (899, 900, 926)
4405 - fix O_DIRECT for ia64 (55)
4406 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4407 - C-z after timeout could hang a process forever; fixed (977)
4409 - client-side I/O cache (678, 924, 929, 941, 970)
4411 - READPAGE and SETATTRs which don't take server-side locks get
4414 2003-02-11 Phil Schwan <phil@clusterfs.com>
4417 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4418 - Fix fsfilt_extN_readpage() to read a full page of directory
4419 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4420 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4421 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4422 - Most intent operations were removed, in favour of a new RPC mode
4423 that does a single RPC to the server and bypasses most of the VFS
4424 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4425 - Aggressively cancel local locks on DLM servers
4426 - mds_reint_unlink sends EA to the client if it's the last nlink.
4427 client uses that EA to unlink OST objects.
4428 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4429 - recursive symlinks were fixed (439)
4430 - fixed NULL deref in DEBUG_REQ
4431 - filter_update_lastobjid no longer calls sync, which annoyed extN
4432 - fixed multi-client small-writes to a single file problem (445)
4433 - fixed mtime updates during file writes (607)
4434 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4435 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4436 - fixed Linux OST size reporting problem (444, 656)
4437 - OST now updates object mtime with writes or setattr (607, 619)
4438 - client verifies file size before zeroing page past EOF (445)
4439 - OST now writes last allocated objid to disk with allocation (108)
4440 - LOV on echo now works (409)
4442 - mds_reint_unlink sends a new buffer, with the EA included. this
4443 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4444 set if a regular file was being unlinked, and it was the last link
4445 - use PtlGet from the target for bulk writes (315)
4446 - OST now updates object mtime with writes or setattr (607, 619)
4447 - LDLM now has a grant-time callback to revalidate locked items, if
4449 - Many MDS operations were reorganized to combat race conditions
4451 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4452 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4454 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4457 - Fully reactivate OST imports after reconnection (512, others)
4458 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4459 - More graceful error handling for truncating on dead OST (515)
4460 - Don't error out unless we're actually accessing dead stripes (474)
4461 - Fix garbage sizes when stripes are missing (410)
4462 - LRU counters were broken, causing constant lock purge (433, 432)
4463 - garbage on read from stripes with failed OSTs (441)
4464 - mark OSCs as active before reconnecting during recovery (438)
4465 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4466 - lfind did not preserve OST order in output (443)
4467 - symlinks cause hung clients, incorrect data (439)
4468 - stop dereferencing request after dropping refcount (457)
4469 - don't LASSERT(spin_is_locked) on non-SMP (455)
4470 - fixes for many rename() bugs
4471 - fstat didn't correctly synchronize attributes (399)
4472 - server must handle lock cancellation during blocking AST prep (487)
4473 - bulk descriptors were free()d too soon (511)
4474 - fix paths in lconf, which would load incorrect modules (451, 507)
4475 - fix confusing lconf 'host not found' error message (386)
4476 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4477 - fix race condition in mdc_blocking_ast() for inode access (526)
4478 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4479 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4480 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4481 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4482 - fix lstripe in several ways (kernel side) (527)
4483 - fix request leak in ldlm_cli_enqueue (262)
4484 - incorrect OSC was marked inactive after OST failure
4485 - call mds_fs_cleanup before unmounting filesystem (524)
4486 - fix races between taking ns_lock and ldlm_lock_change_resource
4487 - fix races updating LOV export open file list
4488 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4489 - fix recovery NULL deref in ldlm_cli_cancel_unused
4490 - fix some DLM races by using new hash table for lock handles (419)
4491 - permit the client to specify desired inodes, at replay
4492 - duplicate requests when we queue them for replay reintegration
4493 - fix last_rcvd offset calculation
4494 - sync after each recovered transaction, so we always make progress
4495 - never, not always, ERESTART requests without transnos
4496 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4498 - skip replay if the MDS says that the client is already connected
4499 - don't check for a recovery-enabled export to match lctl's UUID
4500 - don't INC_USE_COUNT for phantom exports
4501 - don't crash when cleaning up phantom exports (567)
4502 - don't double-finish or set replay data for errored mdc_open requests
4503 - abort requests when they time out, so we don't get old replies
4504 - send/receive replies for AST messages again
4505 - if the client says that it doesn't have the lock, cancel it on the
4507 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4508 mark it as destroyed, it will all work out when decref is called
4509 - fix module use counts (22, 581)
4511 - ASTs now expect a reply (server cancels lock on error reply)
4513 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4516 - fix many simultaneous client startup (392)
4517 - fix dentry->d_it clobbering
4518 - credentials weren't being shipped for readdir/getattr operations
4519 - remove invalid assertions triggered during some concurrent MD
4521 - proper Lustre versions added (336, 389)
4522 - fix memory leak for create error case (398)
4523 - fix LOV locking bug that would get cli/srv out of sync
4524 - fix echo client over LOV (409)
4525 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4526 - fix double-O_EXCL intent crash (424)
4527 - avoid sending multiple lock CANCELs (352)
4529 - MDS can do multi-client recovery (modulo bugs in new code)
4531 - many updates, edits, cleanups
4533 2002-11-18 Phil Schwan <phil@clusterfs.com>
4536 - fix null d_it dereference (346)
4537 - fix full OST/dbench hang (333)
4538 - fix permission problem with file removal (286)
4539 - fix removal of OSCs from LOV when they fail
4540 - fix NULL deref during bulk timeout (214)
4541 - fix problems related to multiple filesystems on one MDS (241)
4542 - fixed serious subtle metadata locking bugs
4543 - free locks on clients when inodes are removed due to memory
4545 - fix inode pointer in lock data (285)
4546 - partial support for multiple MDS on a single host (241)
4547 - data locks weren't cancelled at clear_inode time (290, 311)
4548 - intent locks could lead to unbounded lock growth (205)
4549 - added a maximum lock count, an LRU list, and a flusher
4550 - fix multiple rename (365)
4551 - properly abstracted the echo client
4552 - OSC locked 1 byte too many; fixed
4553 - rewrote brw callback code:
4554 - fixed recovery bugs related to LOVs (306)
4555 - fixed too-many-pages-in-one-write crash (191)
4556 - fixed (again) crash in sync_io_timeout (214)
4557 - probably fixed callback-related race (385)
4559 - Add capability to MDS protocol
4560 - LDLM cancellations and callbacks on different portals
4562 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4565 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4566 - defer open object destruction to close time (601981, 138)
4567 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4568 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4569 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4570 - don't call obd_setattr in ll_file_release for destroyed objects
4571 * protocol change to lustre_msg: move |version| and add |flags|
4572 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4573 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4574 * added replay of create, unlink, link and rename operations during
4575 MDS failover; recovery should be much more robust now
4576 * remove failed OSCs from LOVs (only lov_create uses this so far)
4577 * the lustre-HOWTO was brought (more) up to date (582544)
4579 2002-10-23 Phil Schwan <phil@clusterfs.com>
4582 - in-use dentries weren't being reused properly (617851)
4583 - prevent multiple LDLM setup (599178)
4584 - fix LOV size calculations for truncate (617853)
4585 - fix client handling of MDS intent errors (POSIX)
4586 - fix permission bug in lovstripe.c test (624321)
4587 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4588 - truncate past end of file could corrupt data
4589 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4590 - a race in recovery could return ETIMEDOUT to apps (623947)
4591 - building outside the source directory was fixed
4592 * the lustre-HOWTO was brought (more) up to date (582544)
4593 * major progress was made on recovery functionality
4595 2002-10-10 Phil Schwan <phil@clusterfs.com>
4598 - recovery deadlock fix
4599 - rm -rf causes LBUG fix (617817)
4600 - file open by multiple tasks fix (618962)
4601 - directory permissions bugs (602707 and 620007)
4602 - journal_stop fixed with locking (611313)
4603 - O_APPEND failures resolved (618273, perhaps 614459)
4604 - lconf PATH fix (619770)
4605 - IA64 build fix (621450)
4606 - RPC buffer sizes scale with amount of memory
4608 2002-10-01 Phil Schwan <phil@clusterfs.com>
4611 - locks would be cancelled without throwing away data pages,
4612 resulting in inconsistent data (605627)
4613 - inode attributes were not always being refreshed (605627, 612449)
4614 - lconf now continues to cleanup after lctl reports an error
4615 - MDS now enforces user permissions (602707)
4616 - lprocfs cleanup fixed, but not yet enabled (614157)
4617 - fixed infinite server hang, should a client not respond to an AST
4618 - avoid going into recovery if user calls readlink() with a buffer
4619 that's too small (613941)
4620 - AST RPCs no longer require replies (614867) -- this may be changed
4621 - don't crash server if client sends an IOV that's too big (611336)
4622 - fixed lock conversion deadlock (611892)
4623 - fixed the following of symlinks (614622)
4624 * recovery: the server can remove locks from a client that dies, other
4625 clients can make progress
4626 * more extN patch fixes
4627 * compile-time configurable ptlrpc buffer allocations
4629 - collaborative read cache document
4630 - Lustre Lite Performance CDR document-in-progress
4632 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4635 - fix typo in patch-2.4.18
4637 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4640 - clear ptlrpc request each time in handle_incoming_request()
4641 - unlink of files now destroys the object on the OST
4643 2002-09-19 Peter Braam <braam@clusterfs.com>
4645 * add hard link support
4646 * change obdfile creation method
4647 * kernel patch changed
4649 2002-09-19 Peter Braam <braam@clusterfs.com>
4652 - stack overflow bug in extN fixed
4654 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4656 * documentation updates
4657 - add man pages for config tools
4658 - update tests/README to describe testing with new config tools
4659 - finish metadata API descriptions
4660 * bug fixes and cleanups
4661 - statfs workaround for 16TB limit
4662 - LOV stripe allocation improved, can stripe on subset of OSTs
4663 - LOV file size/IO offset was wrong for files > 4GB in size
4664 - object EA data was being dropped, caused files to be unreadable
4665 - memory overflow with non-LOV OST caused memory corruption
4666 - fixed regression tests to work with new config tools, obdfilter
4667 - fixed bug when directory size became larger than 1 block
4668 - fixed bug (for single client case) when PWD was deleted
4669 - invalidate local directory pages when doing intent-based ops
4670 - avoid LDLM oops when lock callback contained bad data
4672 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4674 * documentation updates
4675 * bug fixes and cleanups
4676 - configuration tools
4679 - 64-bit compile warnings
4680 - 64-bit internal statfs data
4682 * test_brw on persistent OST devices
4684 * lprocfs (disabled)
4686 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4688 * documentation updates
4689 * bug fixes and cleanups
4690 * configuration tools
4692 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4695 * many small fixes to 0_5_4
4696 * io/network handling
4697 * thinkos in MDS operations
4699 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4702 * crucial basic fixes to 0.5.3
4703 * IOR, Iozone work over Elan
4706 2002-08-07 Phil Schwan <phil@clusterfs.com>
4707 * version 0_5_3, our first alpha
4708 * we use the new Portals iovs
4709 * documentation updates
4710 * bug fixes and cleanups
4711 * small changes in the DLM wire protocol
4713 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4714 * version 0_5_1 with some initial stability,
4715 * locking on MD and file I/O.
4716 * documentation updates
4717 * several bug fixes since 0.5.0
4718 * small changes in wire protocol
4720 2002-07-18 Phil Schwan <phil@clusterfs.com>
4722 * delivered as Lustre Light Alpha
4723 * fixed a crash after handling invalid MDS requests
4724 * fixed directory pages for architectures with non-4k pages sizes
4726 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4727 * release version v0_4_4
4728 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4730 * Fixes a number of interruption problems with OST operations.
4731 * Update documentation for portals header changes
4732 * Move all wire protocol structs/defines to lustre_idl.h
4733 * Fixes symlink length bug.
4734 * Add tcpdump to repository.
4736 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4737 * release version v0_4_3
4738 * Fixes statfs for inodes on extN.
4739 * Fixes bug in runtests which would delete /etc/hosts.
4740 * Use 64-bit object IDs wherever possible (not into VFS though)
4741 Remove ost_get_info, which is unused by lustre, and out of date.
4743 2002-07-03 Peter Braam <braam@clusterfs.com>
4744 * release version v0_4_2 Fixes a lookup error (type not passed)
4745 * move forward to head of Portals
4746 * move forward to latest Lustre kernel
4748 2002-06-25 Peter Braam <braam@clusterfs.com>
4749 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob