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: mds_mfd_close() ASSERTION(rc == 0)
18 Details : In mds_mfd_close(), we need protect inode's writecount change
19 within its orphan write semaphore to prevent possible races.
23 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
27 Frequency : rare, on shutdown ost
28 Description: don't hit live lock with umount ost.
29 Details : shrink_dcache_parent can be in long loop with destroy dentries,
30 use shrink_dcache_sb instead.
34 Description: don't panic with use echo client
35 Details : echo client pass NULL as client nid pointer and this produce null
40 Description: fix build on ppc32
41 Details : compile code with -m64 flag produce wrong object file for ppc32.
45 Description: add message levels for liblustreapi
49 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
50 Details : Move the 'good_osts' check before the 'total_bavail' check. This
51 will result in an -EAGAIN and in the exit call path we call
52 alloc_rr() which will with increasing aggressiveness attempt to
53 aquire precreated objects on the minimum number of required OSCs.
57 Description: Use old size assignment to avoid deadlock
58 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
59 the scheduling while holding a spinlock. We do not need locking
60 for size in ll_update_inode() because size is only updated from
61 the MDS for directories or files without objects, so there is no
62 other place to do the update, and concurrent access to such inodes
63 are protected by the inode lock.
67 Description: resolve "_IOWR redefined" build error on SLES10
71 Description: dump the memory debugging after all modules are unloaded to
72 suppress false negative in conf_sanity test 39
74 Severity : enhancement
76 Description: build kernel-ib packages for OFED 1.3 in our release cycle
81 Description: fix SLES kernel versioning
82 Details : the kernel version for our SLES 10 kernel did not include a "-"
83 before the "smp" at the end. while this was not a problem in
84 general, it did mean that software trying to use the kernel
85 version to try to detect a vendor specific kernel would fail.
86 this was most evident by the OFED build scripts.
90 Description: Don't update lov_desc members until making sure they are valid
91 Details : When updating lov_desc members via proc fs, need fix their
92 validities before doing the real update.
96 Description: don't put request into delay list while invalidate in flight.
97 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
98 import in flight. this produce timeout for invalidate and sometimes
103 Frequency : on ppc only
104 Description: not convert ost objects for directory because it's not exist.
105 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
109 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
111 Severity : enhancement
113 Description: Update to support 2.6.22.14 vanilla kernel.
117 Frequency : rare, on recovery
118 Description: read procfs can produce deadlock in some situation
119 Details : Holding lprocfs lock which send rpc can produce block for destroy
120 obd objects and this also block reconnect with -EALREADY. This isn't
121 fix all lprocfs bugs - but make it rare.
123 Severity : enhancement
125 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
128 Frequency : frequent on X2 node
130 Description: mdc_set_open_replay_data LBUG
131 Details : Set replay data for requests that are eligible for replay.
135 Description: lustre_mgs: operation 101 on unconnected MGS
136 Details : When MGC is disconnected from MGS long enough, MGS will evict the
137 MGC, and late on MGC cannot successfully connect to MGS and a lot
138 of the error messages complaining that MGS is not connected.
141 Frequency : on start mds
143 Description: Implement get_info(last_id) in obdfilter.
146 Frequency : occasional
148 Description: Correctly check stale fid, not start epoch if ost not support SOM
149 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
150 overwrited with new generated fid, but mds can anwer with one of these
151 two fids and both is not stale. setattr incorectly start epoch and
152 assume will be called done_writeting, but without SOM done_writing
156 Frequency : rare, depends on device drivers and load
158 Description: MDS or OSS nodes crash due to stack overflow
159 Details : Code changes in 1.8.0 increased the stack usage of some functions.
160 In some cases, in conjunction with device drivers that use a lot
161 of stack the MDS (or possibly OSS) service threads could overflow
162 the stack. One change which was identified to consume additional
163 stack has been reworked to avoid the extra stack usage.
166 Frequency : occasional
168 Description: Do not fail import if osc_interpret_create gets -EAGAIN
169 Details : If osc_interpret_create got -EAGAIN it immediately exits and
170 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
171 oscc_has_objects and see OSC has no objests and call
172 oscc_internal_create to resend create request.
174 Severity : enhancement
176 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
178 Severity : enhancement
180 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
183 Frequency : very rare
185 Description: Fix replay if there is an un-replied request and open
186 Details : In some cases, older replay request will revert the
187 mcd->mcd_last_xid on MDS which is used to record the client's
190 Severity : enhancement
192 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
194 Severity : enhancement
196 Description: Add rhel5 support to HEAD.
198 Serverity : enhancement
200 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
205 Description: Don't allow skipping OSTs if index has been specified.
206 Details : Don't allow skipping OSTs if index has been specified, make locking
207 in internal create lots better.
211 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
212 Details : access to bitfield in structure is always rounded to long
213 and this produce problem with not atomic change any bit.
217 Description: Lustre make rpms failed.
218 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
219 builds Lustre rpms from tarball.
222 Frequency : rare on shutdown ost
224 Description: If llog cancel was not send before clean_exports phase, this can
225 produce deadlock in llog code.
226 Details : If llog thread has last reference to obd and call class_import_put
227 this produce deadlock because llog_cleanup_commit_master wait when
228 last llog_commit_thread exited, but this never success because was
229 called from llog_commit_thread.
233 Description: allow userland application know is lost one of stripes.
234 Details : fill lvb_blocks with error code on ost and return it to
235 application if error flag found.
239 Description: NULL lov_tgts causing MDS oops
240 Details : more safe checks for NULL lov_tgts for avoid oops.
242 Severity : enhancement
244 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
248 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
249 Details : If an OST is down the MDS will hang indefinitely in
250 obd_statfs_rqset() waiting for the statfs data. While for
251 MDS QOS usage of statfs, it should not stuck in waiting.
253 Severity : enhancement
255 Description: remote_acl support
256 Details : Support ACL-based permission check for remote user.
257 Support setfacl/getfacl for remote user with the utils
258 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
259 the system "{s,g}etfacl" utils.
261 Severity : enhancement
263 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
265 Severity : enhancement
267 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
271 Description: make mgs_setparam() handle fsname containing dash
272 Details : fsname containing a dash does not work with lctl conf_param
274 Severity : enhancement
276 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
278 Severity : enhancement
280 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
284 Frequency : rare, at shutdown
285 Description: access already free / zero obd_namespace.
286 Details : if client_disconnect_export was called without force flag set,
287 and exist connect request in flight, this can produce access to
288 NULL pointer (or already free pointer) when connect_interpret
289 store ocd flags in obd_namespace.
293 Frequency : only at startup
294 Description: not alloc memory with spinlock held.
295 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
296 if any spinlock held.
298 Severity : enhancement
300 Description: make lustre randomly fail allocating memory
301 Details : Make lustre randomly failed allocating memory for testing purpose.
303 Severity : enhancement
305 Description: lost problems with lov objid file
306 Details : Fixes some scability and access to not inited memory problems
307 in work with lov objdid file.
312 Description: lfs find does not continue on file error
313 Details : Continue other files processing when a file/dir is absent.
317 Description: Inconsistent usage of lustre_pack_reply()
318 Details : Standardize the usage of lustre_pack_reply() such that it
319 always generate a CERROR on failure.
322 Frequency : occasional
324 Description: MDS hang or stay in waiting lock
325 Details : If client receive lock with CBPENDING flag ldlm need send lock
326 cancel as separate rpc, to avoid situation when cancel request
327 can't processed due all i/o threads stay in wait lock.
331 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
335 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
339 Description: Update to SLES9 kernel-2.6.5-7.287.3.
343 Description: Update to RHEL4 latest kernel.
345 Severity : enhancement
347 Description: Build SLES10 patchless client fails
348 Details : The configure was broken by run ./configure with
349 --with-linux-obj=.... argument for patchless client. When the
350 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
351 can't search header adequately. Use absolute path such as
352 -I($LINUX)/include instead.
356 Description: interrupt oig_wait produce painc on resend.
357 Details : brw_redo_request can be used for resend requests from ptlrpcd and
358 private set, and this produce situation when rq_ptlrpcd_data not
359 copyed to new allocated request and triggered LBUG on assert
360 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
361 ptlrpcd set if request is changed and can be safety changed to use
366 Description: LASSERT_{REQ,REP}SWAB macros are buggy
367 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
368 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
369 these macros and replace them with their difinitions instead.
373 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
374 Details : Update which_patch & target file for SLES10 latest kernel.
377 Description: Remove client patches from SLES 10 kernel.
378 Details : This causes SLES 10 clients to behave as patchless clients
379 even on a Lustre-patched (server) kernel.
381 Severity : enhancement
383 Description: self-adjustable client's lru lists
384 Details : use adaptive algorithm for managing client cached locks lru
385 lists according to current server load, other client's work
386 pattern, memory activities, etc. Both, server and client
387 side namespaces provide number of proc tunables for controlling
390 Severity : enhancement
392 Description: light-weight GSS support
393 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
394 and reduce performance overhead.
396 Severity : enhancement
398 Description: Linux keyring support
399 Details : Support using service of Linux keyring for Lustre GSS internal
400 context refresh/cache mechanism.
404 Description: Fix errors in lfs documentation
405 Details : Fixes man pages
409 Description: don't use GFP_* in generic Lustre code.
410 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
414 Description: obdclass is limited by single OBD_ALLOC(idarray)
415 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
419 Description: warnings with build patchless client with vanila 2.6.19 and up
420 Details : change old ctl_table style and replace ctl_table/ctl_table_header
421 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
425 Frequency : Only for SLES
426 Description: sanity_quota fail test_1
427 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
428 file, which makes TSTID[2] inunique.
432 Frequency : Only for SLES9
433 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
434 Details : Update target/ChangeLog/which_patch .
438 Frequency : Only for test_10 in sanity-quota.sh
439 Description: a bug in quota test code
440 Details : checking a test flag in a wrong place
444 Frequency : Only for test_13 in sanity-quota.sh
445 Description: a bug in quota test script
446 Details : shouldn't check overfull allocated quota
450 Description: jbd statistics
451 Details : Port older jbd statistics patch for sles10
453 Severity : enhancement
455 Description: Add printing inode info into message about error in writepage.
460 Description: reply_lock_interpret crash due to race with it and lock cancel.
461 Details : Do not replay locks that are being cancelled. Do not reference
462 locks by their address during replay, just by their handle.
466 Frequency : When flocks are used.
467 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
468 Details : Flock locks might destroy just granted lock if it could be merged
469 with another existing flock, this is done in completion handler,
470 so teach ldlm_cli_enquque_fini that this is a valid case for
476 Description: Oops in read and write path when failing to allocate lock.
477 Details : Check if lock allocation failed and return error back.
481 Description: lstripe command fails for valid OST index
482 Details : The stripe offset is compared to lov->desc.ld_tgt_count
483 instead of lov->desc.ld_active_tgt_count.
487 Description: sanity.sh failed test 103
488 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
493 Description: df doesn't work properly if diskfs blocksize != 4K
494 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
498 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
499 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
503 Description: obd_health_check_timeout too short
504 Details : set obd_health_check_timeout as 1.5x of obd_timeout
508 Description: llapi_file_create() does not allow some changes
509 Details : add llapi_file_open() that allows specifying the mode and
510 open flags, and also returns an open file handle.
514 Description: merge and cleanup kernel patches.
515 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
519 Description: Add journal checksum support.(Kernel part)
520 Details : The journal checksum feature adds two new flags i.e
521 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
522 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
523 indicates that the commit block contains the checksum for
524 the blocks described by the descriptor blocks. Now commit
525 record can be sent to disk without waiting for descriptor
526 blocks to be written to disk. This behavior is controlled
527 using JBD2_FEATURE_ASYNC_COMMIT flag.
531 Description: OSS needs mutliple precreate threads
532 Details : Add ability to start more than one create thread per OSS.
536 Description: Sanity.sh test_65k failed: setstripe should have succeeded
537 Details : execute lfs setstripe on client
541 Description: mds_obd_create error creating tmp object
542 Details : When the user sets quota on root, llog will be affected and can't
543 create files and write files.
546 Frequency : Always on ia64 patchless client, and possibly others.
548 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
549 Details : This allows the patchless client to be loaded on architectures
554 Description: RedHat Update kernel for RHEL5
555 Details : Add the kernel config file for RHEL5.
559 Description: RedHat Update kernel for RHEL5
560 Details : Modify the kernel config file more closer RHEL5.
564 Description: Build failure against Centos5 (RHEL5)
565 Details : Define PAGE_SIZE when it isn't present.
569 Description: test_7 on sanity-quota.sh can't be used on separate machines
570 Details : add facet to handle it
574 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
575 Details : with NFS, the anon dentry's parent was set to itself in
576 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
577 corresponding dentry other than use rec->ur_name.
579 Severity : enhancement
581 Description: lfs setstripe enhancement
582 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
586 Description: enable data checksumming by default
587 Details : enable checksum by default, allow --disable-checksum
588 configure option and "-o nochecksum" mount option. Checksums
589 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
593 Description: lustre support for RHEL5
594 Details : Add support for RHEL5.
598 Description: Client eviction due to failover config
599 Details : after a connection loss, the lustre client should attempt to
600 reconnect to the last active server first before trying the
601 other potential connections.
605 Description: when mds and osts use different quota unit(32bit and 64bit),
606 quota will be released repeatly.
607 Details : void sending multiple quota reqs to mds, which will keep the
608 status between the reqs.
612 Description: osts not allocated evenly to files
613 Details : change the condition to increase offset_idx
616 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
618 Description: Data corruption for OSTs that are formatted larger than 2TB
620 Details : When generating the bio request for lustre file writes the
621 sector number would overflow a temporary variable before being
622 used for the IO. The data reads correctly from Lustre (which
623 will overflow in a similar manner) but other file data or
624 filesystem metadata may be corrupted in some cases.
628 Description: Tune the kernel for good SCSI performance.
629 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
630 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
635 Description: rewrite ext2-derived code in obdclass/uuid.c
636 Details : rewrite inherited code (uuid parsing code from ext2 utils)
637 from scratch preserving functionality.
641 Description: "lfs find -obd UUID" prints directories
642 Details : "lfs find -obd UUID" will return all directory names instead
643 of just file names. It is incorrect because the directories
644 do not reside on the OSTs.
648 Description: use i_size_read and i_size_write in 2.6 port
649 Details : replace inode->i_size access with i_size_read/write()
652 Frequency : when removing large files
654 Description: scheduling issue during removal of large Lustre files
655 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
656 It causes scheduling issues when removing large files (17TB in the
660 Frequency : only with liblustre clients on XT3
662 Description: evictions taking too long
663 Details : allow llrd to evict clients directly on OSTs
666 Frequency : only on ppc
668 Description: /proc/fs/lustre/devices broken on ppc
669 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
670 We should make sure the type of 'index' is loff_t and then cast
671 explicitly as needed below. Do not assign an explicitly cast
675 Frequency : only for rhel5
677 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
678 Details : Modify the target file & which_kernel.
680 Severity : enhancement
682 Description: omit set fsid for export NFS
683 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
684 as FS_REQUIRES_DEV for avoid problems with generate fsid.
688 Frequency : Always, for kernels after 2.6.16
689 Description: Fix warning idr_remove called for id=.. which is not allocated.
690 Details : Last kernels save old s_dev before kill super and not allow
691 to restore from callback - restore it before call kill_anon_super.
695 Description: buffer overruns could theoretically occur
696 Details : llapi_semantic_traverse() modifies the "path" argument by
697 appending values to the end of the origin string, and a
698 overrun may occur. Adding buffer overrun check in liblustreapi.
702 Description: Fix error on 'ls .' at the top of the Lustre mount.
703 Details : Don't revalidate dentry if it is a root dentry.
707 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
708 Details : Modify vm-tunables-rhel4.patch.
712 Description: Kernel config for 2.6.18-vanilla.
713 Details : Modify targets/2.6-vanilla.target.in.
714 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
715 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
716 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
717 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
721 Description: improve handling recoverable errors
722 Details : if request processig with error which can be recoverable on server
723 request should be resend, otherwise page released from cache and
728 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
729 Details : Modify target file & which_patch.
730 A flaw was found in the IA32 system call emulation provided
731 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
732 value could be stored in the %RAX register, which could trigger an
733 out-of-bounds system call table access. An untrusted local user
734 could exploit this flaw to run code in the kernel
735 (ie a root privilege escalation). (CVE-2007-4573).
739 Description: O_DIRECT bypasses client statistics.
740 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
741 read_ahead_stats were updated. copied the stats section from
742 osc_send_oap_rpc() into async_internal().
746 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
750 Description: change order of libsysio includes
751 Details : '#include sysio.h' should always come before '#include xtio.h'
755 Description: handle "serious error: objid * already exists" more gracefully
756 Details : If LAST_ID value on disk is smaller than the objects existing in
757 the O/0/d* directories, it indicates disk corruption and causes an
758 LBUG(). If the object is 0-length, then we should use the existing
759 object. This will help to avoid a full fsck in most cases.
761 Severity : enhancement
763 Description: adapt the lustre_config script to support the upgrade case
764 Details : Add "-u" option for lustre_config script to support upgrading 1.4
765 server targets to 1.6 in parallel.
769 Description: To avoid grant space > avaible space when the disk is almost
770 full. Without this patch you might see the error "grant XXXX >
771 available" or some LBUG about grant, when the disk is almost
773 Details : In filter_check_grant, for non_grant cache write, we should
774 check the left space by if (*left > ungranted + bytes), instead
775 of (*left > ungranted), because only we are sure the left space
776 is enough for another "bytes", then the ungrant space should be
777 increase. In client, we should update cl_avail_grant only there
778 is OBD_MD_FLGRANT in the reply.
782 Description: Update RHEL 4 kernel to fix local root privilege escalation.
783 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
784 described in CVE-2007-4573. This problem could allow untrusted
785 local users to gain root access.
788 Frequency : when using O_DIRECT and quotas
790 Description: Incorrect file ownership on O_DIRECT output files
791 Details : block usage reported by 'lfs quota' does not take into account
792 files that have been written with O_DIRECT.
797 Description: touch file failed when fs is not full
798 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
799 otherwise we can get ENOSP while fs is not full.
803 Description: parallel lock callbacks
804 Details : Instead of sending blocking and completion callbacks as separated
805 requests, adding them to a set and sending in parallel.
808 Frequency : only for Cray XT3
809 Bugzilla : 12829/13455
810 Description: Changing primary group doesn't change the group lustre assigns to
812 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
813 overridden with the primary group provided by the group upcall,
814 whereas the supplied fsgid can be trusted if it is in the list of
815 supplementary groups returned by the group upcall.
817 Severity : enhancement
819 Description: Allow masking D_WARNING, D_ERROR messages from console
820 Details : Console messages can now be disabled via lnet.printk.
824 Description: User code with malformed file open parameter crashes client node
825 Details : Before packing join_file req, all the related reference should be
826 checked carefully in case some malformed flags cause fake
827 join_file req on client.
831 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
832 Details : ldlm_completion_ast() assumes that a lock is granted when the req
833 mode is equal to the granted mode. However, it should also check
834 that LDLM_FL_CP_REQD is not set.
838 Description: Heavy nfs access might result in deadlocks
839 Details : After ELC code landed, it is now improper to enqueue any mds
840 locks under och_sem, because enqueue might want to decide to
841 cancel open locks for same inode we are holding och_sem for.
845 Description: Client eviction while running blogbench
846 Details : A lot of unlink operations with concurrent I/O can lead to a
847 deadlock causing evictions. To address the problem, the number of
848 oustanding OST_DESTROY requests is now throttled to
849 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
850 callbacks are processed in priority.
854 Description: enable ACLs on MDS by default
855 Details : ACLs must be enabled on MDS by default.
858 Frequency : PPC/PPC64 only
860 Description: conflicts between asm-ppc64/types.h and lustre_types.h
861 Details : fix duplicated definitions between asm-ppc64/types.h and
862 lustre_types.h on PPC.
865 Frequency : PPC/PPC64 only
867 Description: asm-ppc/segment.h does not exist
868 Details : fix compile issue on PPC.
872 Description: better handle error messages in extents code
875 Frequency : RHEL4 only
877 Description: mkfs is very slow on IA64/RHEL4
878 Details : A performance regression has been discovered in the MPT Fusion
879 driver between versions 3.02.73rh and 3.02.99.00rh. As a
880 consequence, we have downgraded the MPT Fusion driver in the RHEL4
881 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
883 Severity : enhancement
885 Description: SNMP support enhancement
886 Details : Adding total number of sampled request for an MDS node in snmp
889 Severity : enhancement
891 Description: Optimize ldlm waiting list processing for PR extent locks
892 Details : When processing waiting list for read extent lock and meeting read
893 lock that is same or wider to it that is not contended, skip
894 processing rest of the list and immediatelly return current
895 status of conflictness, since we are guaranteed there are no
896 conflicting locks in the rest of the list.
900 Description: Time out and refuse to reconnect
901 Details : When the failover node is the primary node, it is possible
902 to have two identical connections in imp_conn_list. We must
903 compare not conn's pointers but NIDs, otherwise we can defeat
904 connection throttling.
908 Description: port llog fixes from b1_6 into HEAD
909 Details : Port llog reference couting and some llog cleanups from b1_6
910 (bug 10800) into HEAD, for protect from panic and access to already
911 free llog structures.
915 Description: Detect stride IO mode in read-ahead
916 Details : When a client does stride read, read-ahead should detect that and
917 read-ahead pages according to the detected stride pattern.
921 Description: data checksumming impacts single node performance
922 Details : add support for several checksum algorithm. Currently, only CRC32
923 and Adler-32 are supported. The checksum type can be changed on
924 the fly via /proc/fs/lustre/osc/*/checksum_type.
928 Description: use adler32 for page checksums
929 Details : when available, use the Adler-32 algorithm instead of CRC32 for
934 Description: build for x2 fails
935 Details : fix compile issue on Cray systems.
939 Description: Properly match for duplicate locks
940 Details : Due to different lock order from skiplists code, we need to
941 traverse entire list for now
944 Frequency : only on PPC/SLES10
946 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
947 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
948 be defined. Add a hack in mkfs_lustre.c to work around this compile
953 Description: LASSERT on MDS when client holding flock lock dies
954 Details : ldlm pool logic depends on number of granted locks equal to
955 number of released locks which is not true for flock locks, so
956 just exclude such locks from consideration.
960 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
961 Details : Use fsfilt_sync() for both the whole filesystem sync and
962 individual file sync to eliminate dangerous inode locking
963 with I_LOCK that can lead to a deadlock.
967 Description: performance in 1.6.3
968 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
969 sure that 1MB requests are not fragmented by the block layer.
971 Severity : enhancement
973 Description: organize the server-side client stats on per-nid basis
974 Details : Change the structure of stats under obdfilter and mds to
982 The "uuid"s file would list the uuids of _active_ exports.
983 And the clear entry is to clear all stats and stale nids.
985 Severity : enhancement
987 Description: eliminate client locks in face of contention
988 Details : file contention detection and lockless i/o implementation
993 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
995 --------------------------------------------------------------------------------
997 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
999 * Support for kernels:
1000 2.6.5-7.283 (SLES 9),
1001 2.6.9-55.EL (RHEL 4),
1002 2.6.16.46-0.14 (SLES 10),
1003 2.6.18.8 vanilla (kernel.org)
1004 * Client support for unpatched kernels:
1005 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1006 2.6.16 - 2.6.22 vanilla (kernel.org)
1007 * Due to recently discovered recovery problems, we do not recommend
1008 using patchless RHEL 4 clients with this or any earlier release.
1009 * Recommended e2fsprogs version: 1.39.cfs8
1010 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1011 * Starting with this release, the ldiskfs backing filesystem required
1012 by Lustre is now in its own package, lustre-ldiskfs. This package
1013 should be installed. It is versioned separately from Lustre and
1014 may be released separately in future.
1018 Description: block reactivating mgc import until all deactivates complete
1019 Details : Fix race when failing back MDT/MGS to itself (testing)
1022 Frequency : at statup only
1024 Description: mds_lov_synchronize race leads to various problems
1025 Details : simultaneous MDT->OST connections at startup can cause the
1026 sync to abort, leaving the OSC in a bad state.
1028 Severity : enhancement
1030 Description: add optional extra BUILD_VERSION info
1031 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1032 to override the lustre version.
1035 Frequency : 2.6.18 servers only
1037 Description: ll_kern_mount() doesn't release the module reference
1038 Details : The ldiskfs module reference count never drops down to 0
1039 because ll_kern_mount() doesn't release the module reference.
1044 Description: server LBUG when using old ost_num_threads parameter
1045 Details : Accept the old ost_num_threads parameter but warn that it
1046 is deprecated, and fix an off-by-one error that caused an LBUG.
1051 Description: Transient SCSI error results in persistent IO issue
1052 Details : iobuf->dr_error is not reinitialized to 0 between two
1056 Frequency : sometimes when underlying device returns I/O errors
1058 Description: OSTs not going read-only during write failures
1059 Details : OSTs are not remounted read-only when the journal commit threads
1060 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1061 instead of the ext3 wrappers.
1065 Description: poor connect scaling with increasing client count
1066 Details : Don't run filter_grant_sanity_check for more than 100 exports
1067 to improve scaling for large numbers of clients.
1070 Frequency : SLES10 only
1072 Description: sanity-quota.sh quotacheck failed: rc = -22
1073 Details : Quotas cannot be enabled on SLES10.
1076 Frequency : liblustre clients only
1078 Description: getdirentries does not give error when run on compute nodes
1079 Details : getdirentries does not fail when the size specified as an argument
1080 is too small to contain at least one entry
1082 Severity : enhancement
1084 Description: Add LNET router traceability for debug purposes
1085 Details : If a checksum failure occurs with a router as part of the
1086 IO path, the NID of the last router that forwarded the bulk data
1087 is printed so it can be identified.
1092 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1093 Details : Due to a race condition, liblustre clients were occasionally
1094 evicted incorrectly.
1096 Severity : enhancement
1098 Description: lfs setstripe use optional parameters instead of postional
1101 Severity : enhancement
1103 Description: Nanosecond timestamp support for ldiskfs
1104 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1105 resolution timestamps. There is not yet support for this at
1106 the Lustre filesystem level.
1109 Frequency : during server recovery
1111 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1112 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1116 Frequency : nfs export on patchless client
1118 Description: connectathon hang when test nfs export over patchless client
1119 Details : Disconnected dentry cannot be found with lookup, so we do not need
1120 to unhash it or make it invalid
1123 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1124 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1125 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1130 Description: open req refcounting wrong on reconnect
1131 Details : If reconnect happened between getting open reply from server and
1132 call to mdc_set_replay_data in ll_file_open, we will schedule
1133 replay for unreferenced request that we are about to free.
1134 Subsequent close will crash in variety of ways.
1135 Check that request is still eligible for replay in
1136 mdc_set_replay_data().
1141 Description: disable writes to filesystem when reading health_check file
1142 Details : the default for reading the health_check proc file has changed
1143 to NOT do a journal transaction and write to disk, because this
1144 can cause reads of the /proc file to hang and block HA state
1145 checking on a healthy but otherwise heavily loaded system. It
1146 is possible to return to the previous behaviour during configure
1147 with --enable-health-write.
1149 Severity : enhancement
1151 Description: 64-bit inode version
1152 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1153 to the inode. This will be required for version-based recovery.
1158 Description: MDS fails to start if a duplicate client export is detected
1159 Details : in some rare cases it was possible for a client to connect to
1160 an MDS multiple times. Upon recovery the MDS would detect this
1161 and fail during startup. Handle this more gracefully.
1163 Severity : enhancement
1165 Description: Add -o localflock option to simulate old noflock
1167 Details : This will achieve local-only flock/fcntl locks
1173 Description: log_commit_thread vs filter_destroy race leads to crash
1174 Details : Take import reference before releasing llog record semaphore
1179 Description: Wrong request locking in request set processing
1180 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1181 imp_delayed_list, in this place should be used imp_lock.
1184 Frequency : when reconnection
1186 Description: Grant Leak when osc reconnect to OST
1187 Details : When osc reconnect ost, OST(filter) should check whether it
1188 should grant more space to client by comparing fed_grant and
1189 cl_avail_grant, and return the granted space to client instead
1190 of "new granted" space, because client will call osc_init_grant
1191 to update the client grant space info.
1194 Frequency : when client reconnect to OST
1196 Description: Grant Leak when osc do resend and replay bulk write
1197 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1198 bulk write request, because the grant info while be sync between
1199 OSC and OST when reconnect, and we should ignore the grant info
1200 these of resend/replay write req.
1205 Description: Grant space more than avaiable left space sometimes.
1206 Details : When then OST is about to be full, if two bulk writing from
1207 different clients came to OST. Accord the avaliable space of the
1208 OST, the first req should be permitted, and the second one
1209 should be denied by ENOSPC. But if the seconde arrived before
1210 the first one is commited. The OST might wrongly permit second
1211 writing, which will cause grant space > avaiable space.
1214 Frequency : when client is evicted
1216 Description: Grant might be wrongly erased when osc is evicted by OST
1217 Details : when the import is evicted by server, it will fork another
1218 thread ptlrpc_invalidate_import_thread to invalidate the
1219 import, where the grant will be set to 0. While the original
1220 thread will update the grant it got when connecting. So if
1221 the former happened latter, the grant will be wrongly errased
1222 because of this race.
1227 Description: Checking Stale with correct fid
1228 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1229 to check whether it is stale, because sometimes, we want the
1230 enqueue happened anyway, and op_data.fid2 will not be initialized.
1232 Severity : enhancement
1234 Description: update patchless client
1235 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1238 Frequency : only with 2.4 kernel
1240 Description: random memory corruption
1241 Details : size of struct ll_inode_info is to big for union inode.u and this
1242 can be cause of random memory corruption.
1247 Description: Memory leak in recovery
1248 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1249 It should also check obd_fail before fsfilt_start, otherwise if
1250 fsfilt_start return -EROFS,(failover mds during mds recovery).
1251 then the req will return with repmsg->transno = 0 and rc = EROFS.
1252 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1253 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1254 be freed no matter whether fsfilt_commit success or not.
1257 Frequency : only with huge count clients
1259 Description: Prevents from taking the superblock lock in llap_from_page for
1261 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1262 for llap_from_page call in ll_removepage prevents from taking the
1263 superblock lock for a soon died page.
1268 Description: Not check open intent error before release open handle
1269 Details : in some rare cases, the open intent error is not checked before
1270 release open handle, which may cause
1271 ASSERTION(open_req->rq_transno != 0), because it tries to release
1272 the failed open handle.
1277 Description: Set cat log bitmap only after create log success.
1278 Details : in some rare cases, the cat log bitmap is set too early. and it
1279 should be set only after create log success.
1283 Description: Accessing a block bevice can re-enable I/O when Lustre is
1284 tearing down a device.
1285 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1290 Description: service threads may hog cpus when there are a lot of requests
1291 Details : Insert cond_resched to give other threads a chance to use some CPU
1296 Description: the cat log was not initialized in recovery
1297 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1298 unlink log on mds will not be initialized until mds post
1299 recovery. And also in mds post recovery, the unlink log will
1300 initialization will be done asynchronausly, so there will be race
1301 between add unlink log and unlink log initialization.
1305 Description: brw_stats were being printed incorrectly
1306 Details : brw_stats were being printed as log2 but all of them were not
1307 recorded as log2. Also remove some code duplication arising from
1308 filter_tally_{read,write}.
1312 Frequency : rare, only in recovery.
1313 Description: ASSERTION(req->rq_type != LI_POISON) failed
1314 Details : imp_lock should be held while iterating over imp_sending_list for
1315 prevent destroy request after get timeout in ptlrpc_queue_wait.
1319 Description: replay-single.sh test 52 fails
1320 Details : A lock's skiplist need to be cleanup when it being unlinked
1321 from its resource list.
1325 Description: Short directio read returns full requested size rather than
1327 Details : Direct I/O operations should return actual amount of bytes
1328 transferred rather than requested size.
1330 Severity : enhancement
1332 Description: metadata RPC reduction (e.g. for rm performance)
1333 Details : decrease the amount of synchronous RPC between clients and servers
1334 by canceling conflicing lock before the operation on the client
1335 and packing thier handles into the main operation RPC to server.
1337 Severity : enhancement
1339 Description: Async OSC create to avoid the blocking unnecessarily.
1340 Details : If a OST has no remain object, system will block on the creating
1341 when need to create a new object on this OST. Now, ways use
1342 pre-created objects when available, instead of blocking on an
1343 empty osc while others are not empty. If we must block, we block
1344 for the shortest possible period of time.
1346 Severity : enhancement
1348 Description: refine locking for avoid write wrong info into lov_objid file
1349 Details : fix possible races with add new target and write/update data in
1352 --------------------------------------------------------------------------------
1354 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1359 Frequency : on some architectures
1361 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1362 Details : Uninitialized flags sometimes cause configuration commands to
1366 Frequency : patchless clients only
1368 Description: missing __iget() symbol export
1369 Details : The __iget() symbol export is missing. To avoid the need for
1370 this on patchless clients the deathrow inode reaper is turned
1371 off, and we depend on the VM to clean up old inodes. This
1372 dependency was during via the fix for bug 12181.
1374 --------------------------------------------------------------------------------
1376 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1378 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1379 INTEROPERATE with older versions automatically. In many cases a
1380 special upgrade step is needed. Please read the
1381 user documentation before upgrading any part of a live system.
1382 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1383 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1384 * WARNING: Lustre configuration and startup changes are required with
1385 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1391 Severity : enhancement
1393 Description: MountConf
1394 Details : Lustre configuration is now managed via mkfs and mount
1395 commands instead of lmc and lconf. New obd types (MGS, MGC)
1396 are added for dynamic configuration management. See
1397 https://mail.clusterfs.com/wikis/lustre/MountConf for
1400 Severity : enhancement
1402 Description: dynamic OST addition
1403 Details : OSTs can now be added to a live filesystem
1405 Severity : enhancement
1407 Description: startup order invariance
1408 Details : MDTs and OSTs can be started in any order. Clients only
1409 require the MDT to complete startup.
1411 Severity : enhancement
1413 Description: parallel, asynchronous orphan cleanup
1414 Details : orphan cleanup is now performed in separate threads for each
1415 OST, allowing parallel non-blocking operation.
1417 Severity : enhancement
1419 Description: optimized stripe assignment
1420 Details : stripe assignments are now made based on ost space available,
1421 ost previous usage, and OSS previous usage, in order to try
1422 to optimize storage space and networking resources.
1424 Severity : enhancement
1426 Description: Permanently set tunables
1427 Details : All writable /proc/fs/lustre tunables can now be permanently
1428 set on a per-server basis, at mkfs time or on a live system.
1430 Severity : enhancement
1432 Description: Lustre message v2
1433 Details : Add lustre message format v2.
1435 Severity : enhancement
1437 Description: client OST exclusion list
1438 Details : Clients can be started with a list of OSTs that should be
1439 declared "inactive" for known non-responsive OSTs.
1443 Description: SPEC SFS validation failure on NFS v2 over lustre.
1444 Details : Changes the blocksize for regular files to be 2x RPC size,
1445 and not depend on stripe size.
1447 Severity : enhancement
1449 Description: Multiple MD RPCs in flight.
1450 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1451 To avoid overly-overloading MDS, introduce a limit on number of
1452 MDS RPCs in flight for a single client and add /proc controls
1453 to adjust this limit.
1455 Severity : enhancement
1457 Description: client read/write statistics
1458 Details : Add client read/write call usage stats for performance
1459 analysis of user processes.
1460 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1461 file access. extents_stats shows chunk size distribution.
1462 extents_stats_per_process show chunk size distribution per
1465 Severity : enhancement
1467 Description: mds statistics
1468 Details : Add detailed mds operations statistics in
1469 /proc/fs/lustre/mds/*/stats.
1473 Description: Failure of copying files with lustre special EAs.
1474 Details : Client side always return success for setxattr call for lustre
1475 special xattr (currently only "trusted.lov").
1480 Description: Refcount LNET uuids
1481 Details : The global LNET uuid list grew linearly with every startup;
1482 refcount repeated list entries instead of always adding to
1485 Severity : enhancement
1487 Description: Dynamic service threads
1488 Details : Within a small range, start extra service threads
1489 automatically when the request queue builds up.
1492 Frequency : mixed-endian client/server environments
1494 Description: mixed-endian crashes
1495 Details : The new msg_v2 system had some failures in mixed-endian
1498 Severity : enhancement
1500 Description: Easy OST removal
1501 Details : OSTs can be permanently deactivated with e.g. 'lctl
1502 conf_param lustre-OST0001.osc.active=0'
1504 Severity : enhancement
1506 Description: MGS proc entries
1507 Details : Added basic proc entries for the MGS showing what filesystems
1510 Severity : enhancement
1512 Description: provide MGS failover
1513 Details : Added config lock reacquisition after MGS server failover.
1515 Severity : enhancement
1517 Description: add Linux 2.4 support
1518 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1522 Description: a large application tries to do I/O to the same resource and dies
1523 in the middle of it.
1524 Details : Check the req->rq_arrival time after the call to
1525 ost_brw_lock_get(), but before we do anything about
1526 processing it & sending the BULK transfer request. This
1527 should help move old stale pending locks off the queue as
1528 quickly as obd_timeout.
1531 Frequency : when an incorrect nid is specified during startup
1533 Description: ptlrpc connect to non-existant node causes kernel crash
1534 Details : LNET can't be re-entered from an event callback, which
1535 happened when we expire a message after the export has been
1536 cleaned up. Instead, hand the zombie cleanup off to another
1539 Severity : enhancement
1541 Description: plain/inodebits lock performance improvement
1542 Details : Grouping plain/inodebits in granted list by their request modes
1543 and bits policy, thus improving the performance of search through
1547 Frequency : only if OST filesystem is corrupted
1549 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1550 Details : for a short time RPCs with bulk IO are in the replay list,
1551 but replay of bulk IOs is unimplemented. If the OST filesystem
1552 is corrupted due to disk cache incoherency and then replay is
1553 started it is possible to trip an assertion. Avoid putting
1554 committed RPCs into the replay list at all to avoid this issue.
1557 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1559 Description: System hang on startup
1560 Details : This bug allowed the liblustre (e.g. catamount) client to
1561 return to the app before handling all startup RPCs. This
1562 could leave the node unresponsive to lustre network traffic
1563 and manifested as a server ptllnd timeout.
1565 Severity : enhancement
1567 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1568 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1569 boolean to control whether to print peer debug info when a
1570 client's RPC times out.
1573 Frequency : only for kernels with patches from Lustre below 1.4.3
1575 Description: Remove old rdonly API
1576 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1579 Frequency : only for devices with external journals
1581 Description: Set external device read-only also
1582 Details : During a commanded failover stop, we set the disk device
1583 read-only while the server shuts down. We now also set any
1584 external journal device read-only at the same time.
1587 Frequency : when upgrading from 1.4 while trying to change parameters
1589 Description: The wrong (new) MDC name was used when setting parameters for
1590 upgraded MDT's. Also allows changing of OSC (and MDC)
1591 parameters if --writeconf is specified at tunefs upgrade time.
1594 Frequency : when setting specific ost indicies
1596 Description: QOS code breaks on skipped indicies
1597 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1598 created with --index need not be sequential.
1600 Severity : enhancement
1602 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1603 Details : The uninit_groups feature works in conjunction with the kernel
1604 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1605 up the pass1 processing of e2fsck. This is a read-only feature
1606 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1607 mount filesystems that have had this feature enabled.
1609 Severity : enhancement
1611 Description: Improve multi-block allocation algorithm to avoid fragmentation
1612 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1613 allocation locality and avoid filesystem fragmentation.
1615 ------------------------------------------------------------------------------
1617 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1619 * Support for kernels:
1620 2.4.21-47.0.1.EL (RHEL 3)
1621 2.6.5-7.283 (SLES 9)
1622 2.6.9-42.0.10.EL (RHEL 4)
1623 2.6.12.6 vanilla (kernel.org)
1624 2.6.16.27-0.9 (SLES 10)
1625 * Recommended e2fsprogs version: 1.39.cfs5
1627 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1631 Frequency : occasional, depends on client load and configuration
1632 Bugzilla : 12181, 12203
1633 Description: data loss for recently-modified files
1635 Details : In some cases it is possible that recently written or created
1636 files may not be written to disk in a timely manner (this should
1637 normally be within 30s unless client IO load is very high).
1638 The problem appears as zero-length files or files that are a
1639 multiple of 1MB in size after a client crash or client eviction
1640 that are missing data at the end of the file.
1642 This problem is more likely to be hit on clients where files are
1643 repeatedly created and unlinked in the same directory, clients
1644 have a large amount of RAM, have many CPUs, the filesystem has
1645 many OSTs, the clients are rebooted frequently, and/or the files
1646 are not accessed by other nodes after being written.
1648 The presence of the problem can be detected by looking at
1649 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1650 smaller than the second (nr_unused) then dirty files will not
1651 be flushed automatically to disk. "sync; sleep 10" should be
1652 run several times on the node before unmounting it to update
1653 Lustre (this is also safe to run on nodes without this problem).
1655 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1656 that can cause this same problem, so customers using that kernel
1657 also need to update the kernel in addition to Lustre. In order
1658 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1660 It is normal that files written just before a client crash (less
1661 than 30s) may not yet have been flushed to disk, even for local
1665 Frequency : frequent on thin XT3 nodes
1667 Description: UUID collision on thin XT3 Linux nodes
1668 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1669 randomly, since we relied on an insufficiently-seeded PRNG.
1674 Description: OSS hangs after "All ost request buffers busy"
1675 Details : A deadlock between quota and journal operations caused OSS
1676 hangs after printing "All ost request buffers busy."
1679 Frequency : always on liblustre builds
1681 Description: Cleanup compiler warnings on liblustre
1684 Frequency : always on liblustre builds on XT3
1686 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1690 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1691 > 32000 subdirectories
1692 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1693 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1694 subdirectory count crosses 32000. This will aid e2fsck to
1695 correctly handle more than 32000 subdirectories.
1698 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1700 Description: System hang on startup
1701 Details : This bug allowed the liblustre (e.g. catamount) client to
1702 return to the app before handling all startup RPCs. This
1703 could leave the node unresponsive to lustre network traffic
1704 and manifested as a server ptllnd timeout.
1706 Severity : enhancement
1708 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1709 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1710 boolean to control whether to print peer debug info when a
1711 client's RPC times out.
1716 Description: make O_SYNC working on 2.6 kernels
1717 Details : 2.6 kernels use different method for mark pages for write,
1718 so need add a code to lustre for O_SYNC work.
1723 Description: Failure to close file and release space on NFS
1724 Details : Put inode details into lock acquired in ll_intent_file_open.
1725 Use mdc_intent_lock in ll_intent_open to properly
1726 detect all kind of errors unhandled by mdc_enqueue.
1731 Description: proc file read during shutdown sometimes raced obd removal,
1733 Details : Add lock to prevent obd access after proc file removal.
1736 Frequency : Only for files larger than 4GB on 32-bit clients.
1738 Description: improperly doing page alignment of locks
1739 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1740 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1745 Description: RPCs being resent when they shouldn't be
1746 Details : Some RPCs that should not be resent are being resent. This
1747 can cause inconsistencies in the RPC state machine. Do not
1748 resend such requests.
1751 Frequency : rare, only with NFS export
1753 Description: Crash on NFS re-export node
1754 Details : under very unusual load conditions an assertion is hit in
1755 ll_intent_file_open()
1758 Frequency : only if OST filesystem is corrupted
1760 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1761 Details : for a short time RPCs with bulk IO are in the replay list,
1762 but replay of bulk IOs is unimplemented. If the OST filesystem
1763 is corrupted due to disk cache incoherency and then replay is
1764 started it is possible to trip an assertion. Avoid putting
1765 committed RPCs into the replay list at all to avoid this issue.
1770 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1771 Details : Large single O_DIRECT read and write calls can fail to allocate
1772 a sufficiently large buffer to process the request. In case of
1773 allocation failure the allocation is retried with a smaller
1774 buffer and broken into smaller requests.
1776 Severity : enhancement
1778 Description: Add -o localflock option to simulate old noflock behaviour.
1779 Details : This will achieve local-only flock/fcntl locks coherentness.
1784 Description: versioning check is incomplete
1785 Details : Checking the version difference of client vs. server, report
1786 error if the gap is too big.
1791 Description: add support PG_writeback bit
1792 Details : add support for PG_writeback bit for Lustre, for more carefull
1793 work with page cache in 2.6 kernel. This also fix some deadlocks
1794 and remove hack for work O_SYNC with 2.6 kernel.
1796 Severity : enhancement
1798 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1799 Details : The uninit_groups feature works in conjunction with the kernel
1800 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1801 up the pass1 processing of e2fsck. This is a read-only feature
1802 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1803 mount filesystems that have had this feature enabled.
1805 Severity : enhancement
1807 Description: Improve multi-block allocation algorithm to avoid fragmentation
1808 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1809 allocation locality and avoid filesystem fragmentation.
1811 ------------------------------------------------------------------------------
1813 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1815 * Support for kernels:
1816 2.6.9-42.0.3.EL (RHEL 4)
1817 2.6.5-7.276 (SLES 9)
1818 2.4.21-47.0.1.EL (RHEL 3)
1819 2.6.12.6 vanilla (kernel.org)
1820 2.6.16.21-0.8 (SLES10)
1821 * Recommended e2fsprogs version: 1.39.cfs2-0
1823 * The backwards-compatible /proc/sys/portals symlink has been removed
1824 in this release. Before upgrading, please ensure that you change
1825 any configuration scripts or /etc/sysctl.conf files that access
1826 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1827 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1828 in advance of the upgrade on any system running Lustre 1.4.6 or
1829 newer, since /proc/sys/lnet was added in that version.
1830 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1834 Frequency : only when quota is used
1836 Description: avoid scanning export list for quota master
1837 Details : Change the algorithms to avoid scanning export list in order
1838 to improve the efficiency.
1841 Frequency : MDS failover only, very rarely
1843 Description: "went back in time" messages on mds failover
1844 Details : The greatest transno may be lost when the current operation
1845 finishes with an error (transno==0) and the client's last_rcvd
1846 record is over-written. Save the greatest transno in the
1847 mds_last_transno for this case.
1850 Frequency : always for specific kernels and striping counts
1852 Description: client may get "Matching packet too big" without ACL support
1853 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1854 when trying to access files in certain configurations. The
1855 clients should in fact be denied when mounting because they do
1856 not understand ACLs.
1859 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1861 Description: many clients connecting with IO in progress causes connect timeouts
1862 Details : Avoid synchronous journal commits to avoid delays caused by many
1863 clients connecting/disconnecting when bulk IO is in progress.
1864 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1865 OST_IO_PORTAL to avoid delays behind potentially many pending
1869 Frequency : occasionally with multiple writers to a single file
1871 Description: shared writes to file may result in wrong size reported by stat()
1872 Details : Allow growing of kms when extent lock is cancelled
1875 Frequency : always with random mmap IO to multi-striped file
1877 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1878 Details : Only if the hole is at the end of OST object so that kms is too
1879 small. Fix is to increase kms accordingly in ll_nopage.
1882 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1884 Description: writes to a missing object would leak memory on the OST
1885 Details : If there is an inconsistency between the MDS and OST filesystems,
1886 such that the MDS references an object that doesn't exist, writes
1887 to that object will leak memory due to incorrect cleanup in the
1888 error handling path, eventually running out of memory on the OST.
1893 Description: Creating too long symlink causes lustre errors
1894 Details : Check symlink and name lengths before sending requests to MDS.
1897 Frequency : only if flock is enabled (not on by default)
1899 Description: posix locks not released on fd closure on 2.6.9+
1900 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1901 kernels, this caused such locks not to be released on fd close and
1902 then assertions on fs unmount about still used locks.
1905 Frequency : MDS failover only, very rarely
1907 Description: clients may get ASSERTION(granted_lock != NULL)
1908 Details : When request was taking a long time, and a client was resending
1909 a getattr by name lock request. The were multiple lock requests
1910 with the same client lock handle and
1911 mds_getattr_name->fixup_handle_for_resent_request found one of the
1912 lock handles but later failed with ASSERTION(granted_lock != NULL).
1917 Description: handle->h_buffer_credits > 0, assertion failure
1918 Details : h_buffer_credits is zero after truncate, causing assertion
1919 failure. This patch extends the transaction or creates a new
1923 Frequency : NFS re-export or patchless client
1924 Bugzilla : 11179, 10796
1925 Description: Crash on NFS re-export node (__d_move)
1926 Details : We do not want to hash the dentry if we don't have a lock.
1927 But if this dentry is later used in d_move, we'd hit uninitialised
1928 list head d_hash, so we just do this to init d_hash field but
1929 leave dentry unhashed.
1932 Frequency : NFS re-export or patchless client
1934 Description: NFS exports has problem with symbolic link
1935 Details : lustre client didn't properly install dentry when re-exported
1936 to NFS or running patchless client.
1939 Frequency : NFS re-export or patchless client
1941 Description: Various nfs/patchless fixes.
1942 Details : fixes reuse disconected alias for lookup process - this fixes
1943 warning "find_exported_dentry: npd != pd",
1944 fix permission error with open files at nfs.
1945 fix apply umask when do revalidate.
1948 Frequency : occasional
1950 Description: Crash on NFS re-export node
1951 Details : calling clear_page() on the wrong pointer triggered oops in
1952 generic_mapping_read().
1955 Frequency : rarely, using O_DIRECT IO
1957 Description: unaligned directio crashes client with LASSERT
1958 Details : check for unaligned buffers before trying any requests.
1961 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1963 Description: stale data returned from RAID cache
1964 Details : If only a small amount of IO is done to the RAID device before
1965 reading it again it is possible to get stale data from the RAID
1966 cache instead of reading it from disk.
1969 Frequency : always for sles10 kernel
1971 Description: sles10 support
1972 Details : ll_follow_link: compile fixes and using of nd_set_link
1973 under newer kernels.
1976 Frequency : depends on arch, kernel and compiler version, always on sles10
1979 Description: recursive or deep enough symlinks cause stack overflow
1980 Details : getting rid of large stack-allocated variable in
1984 Frequency : depends on hardware
1986 Description: lustre write performance loss in the SLES10 kernel
1987 Details : the performance loss is caused by using of write barriers in the
1988 ext3 code. The SLES10 kernel turns barrier support on by
1989 default. The fix is to undo that change for ldiskfs.
1991 ------------------------------------------------------------------------------
1993 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1995 * Support for kernels:
1996 2.6.9-42.0.3EL (RHEL 4)
1997 2.6.5-7.276 (SLES 9)
1998 2.4.21-47.0.1.EL (RHEL 3)
1999 2.6.12.6 vanilla (kernel.org)
2003 Frequency : quota enabled and large files being deleted
2005 Description: releasing more than 4GB of quota at once hangs OST
2006 Details : If a user deletes more than 4GB of files on a single OST it
2007 will cause the OST to spin in an infinite loop. Release
2008 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2013 Description: statfs data retrieved from /proc may be stale or zero
2014 Details : When reading per-device statfs data from /proc, in the
2015 {kbytes,files}_{total,free,avail} files, it may appear
2016 as zero or be out of date.
2019 Frequency : systems with MD RAID1 external journal devices
2021 Description: lconf's call to blkid is confused by RAID1 journal devices
2022 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2023 of returning all block devices that match the journal UUID.
2026 Frequency : always, for aggregate stripe size over 4GB
2028 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2029 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2030 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2033 Frequency : always if "lfs find" used on a local file/directory
2035 Description: "lfs find" segfaults if used on a local file/directory
2036 Details : The case where a directory component was not specified wasn't
2037 handled correctly. Handle this properly.
2040 Frequency : always on ppc64
2042 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2043 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2044 perform endianness conversion.
2047 Frequency : rarely (truncate to non-zero file size after write under load)
2048 Bugzilla : 10730, 10687
2049 Description: Files padded with zeros to next 4K multiple
2050 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2051 it is possible that files that are truncated to a non-zero size
2052 immediately after being written are filled with zero bytes beyond
2053 the truncated size. No file data is lost.
2055 Severity : enhancement
2057 Description: Allow recovery/failover for liblustre clients.
2058 Details : liblustre clients were unaware of failover configurations until
2061 Severity : enhancement
2063 Description: user file locks should fail when not mounting with flock option
2064 Details : Set up an error-returning stub in ll_file_operations.lock field
2065 to prevent incorrect behaviour when client is mounted without
2066 flock option. Also, set up properly f_op->flock field for
2070 Frequency : always on ia64
2072 Description: "lfs df" loops on printing out MDS statfs information
2073 Details : The obd_ioctl_data was not initialized and in some systems
2074 this caused a failure during the ioctl that did not return
2075 an error. Initialize the struct and return an error on failure.
2078 Frequency : SLES 9 only
2080 Description: Error of copying files with lustre special EAs as root
2081 Details : Client side always return success for setxattr call for lustre
2082 special xattr (currently only "trusted.lov").
2085 Frequency : rarely on clusters with both ia64+i386 clients
2087 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2088 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2089 clients are doing concurrent writes to the same file it is
2090 possible that the ia64 clients may LASSERT because the OST
2091 extent locks are not PAGE_SIZE aligned. Ensure that grown
2092 locks are always aligned on the request boundary.
2095 Frequency : specific use, occasional
2097 Description: Overwriting in use executable truncates on-disk binary image
2098 Details : If one node attempts to overwrite an executable in use by
2099 another node, we now correctly return ETXTBSY instead of
2100 truncating the file.
2102 Severity : enhancement
2104 Description: Async OSC create to avoid the blocking unnecessarily.
2105 Details : If a OST has no remain object, system will block on the creating
2106 when need to create a new object on this OST. Now, ways use
2107 pre-created objects when available, instead of blocking on an
2108 empty osc while others are not empty. If we must block, we block
2109 for the shortest possible period of time.
2114 Description: chmod on Lustre root is propagated to other clients
2115 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2116 invalid by the follow_mount time.
2119 Frequency : liblustre clients only
2121 Description: Race in 'instant cancel' lock handling could lead to such locks
2122 never to be granted in case of SMP MDS
2123 Details : Do not destroy not yet granted but cbpending locks in
2127 Frequency : replay/resend of open
2129 Description: non null lock assetion failure in mds_intent_policy
2130 Details : Trying to replay/resend lockless open requests resulted in
2131 mds_open() returning 0 with no lock. Now it sets a flag if
2132 a lock is going to be returned.
2134 Severity : enhancement
2136 Description: Checksum enhancements
2137 Details : New checksum enhancements allow for resending RPCs that failed
2140 Severity : enhancement
2142 Description: Tunables on number of dirty pages in cacche
2143 Details : Allow to set limit on number of dirty pages cached.
2148 Description: client crash on unmount - lock still has references
2149 Details : In some error handling cases it was possible to leak a lock
2150 reference on a client while accessing a file. This was not
2151 harmful to the client during operation, but would cause the
2152 client to crash when the filesystem is unmounted.
2155 Frequency : specific case, rare
2157 Description: ETXTBSY on mds though file not in use by client
2158 Details : ETXTBSY is no longer incorrectly returned when attempting to
2159 chmod or chown a directory that the user previously tried to
2160 execute or a currently-executing binary.
2163 Frequency : extremely rare except on liblustre-based clients
2165 Description: Lustre space not freed when files are deleted
2166 Details : Clean up open-unlinked files after client eviction. Previously
2167 the unlink was skipped and the files remained as orphans.
2172 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2173 Details : In some cases it was possible to send a blocking callback to
2174 a client doing a glimpse, even though that client didn't get
2175 a lock granted. When the glimpse lock is cancelled on the OST
2176 the freed lock is left on the waiting list and corrupted the list.
2179 Frequency : all core dumps
2181 Description: Broke core dumps to lustre
2182 Details : Negative dentry may be unhashed if parent does not have UPDATE
2183 lock, but some callers, e.g. do_coredump, expect dentry to be
2184 hashed after successful create, hash it in ll_create_it.
2186 ------------------------------------------------------------------------------
2188 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2190 * Support for kernels:
2191 2.6.9-42.0.2.EL (RHEL 4)
2192 2.6.5-7.276 (SLES 9)
2193 2.4.21-40.EL (RHEL 3)
2194 2.6.12.6 vanilla (kernel.org)
2198 Frequency : always on RHEL 3
2200 Description: Number of open files grows over time
2201 Details : The number of open files grows over time, whether or not
2202 Lustre is started. This was due to a filp leak introduced
2203 by one of our kernel patches.
2205 ------------------------------------------------------------------------------
2207 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2209 * Support for kernels:
2210 2.6.9-42.EL (RHEL 4)
2211 2.6.5-7.276 (SLES 9)
2212 2.4.21-40.EL (RHEL 3)
2213 2.6.12.6 vanilla (kernel.org)
2218 Bugzilla : 5719, 9635, 9792, 9684
2219 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2220 Details : If a server is under heavy load and cannot reply to new
2221 connection requests before the client resends the (re)connect,
2222 the connection handling code can behave badly if two service
2223 threads are concurrently handing separate (re)connections from
2224 the same client. Add better locking to the connection handling
2225 code, and ensure that only a single connection will be processed
2226 for a given client UUID, even if the lock is dropped.
2228 Severity : enhancement
2230 Description: add TCP zero-copy support to kernel
2231 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2232 sends if the hardware supports scatter-gather and checksumming.
2233 This allows socklnd to do client-write and server-read more
2234 efficiently and reduce CPU utilization from skbuf copying.
2237 Frequency : only if NFS exporting from client
2239 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2240 Details : If setting attributes on a file created under NFS that had
2241 never been opened it would be possible to oops the client
2242 if the file had no objects.
2245 Frequency : always for liblustre
2247 Description: liblustre client does MDS+OSTs setattr RPC for each write
2248 Details : When doing a write from a liblustre client, the client
2249 incorrectly issued an RPC to the MDS and each OST the file was
2250 striped over in order to update the timestamps. When writing
2251 with small chunks and many clients this could overwhelm the MDS
2252 with RPCs. In all cases it would slow down the write because
2253 these RPCs are unnecessary.
2255 Severity : enhancement
2257 Description: allow number of MDS service threads to be changed at module load
2258 Details : It is now possible to change the number of MDS service threads
2259 running. Adding "options mds mds_num_threads={N}" to the MDS's
2260 /etc/modprobe.conf will set the number of threads for the next
2261 time Lustre is restarted (assuming the "mds" module is also
2262 reloaded at that time). The default number of threads will
2263 stay the same, 32 for most systems.
2268 Description: OST crash if filesystem is unformatted or corrupt
2269 Details : If an OST is started on a device that has never been formatted
2270 or if the filesystem is corrupt and cannot even mount then the
2271 error handling cleanup routines would dereference a NULL pointer.
2276 Description: NULL pointer deref in llap_from_page.
2277 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2278 incorrect) mapping if the page was truncated/reclaimed while it was
2279 searched for. Check for this condition and skip such pages when
2280 doing readahead. Introduce extra check to llap_from_page() to
2281 verify page->mapping->host is non-NULL (so page is not anonymous).
2284 Frequency : Sometimes when using sys_sendfile
2286 Description: "page not covered by a lock" warnings from ll_readpage
2287 Details : sendfile called ll_readpage without right page locks present.
2288 Now we introduced ll_file_sendfile that does necessary locking
2289 around call to generic_file_sendfile() much like we do in
2293 Frequency : with certain MDS communication failures at client mount time
2295 Description: NULL pointer deref after failed client mount
2296 Details : a client connection request may delayed by the network layer
2297 and not be sent until after the PTLRPC layer has timed out the
2298 request. If the client fails the mount immediately it will try
2299 to clean up before the network times out the request. Add a
2300 reference from the request import to the obd device and delay
2301 the cleanup until the network drops the request.
2304 Frequency : occasionally during client (re)connect
2306 Description: assertion failure during client (re)connect
2307 Details : processing a client connection request may be delayed by the
2308 client or server longer than the client connect timeout. This
2309 causes the client to resend the connection request. If the
2310 original connection request is replied in this interval, the
2311 client may trip an assertion failure in ptlrpc_connect_interpret()
2312 which thought it would be the only running connect process.
2315 Frequency : only with obd_echo servers and clients that are rebooted
2317 Description: kernel BUG accessing uninitialized data structure
2318 Details : When running an obd_echo server it did not start the ping_evictor
2319 thread, and when a client was evicted an uninitialized data
2320 structure was accessed. Start the ping_evictor in the RPC
2321 service startup instead of the OBD startup.
2323 Severity : enhancement
2324 Bugzilla : 10193 (patchless)
2325 Description: Remove dependency on various unexported kernel interfaces.
2326 Details : No longer need reparent_to_init, exit_mm, exit_files,
2327 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2330 Frequency : rare (only users of deprecated and unsupported LDAP config)
2332 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2333 Details : LDAP apparently contains 'lustreName' attributes instead of
2334 'name'. A simple remapping of the name is sufficient.
2337 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2339 Description: waiting_locks_callback trips kernel BUG if client is evicted
2340 Details : Running with the dump_on_timeout debug flag turned on makes
2341 it possible that the waiting_locks_callback() can try to dump
2342 the Lustre kernel debug logs from an interrupt handler. Defer
2343 this log dumping to the expired_lock_main() thread.
2345 Severity : enhancement
2347 Description: Support NFS exporting on 2.6 kernels.
2348 Details : Implement non-rawops metadata methods for NFS server to use without
2349 changing NFS server code.
2352 Frequency : very rare (synthetic metadata workload only)
2354 Description: two racing renames might cause an MDS thread to deadlock
2355 Details : Running the "racer" program may cause one MDS thread to rename
2356 a file from being the source of a rename to being the target of
2357 a rename at exactly the same time that another thread is doing
2358 so, and the second thread has already enqueued these locks after
2359 doing a lookup of the target and is trying to relock them in
2360 order. Ensure that we don't try to re-lock the same resource.
2363 Frequency : only very large systems with liblustre clients
2365 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2366 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2367 all OSTs in parallel. This allows the eviction of stale liblustre
2368 clients to proceed much faster than if they were done in series,
2369 and also offers similar improvements for other set_info RPCs.
2374 Description: excessive CPU usage during initial read phase on client
2375 Details : During the initial read phase on a client, it would agressively
2376 retry readahead on the file, consuming too much CPU and impacting
2377 performance (since 1.4.5.8). Improve the readahead algorithm
2378 to avoid this, and also improve some other common cases (read
2379 of small files in particular, where "small" is files smaller than
2380 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2385 Description: MDS crash when receiving packet with unknown intent.
2386 Details : Do not LBUG in unknown intent case, just return -EFAULT
2388 Severity : enhancement
2389 Bugzilla : 9293, 9385
2390 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2391 Details : Do not serialize getattr (non-intent version) and statfs.
2394 Frequency : occasional, when OST network is overloaded/intermittent
2396 Description: client evicted by OST after bulk IO timeout
2397 Details : If a client sends a bulk IO request (read or write) the OST
2398 may evict the client if it is unresposive to its data GET/PUT
2399 request. This is incorrect if the network is overloaded (takes
2400 too long to transfer the RPC data) or dropped the OST GET/PUT
2401 request. There is no need to evict the client at all, since
2402 the pinger and/or lock callbacks will handle this, and the
2403 client can restart the bulk request.
2406 Frequency : Always when mmapping file with no objects
2408 Description: client crashes when mmapping file with no objects
2409 Details : Check that we actually have objects in a file before doing any
2410 operations on objects in ll_vm_open, ll_vm_close and
2416 Description: Request leak when working with deleted CWD
2417 Details : Introduce advanced request refcount tracking for requests
2418 referenced from lustre intent.
2420 Severity : Enhancement
2422 Description: Cache open file handles on client.
2423 Details : MDS now will return special lock along with openhandle, if
2424 requested and client is allowed to hold openhandle, even if unused,
2425 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2426 closing files for every read/write openration.
2428 Severity : Enhancement
2430 Description: Cache open negative dentries on client when possible.
2431 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2432 negative dentries on lock revocation.
2437 Description: Remounting a client read-only wasn't possible with a zconf mount
2438 Details : It wasn't possible to remount a client read-only with llmount.
2440 Severity : enhancement
2441 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2442 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2443 MPICH is included to add Lustre-specific ADIO interfaces.
2444 This is based closely on the UFS ADIO layer and only differs
2445 in file creation, in order to allow the OST striping to be set.
2446 This is user-contributed code and not supported by CFS.
2451 Description: extended inode attributes (immutable, append-only) work improperly
2452 when 2.4 and 2.6 kernels are used on client/server or vice versa
2453 Details : Introduce kernel-independent values for these flags.
2455 Severity : enhancement
2458 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2459 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2460 granularity. Now, Fractional MB granularity is be supported,
2461 this is very useful for low memory system.
2463 Severity : enhancement
2465 Description: Getattr by fid
2466 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2467 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2468 have locks and avoids getattr rpc requests for it on every stat.
2471 Frequency : Always, for filesystems larger than 2TB
2473 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2474 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2475 filesystems larger than 2TB to be mounted with mballoc enabled.
2478 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2480 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2481 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2482 filesystems larger than 2TB is mounted then the kernel will
2483 silently corrupt the start of the filesystem. CONFIG_LBD is
2484 enabled for all CFS-supported kernels, but the possibility of
2485 this happening with a modified kernel config exists.
2487 Severity : enhancement
2489 Description: add client O_DIRECT support for 2.6 kernels
2490 Details : It is now possible to do O_DIRECT reads and writes to files
2491 in the Lustre client mountpoint on 2.6 kernel clients.
2493 Severity : enhancement
2495 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2496 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2497 parallel, not waiting for response from every OST before sending
2498 a rpc to the next OST.
2503 Description: setattr vs write race when updating file timestamps
2504 Details : Client processes that update a file timestamp into the past
2505 right after writing to the file (e.g. tar) it is possible that
2506 the updated file modification time can be reset to the current
2507 time due to a race between processing the setattr and write RPC.
2509 Severity : enhancement
2511 Description: Bring 'lfs find' closer in line with regular Linux find.
2512 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2513 -print0 options and obtains all the needed info through the lustre
2516 Severity : enhancement
2518 Description: support up to 1024 configured devices on one node
2519 Details : change obd_dev array from statically allocated to dynamically
2520 allocated structs as they are first used to reduce memory usage
2525 Description: Flush dirty partially truncated pages during truncate
2526 Details : Immediatelly flush partially truncated pages in filter_setattr,
2527 this way we completely avoid having any pages in page cache on OST
2528 and can retire ugly workarounds during writes to flush such pages.
2533 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2534 Details : protect inode from truncation within vfs_unlink() context
2535 just take a reference before calling vfs_unlink() and release it
2536 when parent's i_sem is free.
2541 Description: last_id value checked outside lock on OST caused LASSERT failure
2542 Details : If there were multiple MDS->OST object precreate requests in
2543 flight, it was possible that the OST's last object id was checked
2544 outside a lock and incorrectly tripped an assertion. Move checks
2545 inside locks, and discard old precreate requests.
2548 Frequency : always, if extents are used on OSTs
2550 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2551 Details : OSTs using the extents format would not zero the high 16 bits of
2552 the index physical block number. This is not a problem for any
2553 OST filesystems smaller than 16TB, and no kernels support ext3
2554 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2555 new/modified files) and can be fixed for existing filesystems
2556 with e2fsprogs-1.39-cfs1.
2561 Description: import connection selection may be incorrect if timer wraps
2562 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2563 import connections to be ignored if the 32-bit jiffies counter
2564 wraps. Use a 64-bit jiffies counter.
2567 Frequency : very large clusters immediately after boot
2569 Description: LNET request buffers exhausted under heavy short-term load
2570 Details : If a large number of client requests are generated on a service
2571 that has previously never seen so many requests it is possible
2572 that the request buffer growth cannot keep up with the spike in
2573 demand. Instead of dropping incoming requests, they are held in
2574 the LND until the RPC service can accept more requests.
2577 Frequency : Sometimes during replay
2579 Description: Assertion failure in ll_local_open after replay.
2580 Details : If replay happened on an open request reply before we were able
2581 to set replay handler, reply will become not swabbed tripping the
2582 assertion in ll_local_open. Now we set the handler right after
2583 recognising of open request
2586 Frequency : very rare
2588 Description: kernel reports "badness in vsnprintf"
2589 Details : Reading from the "recovery_status" /proc file in small chunks
2590 may cause a negative length in lprocfs_obd_rd_recovery_status()
2591 call to vsnprintf() (which is otherwise harmless). Exit early
2592 if there is no more space in the output buffer.
2594 Severity : enhancement
2596 Description: clear OBD RPC statistics by writing to them
2597 Details : It is now possible to clear the OBD RPC statistics by writing
2598 to the "stats" file.
2603 Description: Client mtime is not the same on different clients after utimes
2604 Details : In some cases, the client was using the utimes() syscall on
2605 a file cached on another node. The clients now validate the
2606 ctime from the MDS + OSTs to determine which one is right.
2611 Description: Inability to activate failout mode
2612 Details : lconf script incorrectly assumed that in pythong string's numeric
2613 value is used in comparisons.
2616 Frequency : always with multiple stripes per file
2618 Description: Inefficient object allocation for mutli-stripe files
2619 Details : When selecting which OSTs to stripe files over, for files with
2620 a stripe count that divides evenly into the number of OSTs,
2621 the MDS is always picking the same starting OST for each file.
2622 Return the OST selection heuristic to the original design.
2627 Description: mount failures may take full timeout to return an error
2628 Details : Under some heavy load conditions it is possible that a
2629 failed mount can wait for the full obd_timeout interval,
2630 possibly several minutes, before reporting an error.
2631 Instead return an error as soon as the status is known.
2633 Frequency : quota enabled and large files being deleted
2635 Description: releasing more than 4GB of quota at once hangs OST
2636 Details : If a user deletes more than 4GB of files on a single OST it
2637 will cause the OST to spin in an infinite loop. Release
2638 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2643 Description: statfs data retrieved from /proc may be stale or zero
2644 Details : When reading per-device statfs data from /proc, in the
2645 {kbytes,files}_{total,free,avail} files, it may appear
2646 as zero or be out of date.
2649 Frequency : systems with MD RAID1 external journal devices
2651 Description: lconf's call to blkid is confused by RAID1 journal devices
2652 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2653 of returning all block devices that match the journal UUID.
2656 Frequency : always, for aggregate stripe size over 4GB
2658 Description: assertion fails when trying to use 4GB stripe size
2659 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2660 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2663 Frequency : always on ppc64
2665 Description: the first write on an ext3 filesystem with mballoc got stuck
2666 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2667 perform endianness conversion.
2669 ------------------------------------------------------------------------------
2671 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2673 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2674 INTEROPERATE with older versions automatically. Please read the
2675 user documentation before upgrading any part of a live system.
2676 * WARNING: Lustre networking configuration changes are required with
2677 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2680 * Support for kernels:
2681 2.6.9-22.0.2.EL (RHEL 4)
2682 2.6.5-7.244 (SLES 9)
2683 2.6.12.6 vanilla (kernel.org)
2686 Severity : enhancement
2687 Bugzilla : 7981/8208
2688 Description: Introduced Lustre Networking (LNET)
2689 Details : LNET is new networking infrastructure for Lustre, it includes
2690 a reorganized network configuration mode (see the user
2691 documentation for full details) as well as support for routing
2692 between different network fabrics. Lustre Networking Devices
2693 (LNDs) for the supported network fabrics have also been
2694 created for this new infrastructure.
2696 Severity : enhancement
2697 Description: Introduced Access control lists
2698 Details : clients can set ACLs on files and directories in order to have
2699 more fine-grained permissions than the standard Unix UGO+RWX.
2700 The MDS must be started with the "-o acl" mount option.
2702 Severity : enhancement
2703 Description: Introduced filesystem quotas
2704 Details : Administrators may now establish per-user quotas on the
2707 Severity : enhancement
2709 Description: Configuration change for the XT3
2710 The PTLLND is now used to run Lustre over Portals on the XT3
2711 The configure option(s) --with-cray-portals are no longer used.
2712 Rather --with-portals=<path-to-portals-includes> is used to
2713 enable building on the XT3. In addition to enable XT3 specific
2714 features the option --enable-cray-xt3 must be used.
2719 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2720 Details : A race between allocating a new llcd and re-getting the llcd_lock
2721 allowed another thread to grab newly-allocated llcd.
2723 Severity : enhancement
2725 Description: 2.6 OST async journal commit and locking fix to improve performance
2726 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2727 2.6 kernels are now async as they already were in 2.4 kernels so
2728 that they can commit concurrently with the network bulk transfer.
2729 For block-allocated files the filter allocation semaphore is held
2730 to avoid filesystem fragmentation during allocation. BKL lock
2731 removed for 2.6 xattr operations where it is no longer needed.
2736 Description: lconf incorrectly determined whether two IP networks could talk
2737 Details : In some more complicated routing and multiple-network
2738 configurations, lconf will avoid trying to make a network
2739 connection to a disjoint part of the IP space. It was doing the
2740 math incorrectly for one set of cases.
2745 Description: Fix for potential infinite loop processing records in an llog.
2746 Details : If an llog record is corrupted/zeroed, it is possible to loop
2747 forever in llog_process(). Validate the llog record length
2748 and skip the remainder of the block on error.
2751 Frequency : occasional (liblustre only)
2753 Description: liblustre could not open files whose last component is a symlink
2754 Details : sysio_path_walk() would incorrectly pass the open intent to
2755 intermediate path components.
2758 Frequency : rare (liblustre only with non-standard tuning)
2759 Bugzilla : 7201 (7350)
2760 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2761 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2762 after it has been released, only do this for referenced locks
2764 Severity : enhancement
2766 Description: specifying an (invalid) directory default stripe_size of -1
2767 would reset the directory default striping
2768 Details : stripe_size -1 was used internally to signal directory stripe
2769 removal, now use "all default" to signal dir stripe removal
2770 as a directory striping of "all default" is not useful
2773 Frequency : common for large clusters running liblustre clients
2775 Description: doing an ls when liblustre clients are running is slow
2776 Details : sending a glimpse AST to a liblustre client waits for every AST
2777 to time out, as liblustre clients will not respond. Since they
2778 cannot cache data we refresh the OST lock LVB from disk instead.
2780 Severity : enhancement
2782 Description: doing an ls at the same time as file IO can be slow
2783 Details : enqueue and other "small" requests can be blocked behind many
2784 large IO requests. Create a new OST IO portal for non-IO
2785 requests so they can be processed faster.
2788 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2790 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2791 created file with mode 000
2792 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2793 allow an ftruncate() as a non-root user to a file with mode 000.
2794 Lustre now respects this flag to disable mode checks when
2795 truncating a file owned by the user
2798 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2801 Description: Revoking locks from clients that went dead or catatonic might take
2803 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2804 cancellation of such locks instant on servers without waiting for
2805 any reply from clients. Clients drops these locks when cancel
2806 notification from server is received without replying.
2809 Frequency : liblustre-only, when liblustre client dies or becomes busy
2811 Description: Doing ls on Linux clients can take a long time with active
2813 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2814 granting such locks to it in the first place if possible. Locks
2815 are taken by proxy on the OST during the read or write and
2816 dropped immediately afterward. Add connect flags handling, do
2817 not grant locks to liblustre clients for glimpse ASTs.
2819 Severity : enhancement
2821 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2822 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2823 sequential reads and behaves badly with applications doing random
2824 reads. Improve it by reading ahead at least read region, and
2825 avoiding excessive large RPC for small reads.
2827 Severity : enhancement
2829 Description: Creating more than 1000 files for a single job may cause a load
2830 imbalance on the OSTs if there are also a large number of OSTs.
2831 Details : qos_prep_create() uses an OST index reseed value that is an
2832 even multiple of the number of available OSTs so that if the
2833 reseed happens in the middle of the object allocation it will
2834 still utilize the OSTs as uniformly as possible.
2839 Description: OST or MDS may oops in ping_evictor_main()
2840 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2841 but doesn't restart at beginning of obd_exports_timed list
2844 Severity : enhancement
2846 Description: improve by-nid export eviction on the MDS and OST
2847 Details : allow multiple exports with the same NID to be evicted at one
2848 time without re-searching the exports list.
2851 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2853 Description: MDS may oops in groups_free()
2854 Details : in rare race conditions a newly allocated group_info struct is
2855 freed again, and this can be NULL. The 2.4 compatibility code
2856 for groups_free() checked for a NULL pointer, but 2.6 did not.
2859 Frequency : common for liblustre clients doing little filesystem IO
2860 Bugzilla : 9352, 7313
2861 Description: server may evict liblustre clients accessing contended locks
2862 Details : if a client is granted a lock or receives a completion AST
2863 with a blocking AST already set it would not reply to the AST
2864 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2867 Frequency : lfs setstripe, only systems with more than 160 OSTs
2869 Description: unable to set striping with a starting offset beyond OST 160
2870 Details : llapi_create_file() incorrectly limited the starting stripe
2871 index to the maximum single-file stripe count.
2874 Frequency : LDAP users only
2876 Description: lconf did not handle in-kernel recovery with LDAP properly
2877 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2879 Severity : enhancement
2881 Description: bind OST threads to NUMA nodes to improve performance
2882 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2883 node and do their allocations there to localize memory access
2885 Severity : enhancement
2887 Description: llmount can determine client NID directly from Myrinet (GM)
2888 Details : the client NID code from gmnalnid was moved directly into
2889 llmount, removing the need to use this or specifying the
2890 client NID explicitly when mounting GM clients with zeroconf
2893 Frequency : if client is started with down MDS
2895 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2896 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2897 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2898 ptlrpc_import_delayed_req() because we didn't check if the
2899 request was interrupted, and we also didn't break out of the
2900 event loop if there was no timeout
2905 Description: data loss during non-page-aligned writes to a single file from
2906 both multiple nodes and multiple threads on one node at same time
2907 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2908 inconsistency led to false short-reads, that were cached and later
2909 used by ->prepare_write() to fill in partially written page,
2910 leading to data loss.
2913 Frequency : always, if lconf --abort_recovery used
2915 Description: lconf --abort_recovery fails with 'Operation not supported'
2916 Details : lconf was attempting to abort recovery on the MDT device and not
2919 Severity : enhancement
2921 Description: remove cleanup logs
2922 Details : replace lconf-generated cleanup logs with lustre internal
2923 cleanup routines. Eliminates the need for client-cleanup and
2926 Severity : enhancement
2928 Description: add support for EAs (user and system) on lustre filesystems
2929 Details : it is now possible to store extended attributes in the Lustre
2930 client filesystem, and with the user_xattr mount option it
2931 is possible to allow users to store EAs on their files also
2933 Severity : enhancement
2935 Description: Add possibility (config option) to show minimal available OST free
2937 Details : When compiled with --enable-mindf configure option, statfs(2)
2938 (and so, df) will return least minimal free space available from
2939 all OSTs as amount of free space on FS, instead of summary of
2940 free spaces of all OSTs.
2942 Severity : enhancement
2944 Description: do not expand extent locks acquired on OST-side
2945 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2946 by server: they are not cached anyway.
2949 Frequency : when mmap is used/binaries executed from Lustre
2951 Description: Unmmap pages before throwing them away from read cache.
2952 Details : llap_shrink cache now attempts to unmap pages before discarding
2953 them (if unmapping failed - do not discard). SLES9 kernel has
2954 extra checks that trigger if this unmapping is not done first.
2959 Description: lconf didn't resolve symlinks before checking to see whether a
2960 given mountpoint was already in use
2963 Frequency : when migrating failover services
2964 Bugzilla : 6395, 9514
2965 Description: When migrating a subset of services from a node (e.g. failback
2966 from a failover service node) the remaining services would
2967 time out and evict clients.
2968 Details : lconf --force (implied by --failover) sets the global obd_timeout
2969 to 5 seconds in order to quickly disconnect, but this caused
2970 other RPCs to time out too quickly. Do not change the global
2971 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2973 Severity : enhancement
2974 Frequency : if MDS is started with down OST
2975 Bugzilla : 9439,5706
2976 Description: Allow startup/shutdown of an MDS without depending on the
2977 availability of the OSTs.
2978 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2979 Add appropriate locking and lov-osc refcounts for safe
2980 cleaning. Add osc abort_inflight calls in case the
2981 synchronize never started.
2984 Frequency : occasional (Cray XT3 only)
2986 Description: root not authorized to access files in CRAY_PORTALS environment
2987 Details : The client process capabilities were not honoured on the MDS in
2988 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2989 been accessed by an authorized user then root was able to access
2990 the file on the local client also. The root user capabilities
2991 are now allowed on the MDS, as this environment has secure UID.
2994 Frequency : occasional
2996 Description: ldiskfs "too long searching" message happens too often
2997 Details : A debugging message (otherwise harmless) prints too often on
2998 the OST console. This has been reduced to only happen when
2999 there are fragmentation problems on the filesystem.
3004 Description: Division by zero in statfs when all OSCs are inactive
3005 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3006 lov_statfs divides by value returned by lov_get_stripecnt().
3010 Bugzilla : 9489, 3273
3011 Description: First write from each client to each OST was only 4kB in size,
3012 to initialize client writeback cache, which caused sub-optimal
3013 RPCs and poor layout on disk for the first writen file.
3014 Details : Clients now request an initial cache grant at (re)connect time
3015 and so that they can start streaming writes to the cache right
3016 away and always do full-sized RPCs if there is enough data.
3017 If the OST is rebooted the client also re-establishes its grant
3018 so that client cached writes will be honoured under the grant.
3023 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3024 Details : Now I/O RPCs go to different portal number and (presumably) fast
3025 lock requests (and glimses) and other RPCs get their own service
3026 threads pool that should be able to service those RPCs
3029 Severity : enhancement
3031 Description: Ability to exchange lustre version between client and servers and
3032 issue warnings at client side if client is too old. Also for
3033 liblustre clients there is ability to refuse connection of too old
3035 Details : New 'version' field is added to connect data structure that is
3036 filled with version info. That info is later checked by server and
3040 Frequency : rare, liblustre only.
3041 Bugzilla : 9296, 9581
3042 Description: Two simultaneous writes from liblustre at offset within same page
3043 might proceed at the same time overwriting eachother with stale
3045 Details : I/O lock withing llu_file_prwv was released too early, before data
3046 actually was hitting the wire. Extended lock-holding time until
3047 server acknowledges receiving data.
3050 Frequency : extremely rare. Never observed in practice.
3052 Description: avoid generating lustre_handle cookie of 0.
3053 Details : class_handle_hash() generates handle cookies by incrementing
3054 global counter, and can hit 0 occasionaly (this is unlikely, but
3055 not impossible, because initial value of cookie counter is
3056 selected randonly). Value of 0 is used as a sentinel meaning
3057 "unassigned handle" --- avoid it. Also coalesce two critical
3058 sections in this function into one.
3060 Severity : enhancement
3062 Description: allow liblustre clients to delegate truncate locking to OST
3063 Details : To avoid overhead of locking, liblustre client instructs OST to
3064 take extent lock in ost_punch() on client's behalf. New connection
3065 flag is added to handle backward compatibility.
3067 Severity : enhancement
3068 Bugzilla : 4928, 7341, 9758
3069 Description: allow number of OST service threads to be specified
3070 Details : a module parameter allows the number of OST service threads
3071 to be specified via "options ost ost_num_threads={N}" in the
3072 OSS's /etc/modules.conf or /etc/modprobe.conf.
3076 Bugzilla : 6146, 9635, 9895
3077 Description: servers crash with bad pointer in target_handle_connect()
3078 Details : In rare cases when a client is reconnecting it was possible that
3079 the connection request was the last reference for that export.
3080 We would temporarily drop the export reference and get a new
3081 one, but this may have been the last reference and the export
3082 was just destroyed. Get new reference before dropping old one.
3084 Severity : enhancement
3085 Frequency : if client is started with failover MDS
3087 Description: Allow multiple MDS hostnames in the mount command
3088 Details : Try to read the configuration from all specified MDS
3089 hostnames during a client mount in case the "primary"
3092 Severity : enhancement
3094 Description: Stop sending data to evicted clients as soon as possible.
3095 Details : Check if the client we are about to send or are sending data to
3096 was evicted already. (Check is done every second of waiting,
3097 for which l_wait_event interface was extended to allow checking
3098 of exit condition at specified intervals).
3101 Frequency : rare, normally only when NFS exporting is done from client
3103 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3105 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3108 Frequency : always, if srand() is called before liblustre initialization
3110 Description: Liblustre uses system PRNG disturbing its usage by user application
3111 Details : Introduce internal to lustre fast and high-quality PRNG for
3112 lustre usage and make liblustre and some other places in generic
3113 lustre code to use it.
3115 Severity : enhancement
3116 Bugzilla : 9477, 9557, 9870
3117 Description: Verify that the MDS configuration logs are updated when xml is
3118 Details : Check if the .xml configuration logs are newer than the config
3119 logs stored on the MDS and report an error if this is the case.
3120 Request --write-conf, or allow starting with --old_conf.
3122 Severity : enhancement
3124 Description: Handle symlinks in the path when checking if Lustre is mounted.
3125 Details : Resolve intermediate symlinks when checking if a client has
3126 mounted a filesystem to avoid duplicate client mounts.
3131 Description: lconf can hit an error exception but still return success.
3132 Details : The lconf command catches the Command error exception at the top
3133 level script context and will exit with the associated exit
3134 status, but doesn't ensure that this exit status is non-zero.
3139 Description: failure of ptlrpc thread startup can cause oops
3140 Details : Starting a ptlrpc service thread can fail if there are a large
3141 number of threads or the server memory is very fragmented.
3142 Handle this without oopsing.
3145 Frequency : always, only if liblustre and non-default acceptor port was used
3147 Description: liblustre cannot connect to servers with non-default acceptor port
3148 Details : tcpnal_set_default_params() was not called and was therefore
3149 ignoring the environment varaible TCPNAL_PORT, as well as other
3150 TCPNAL_ environment variables
3155 Description: two objects could be created on the same OST for a single file
3156 Details : If an OST is down, in some cases it was possible to create two
3157 objects on a single OST for a single file. No problems other
3158 than potential performance impact and spurious error messages.
3162 Bugzilla : 5681, 9562
3163 Description: Client may oops in ll_unhash_aliases
3164 Details : Client dcache may become inconsistent in race condition.
3165 In some cases "getcwd" can fail if the current directory is
3171 Description: Inode refcounting problems in NFS export code
3172 Details : link_raw functions used to call d_instantiate without obtaining
3173 extra inode reference first.
3177 Bugzilla : 9942, 9903
3178 Description: Referencing freed requests leading to crash, memleaks with NFS.
3179 Details : We used to require that call to ll_revalidate_it was always
3180 followed by ll_lookup_it. Also with revalidate_special() it is
3181 possible to call ll_revalidate_it() twice for the same dentry
3182 even if first occurence returned success. This fix changes semantic
3183 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3184 reference on a request referred from the intent.
3185 ll_intent_release() then releases such a request.
3188 Frequency : rare, normally benchmark loads only
3190 Description: unlinked inodes were kept in memory on the client
3191 Details : If a client is repeatedly creating and unlinking files it
3192 can accumulate a lot of stale inodes in the inode slab cache.
3193 If there is no other client load running this can cause the
3194 client node to run out of memory. Instead flush old inodes
3195 from client cache that have the same inode number as a new inode.
3198 Frequency : SLES9 2.6.5 kernel and long filenames only
3199 Bugzilla : 9969, 10379
3200 Description: utime reports stale NFS file handle
3201 Details : SLES9 uses out-of-dentry names in some cases, which confused
3202 the lustre dentry revalidation. Change it to always use the
3206 Frequency : rare, unless heavy write-truncate concurrency is continuous
3207 Bugzilla : 4180, 6984, 7171, 9963, 9331
3208 Description: OST becomes very slow and/or deadlocked during object unlink
3209 Details : filter_destroy() was holding onto the parent directory lock
3210 while truncating+unlinking objects. For very large objects this
3211 may block other threads for a long time and slow overall OST
3212 responsiveness. It may also be possible to get a lock ordering
3213 deadlock in this case, or run out of journal credits because of
3214 the combined truncate+unlink. Solution is to do object truncate
3215 first in one transaction without parent lock, and then do the
3216 final unlink in a new transaction with the parent lock. This
3217 reduces the lock hold time dramatically.
3220 Frequency : rare, 2.4 kernels only
3222 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3223 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3224 before unmounting their filesystem in order to avoid deadlock.
3225 The kernel_locked() function in 2.4 kernels only checks whether
3226 the kernel lock is held, not whether it is this process that is
3227 holding it as 2.6 kernels do.
3232 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3233 Details : The client ptlrpc code may be trying to reconnect to a down
3234 server before a previous connection attempt has timed out.
3235 Increase the reconnect interval to be longer than the connection
3236 timeout interval to avoid sending duplicate connections to
3240 Frequency : echo_client brw_test command
3242 Description: fix echo_client to work with OST preallocated code
3243 Details : OST preallocation code (5137) didn't take echo_client IO path
3244 into account: echo_client calls filter methods outside of any
3245 OST thread and, hence, there is no per-thread preallocated
3246 pages and buffers to use. Solution: hijack pga pages for IO. As
3247 a byproduct, this avoids unnecessary data copying.
3251 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3252 Description: Client can oops in mdc_commit_close() after open replay
3253 Details : It was possible for the MDS to return an open request with no
3254 transaction number in mds_finish_transno() if the client was
3255 evicted, but without actually returning an error. Clients
3256 would later try to replay that open and may trip an assertion
3257 Simplify the client close codepath, and always return an error
3258 from the MDS in case the open is not successful.
3261 Frequency : rare, 2.6 OSTs only
3263 Description: OST may deadlock under high load on fragmented files
3264 Details : If there was a heavy load and highly-fragmented OST filesystems
3265 it was possible to have all the OST threads deadlock waiting on
3266 allocation of biovecs, because the biovecs were not released
3267 until the entire RPC IO was completed. Instead, release biovecs
3268 as soon as they are complete to ensure forward IO progress.
3270 Severity : enhancement
3272 Description: Support for specifying external journal device at mount
3273 Details : If an OST or MDS device is formatted with an external journal
3274 device, this device major/minor is stored in the ext3 superblock
3275 and may not be valid for failover. Allow detecting and
3276 specifying the external journal at mount time.
3281 Description: Mounting an MDS with pending unlinked files may cause oops
3282 Details : target_finish_recovery() calls mds_postrecov() which returned
3283 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3284 considers this an error and immediately begins cleaning up the
3285 lov, just after starting the mds_lov process
3287 Severity : enhancement
3289 Description: Implement 'lfs df' to report actual free space on per-OST basis
3290 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3291 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3292 usage of objects; '-h' to report in human readable format.
3294 ------------------------------------------------------------------------------
3296 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3303 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3304 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3305 at startup during orphan unlinking. Instead, skip buddy update
3306 before setup, as it will be regenerated after recovery is complete.
3311 Description: If an OST is inactive, its locks might reference stale inodes.
3312 Details : lov_change_cbdata() must iterate over all namespaces, even if
3313 they are inactive to clear inode references from the lock.
3315 Severity : enhancement
3316 Frequency : occasional, if non-standard max_dirty_mb used
3318 Description: Client will block write RPCs if not enough grant
3319 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3320 then the client will block writes while waiting for another RPC
3321 to complete instead of consuming its dirty limit. With change
3322 we get improved performance when max_dirty_mb is small.
3324 Severity : enhancement
3325 Bugzilla : 3389, 6253
3326 Description: Add support for supplementary groups on the MDS.
3327 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3328 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3329 lookups for user supplementary groups into a cache.
3333 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3334 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3335 for liblustre O_EXCL is enforced by MDS.
3339 Description: readdir never returns NULL in liblustre.
3340 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3341 dentry in struct dirent.
3345 Frequency : liblustre only, depends on application IO pattern
3346 Description: liblustre clients evicted if not contacting servers
3347 Details : Don't put liblustre clients into the ping_evictor list, so
3348 they will not be evicted by the pinger ever.
3350 Severity : enhancement
3352 Description: Add ability to evict clients by NID from MDS.
3353 Details : By echoing "nid:$NID" string into
3354 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3355 $NID would be instantly evicted from this MDS and from all active
3356 OSTs connected to it.
3360 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3361 Details : lookup_it_finish() used to query file size from OSTs that was not
3366 Description: service threads change working directory to that of init
3367 Details : Starting lustre service threads may pin the working directory
3368 of the parent thread, making that filesystem busy. Threads
3369 now change to the working directory of init to avoid this.
3373 Frequency : during shutdown only
3374 Description: shutdown with a failed MDS or OST can cause unmount to hang
3375 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3380 Frequency : 2.6 only
3381 Description: chmod/chown may include an extra supplementary group
3382 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3383 supplementary group and if none is specified this is used.
3386 Bugzilla : 5479 (6816)
3388 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3389 Details : If lookup is in progress on a file that is unlinked we might try
3390 to revalidate the inode and fail in revalidate after lookup is
3391 complete and ll_file_open() enqueues the open again but
3392 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3395 Frequency : always, if lconf --abort_recovery used
3397 Description: lconf --abort_recovery fails with 'Operation not supported'
3398 Details : lconf was attempting to abort recovery on the MDT device and not
3401 ------------------------------------------------------------------------------
3403 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3408 Frequency : rare (only unsupported configurations with a node running as an
3410 Bugzilla : 6514, 5137
3411 Description: Mounting a Lustre file system on a node running as an OST could
3413 Details : OSTs now preallocates memory needed to write out data at
3414 startup, instead of when needed, to avoid having to
3415 allocate memory in possibly low memory situations.
3416 Specifically, if the file system is mounted on on OST,
3417 memory pressure could force it to try to write out data,
3418 which it needed to allocate memory to do. Due to the low
3419 memory, it would be unable to do so and the node would
3420 become unresponsive.
3422 Severity : enhancement
3424 Description: Addition of lconf --service command line option
3425 Details : lconf now accepts a '--service <arg>' option, which is
3426 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3428 Severity : enhancement
3430 Description: Failover mode is now the default for OSTs.
3431 Details : By default, OSTs will now run in failover mode. To return to
3432 the old behaviour, add '--failout' to the lmc line for OSTs.
3434 Severity : enhancement
3436 Description: Health checks are now provided for MDS and OSTs
3437 Details : Additional detailed health check information on MSD and OSTs
3438 is now provided through the procfs health_check value.
3441 Frequency : occasional, depends on IO load
3443 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3444 after numerous create/delete cycles
3445 Details : The ext3 inode allocation policy would not allocate new inodes
3446 very well on the OSTs because there are no new directories
3447 being created. Instead we look for groups with free space if
3448 the parent directories are nearly full.
3453 Description: Network or server problems during mount may cause partially
3454 mounted clients instead of returning an error.
3455 Details : The config llog parsing code may overwrite the error return
3456 code during mount error handling, returning success instead
3462 Description: MDS can fail to allocate large reply buffers
3463 Details : After long uptimes the MDS can fail to allocate large reply
3464 buffers (e.g. zconf client mount config records) due to memory
3465 fragmentation or consumption by the buffer cache. Preallocate
3466 some large reply buffers so that these replies can be sent even
3467 under memory pressure.
3471 Frequency : rare (liblustre)
3472 Description: fsx running with liblustre complained that using truncate() to
3473 extend the file doesn't work. This patch corrects that issue.
3474 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3475 ATTR_SIZE and lsm use in llu_setattr_raw.
3479 Frequency : rare, only 2.6 kernels
3480 Description: Unusual file access patterns on the MDS may result in inode
3481 data being lost in very rare circumstances.
3482 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3483 code in the 2.6 kernel caused the inode and/or EA data not to
3484 be read from disk, causing single-file corruption.
3488 Frequency : rare, only 2.6 filesystems using extents
3489 Description: Heavy concurrent write and delete load may cause data corruption.
3490 Details : It was possible under high-load situations to have an extent
3491 metadata block in the block device cache from a just-unlinked
3492 file overwrite a newly-allocated data block. We now unmap any
3493 metadata buffers that alias just-allocated data blocks.
3497 Frequency : filesystems with default stripe_count larger than 77
3498 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3499 Details : lconf specifies an inode size of 4096 bytes when the default
3500 stripe_count is larger than 77. This conflicts with the default
3501 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3502 this case to avoid pinning too much memory for large EAs.
3504 ------------------------------------------------------------------------------
3506 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3511 Frequency : rare (extremely heavy IO load with hundreds of clients)
3513 Description: Client is evicted, gets IO error writing to file
3514 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3515 caused clients to be evicted for AST timeouts. The fixes in
3516 bug 5192 mean we no longer need to have such short AST timeouts
3517 so ldlm_timeout has been increased.
3520 Frequency : occasional during --force or --failover shutdown under load
3521 Bugzilla : 5949, 4834
3522 Description: Server oops/LBUG if stopped with --force or --failover under load
3523 Details : a collection of import/export refcount and cleanup ordering
3524 issues fixed for safer force cleanup
3527 Frequency : only filesystems larger than 120 OSTs
3528 Bugzilla : 5990, 6223
3529 Description: lfs getstripe would oops on a very large filesystem
3530 Details : lov_getconfig used kfree on vmalloc'd memory
3533 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3534 Bugzilla : 6242, 6243
3535 Description: reading from files that had been truncated to a non-zero size
3536 but never opened returned no data
3537 Details : ll_file_read() reads zeros from no-object files to EOF
3542 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3543 Details : There's little harm in aborting MDS/OSS recovery and letting it
3544 try again, so I removed the LASSERT and return an error instead.
3546 Severity : enhancement
3548 Description: New debugging infrastructure for tracking down data corruption
3549 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3550 (b) cover more of the client-side code path, and (c) try to narrow
3551 down where problems occurred
3555 Bugzilla : 3819, 4364, 4397, 6313
3556 Description: Racing close and eviction MDS could cause assertion in mds_close
3557 Details : It was possible to get multiple mfd references during close and
3558 client eviction, leading to one thread referencing a freed mfd.
3560 Severity: : enhancement
3561 Bugzilla : 3262, 6359
3562 Description: Attempts to reconnect to servers are now more aggressive.
3563 Details : This builds on the enhanced upcall-less recovery that was added
3564 in 1.4.2. When trying to reconnect to servers, clients will
3565 now try each server in the failover group every 10 seconds. By
3566 default, clients would previously try one server every 25 seconds.
3571 Description: After recovery, certain operations trigger a failed
3572 assertion on a client.
3573 Details : Failing over an mds, using lconf -d --failover, while a
3574 client was doing a readdir() call would cause the client to
3575 LBUG after recovery completed and the readdir() was resent.
3577 Severity : enhancement
3579 Description: Default groups are now added by lconf
3580 Details : You can now run lconf --group <servicename> without having to
3581 manually add groups with lmc.
3584 Frequency : occasional
3586 Description: Nodes with an elan id of 0 trigger a failed assertion
3589 Frequency : always when accessing e.g. tty/console device nodes
3591 Description: tty and some other devices nodes cannot be used on lustre
3592 Details : file's private_data field is used by device data and lustre
3593 values in there got lost. New field was added to struct file to
3594 store fs-specific private data.
3597 Frequency : when exporting Lustre via NFS
3599 Description: NFSD failed occasionally when looking up a path component
3600 Details : NFSD is looking up ".." which was broken in ext3 directories
3601 that had grown large enough to become hashed.
3604 Frequency : Clusters with multiple interfaces not on the same subnet
3606 Description: Nodes will repeatedly try to reconnect to an interface which it
3607 cannot reach and report an error to the log.
3608 Details : Extra peer list entries will be created by lconf with some peers
3609 unreachable. lconf now validates the peer before adding it.
3612 Frequency : Only if a default stripe is set on the filesystem root.
3614 Description: Setting a default stripe on the filesystem root prevented the
3615 filesystem from being remounted.
3616 Details : The client was sending extra request flags in the root getattr
3617 request and did not allocate a reply buffer for the dir EA.
3620 Frequency : occasional, higher if lots of files are accessed by one client
3621 Bugzilla : 6159, 6097
3622 Description: Client trips assertion regarding lsm mismatch/magic
3623 Details : While revalidating inodes the VFS looks up inodes with ifind()
3624 and in rare cases can find an inode that is being freed.
3625 The ll_test_inode() code will free the lsm during ifind()
3626 when it finds an existing inode and then the VFS later attaches
3627 this free lsm to a new inode.
3631 Bugzilla : 6422, 7030
3632 Description: MDS deadlock between mkdir and client eviction
3633 Details : Creating a new file via mkdir or mknod (starting a transaction
3634 and getting the ns lock) can deadlock with client eviction
3635 (gets ns lock and trying to finish a synchronous transaction).
3638 Frequency : occasional
3639 Description: While starting a server, the fsfilt_ext3 module could not be
3641 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3642 kernels. Previously, lconf would still use the ext3 name
3643 when trying to load modules. Now, it will correctly use
3644 ext3 on 2.4 and ldiskfs on 2.6.
3646 Severity : enhancement
3647 Description: The default stripe count has been changed to 1
3648 Details : The interpretation of the default stripe count (0, to lfs
3649 or lmc) has been changed to mean striping across a single
3650 OST, rather than all available. For general usage we have
3651 found a stripe count of 1 or 2 works best.
3653 Severity : enhancement
3654 Description: Add support for compiling against Cray portals.
3655 Details : Conditional compiling for some areas that are different
3659 Frequency : occasional
3660 Bugzilla : 6409, 6834
3661 Description: Creating files with an explicit stripe count may lead to
3662 a failed assertion on the MDS
3663 Details : If some OSTs are full or unavailable, creating files may
3664 trigger a failed assertion on the MDS. Now, Lustre will
3665 try to use other servers or return an error to the
3669 Frequency : occasional
3671 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3672 caused lock timeout during readahead (since 1.4.2).
3673 Details : Processes doing readahead might match a lock that hasn't been
3674 granted yet if there are overlapping and conflicting lock
3675 requests. The readahead process waits on ungranted lock
3676 (original lock is CBPENDING), while OST waits for that process
3677 to cancel CBPENDING read lock and eventually evicts client.
3679 Severity : enhancement
3681 Description: Initial enabling of flock support for clients
3682 Details : Implements fcntl advisory locking and file status functions.
3683 This feature is provided as an optional mount flag (default
3684 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3685 locking are implemented yet, and those that are are not guaranteed
3686 to be completely correct in production environments.
3687 mount -t lustre -o [flock|noflock] ...
3690 Frequency : occasional
3692 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3693 assertion in the ext3 JBD (journaling) layer.
3694 Details : The b_committed_data struct is protected by the big kernel lock
3695 in 2.4 kernels, serializing journal_commit_transaction() and
3696 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3697 access to this struct. In 2.6 kernels there is finer grained
3698 locking to improve SMP performance of the JBD layer.
3702 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3707 Description: merge and cleanup kernel patches.
3710 -----------------------------------------------------------------------------
3712 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3714 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3715 versions of Lustre. Please update all servers and clients to
3716 version 1.4.2 or later at the same time. You must also run
3717 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3718 to update the configuration logs.
3720 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3721 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3722 - swab llog records of type '0' so we get proper header size/idx (5861)
3723 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3724 - fix rename of one directory over another leaking an inode (5953)
3725 - avoid SetPageDirty on 2.6 (5981)
3726 - don't re-add just-being-destroyed locks to the waiting list (5653)
3727 - when creating new directories, inherit the parent's custom
3728 striping settings if present parent (3048)
3729 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3730 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3731 - don't hold client locks on temporary worklist from l_lru (5666)
3732 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3733 - automatically evict dead clients (5921)
3734 - Update file size properly in create+truncate+fstat case (6196)
3735 - Do not unhash mountpoint dentries, do not allow removal of
3737 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3738 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3739 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3740 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3741 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3742 - Correctly handle reads of files with no objects (6243)
3743 - lctl recover will also mark a device active if deactivate used (5933)
3745 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3746 - allow --write-conf on an MDS with different nettype than client (5619)
3747 - don't write config llogs to MDS for mounts not from that MDS (5617)
3748 - lconf should create multiple TCP connections from a client (5201)
3749 - init scripts are now turned off by default; run chkconfig --on
3750 lustre and chkconfig --on lustrefs to use them
3751 - upcalls are no longer needed for clients to recover to failover
3753 - add --abort-recovery option to lconf to abort recovery on device
3755 - add support for an arbitrary number of OSTs (3026)
3756 - Quota support protocol changes.
3757 - forward compatibility changes to wire structs (6007)
3758 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3759 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3760 - improved "lustre status" script
3761 - initialize blocksize for non-regular files (6062)
3762 - added --disable-server and --disable-client configure options (5782)
3763 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3764 - Vanilla 2.4.29 support
3765 - increase maximum number of obd devices to 520 (6242)
3766 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3767 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3768 SLES 9 2.6.5 kernels
3769 - limit stripes per file to 160 (the maximum EA size) (6093)
3771 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3774 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3775 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3776 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3777 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3778 - small ext3 extents cleanups and fixes (5733)
3779 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3780 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3781 - avoid needless client->OST connect, fix handle mismatch (5317)
3782 - fix DLM error path that led to out-of-sync client, long delays (5779)
3783 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3784 - fix several locking issues related to i_size (5492,5624,5654,5672)
3785 - don't move pending lock onto export if it is already evicted (5683)
3786 - fix kernel oops when creating .foo in unlinked directory (5548)
3787 - fix deadlock in obdfilter statistics vs. object create (5811)
3788 - use time_{before,after} to avoid timer jiffies wrap (5882)
3789 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3790 - Do not leak request if server was not able to process it (5154)
3791 - If mds_open unable to find parent dir, make that negative lookup(5154)
3792 - don't create new directories with extent-mapping (5909, 5936)
3794 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3795 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3796 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3797 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3798 - support for RHEL4 kernel on i686 (5773)
3799 - provide error messages when incompatible logs are encountered (5898)
3801 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3802 * version 1.4.0.10 (1.4.1 release candidate 1)
3804 - don't keep a lock reference when lock is not granted (4238)
3805 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3806 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3807 - drop import inflight refcount on signal_completed_replay error (5255)
3808 - unlock page after async write error during send (3677)
3809 - handle missing objects in filter_preprw_read properly (5265)
3810 - no transno return for symlink open, don't save no-trasno open (3440)
3811 - don't try to complete elan receive that already failed (4012)
3812 - free RPC server reply state on error (5406)
3813 - clean up thread from ptlrpc_start_thread() on error (5160)
3814 - readahead could read extra page into cache that wasn't ejected (5388)
3815 - prevent races in class_attach/setup/cleanup/detach (5260)
3816 - don't dereference de->d_inode after l_dput of de (5458)
3817 - use "int" for stripe value returned from lock_to_stripe (5544)
3818 - mballoc allocation and error-checking fixes in 2.6 (5504)
3819 - block device patches to fix I/O request sizes in 2.6 (5482)
3820 - look up hostnames for IB nals (5602)
3821 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3822 - don't start multiple acceptors for the same port (5277)
3823 - fix incorrect LASSERT in mds_getattr_name (5635)
3824 - export a proc file for general "ping" checking (5628)
3825 - fix "lfs check" to not block when the MDS is down (5628)
3827 - service request history (4965)
3828 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3829 - create an "evict_client" /proc entry on OSTs, like the MDS has
3830 - fix mount usage message, return errors per mount(8) (5168)
3831 - change grep [] to grep "[]" in tests so they work in more UMLs
3832 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3833 - remove ancient LOV_MAGIC_V0 EA support (5047)
3834 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3835 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3836 - allow readahead allocations to fail when low on memory (5383)
3837 - mmap locking landed again, after considerable improvement (2828)
3838 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3840 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3843 - send OST transaction number in read/write reply to free req (4966)
3844 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3845 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3846 - the watchdog thread now runs as interruptible (5246)
3847 - flock/lockf fixes (but it's still disabled, pending 5135)
3848 - don't use EXT3 constants in llite code (5094)
3849 - memory shortage at startup could cause assertion (5176)
3851 - reorganization of lov code
3852 - single portals codebase
3854 - add extents/mballoc support (5025)
3855 - direct I/O reads in the obdfilter (4048)
3856 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3858 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3861 - send OST transaction number in read/write reply to free req (4966)
3862 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3863 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3864 - the watchdog thread now runs as interruptible (5246)
3865 - handle missing objects in filter_preprw_read properly (5265)
3866 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3867 - drop import inflight refcount on signal_completed_replay error (5255)
3868 - unlock page after async write error during send (3677)
3869 - return original error code on reconstructed replies (3761)
3870 - no transno return for symlink open, don't save no-trasno open (3440)
3872 - add pid to ldlm debugging output (4922)
3873 - bump the watchdog timeouts -- we can't handle 30sec yet
3874 - extra debugging for orphan dentry/inode bug (5259)
3876 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3879 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3880 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3881 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3882 - fix NULL dereference in /proc/sys/portals/routes (4827)
3883 - allow failed mdc_close() operations to be interrupted (4561)
3884 - stop precreate on OST before MDS would time out on it (4778)
3885 - don't send partial-page writes before EOF from client (4410)
3886 - discard client grant for sub-page writes on large-page clients (4520)
3887 - don't free dentries not owned by NFS code, check generation (4806)
3888 - fix lsm leak if mds_create_objects() fails (4801)
3889 - limit debug_daemon file size, always print CERROR messages (4789)
3890 - use transno after validating reply (3892)
3891 - process timed out requests if import state changes (3754)
3892 - update mtime on OST during writes, return in glimpse (4829)
3893 - add mkfsoptions to LDAP (4679)
3894 - use ->max_readahead method instead of zapping global ra (5039)
3895 - don't interrupt __l_wait_event() during strace
3897 - add software watchdogs to catch hung threads quickly (4941)
3898 - make lustrefs init script start after nfs is mounted
3899 - fix CWARN/ERROR duplication (4930)
3900 - return async write errors to application if possible (2248)
3901 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3902 - print NAL number in %x format (4645)
3903 - update barely-supported suse-2.4.21-171 series (4842)
3904 - support for sles 9 %post scripts
3905 - support for building 2.6 kernel-source packages
3906 - support for sles km_* packages
3908 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3911 - ignore -ENOENT errors in osc_destroy (3639)
3912 - notify osc create thread that OSC is being cleaned up (4600)
3913 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3914 - reconstruct ost_handle() like mds_handle() (4657)
3915 - create a new thread to do import eviction to avoid deadlock (3969)
3916 - let lconf resolve symlinked-to devices (4629)
3917 - don't unlink "objects" from directory with default EA (4554)
3918 - hold socknal file ref over connect in case target is down (4394)
3919 - allow more than 32000 subdirectories in a single directory (3244)
3920 - fix blocks count for O_DIRECT writes (3751)
3921 - OST returns ENOSPC from object create when no space left (4539)
3922 - don't send truncate RPC if file size isn't changing (4410)
3923 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3924 - bind to privileged port in socknal and tcpnal (3689)
3926 - rate limit CERROR/CWARN console message to avoid overload (4519)
3927 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3928 - basic mmap support (3918)
3929 - kernel patch series update from b1_4 (4711)
3931 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3934 - avoid crash during MDS cleanup with OST shut down (2775)
3935 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3936 - don't use bad inodes on the MDS (3744)
3937 - dynamic object preallocation to improve recovery speed (4236)
3938 - don't hold spinlock over lock dumping or change debug flags (4401)
3939 - don't zero obd_dev when it is force cleaned (3651)
3940 - print grants to console if they go negative (4431)
3941 - "lctl deactivate" will stop automatic recovery attempts (3406)
3942 - look for existing locks in ldlm_handle_enqueue() (3764)
3943 - don't resolve lock handle twice in recovery avoiding race (4401)
3944 - revalidate should check working dir is a directory (4134)
3946 - don't always mark "slow" obdfilter messages as errors (4418)
3948 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3951 - don't close LustreDB during write_conf until it is done (3860)
3952 - fix typo in lconf for_each_profile (3821)
3953 - allow dumping logs from multiple threads at one time (3820)
3954 - don't allow multiple threads in OSC recovery (3812)
3955 - fix debug_size parameters (3864)
3956 - fix mds_postrecov to initialize import for llog ctxt (3121)
3957 - replace config semaphore with spinlock (3306)
3958 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3959 - don't allow enqueue to complete on a destroyed export (3822)
3960 - down write_lock before checking llog header bitmap (3825)
3961 - recover from lock replay timeout (3764)
3962 - up llog sem before sending rpc (3652)
3963 - reduce ns lock hold times when setting kms (3267)
3964 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3965 - fix NULL deref and obd_dev leak on setup error (3312)
3966 - replace some LBUG about llog ops with error handling (3841)
3967 - don't match INVALID dentries from d_lookup and spin (3784)
3968 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3969 - fix invalid assertion in ptlrpc_set_wait (3880)
3971 - add libwrap support for the TCP acceptor (3996)
3972 - add /proc/sys/portals/routes for non-root route listing (3994)
3973 - allow setting MDS UUID in .xml (2580)
3974 - print the stack of a process that LBUGs (4228)
3976 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3979 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3980 - make sure to unset replay flag from failed open requests (3440)
3981 - if default stripe count is 0, use OST count for inode size (3636)
3982 - update parent mtime/ctime on client for create/unlink (2611)
3983 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3984 - free recovery state on server during a forced cleanup (3571)
3985 - unregister_reply for resent reqs (3063)
3986 - loop back devices mounting and status check on 2.6 (3563)
3987 - fix resource-creation race that can provoke i_size == 0 (3513)
3988 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3989 - more debugging for page-accounting assertion (3746)
3990 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3991 - avoid deadlock after precreation failure (3758)
3992 - fix race and lock order deadlock in orphan handling (3450, 3750)
3993 - add validity checks when grabbing inodes from l_ast_data (3599)
3995 - add /proc/.../recovery_status to obdfilter (3428)
3996 - lightweight CDEBUG infrastructure, debug daemon (3668)
3997 - change default OSC RPC parameters to be better on small clusters
3998 - turn off OST read cache for files smaller than 32MB
3999 - install man pages and include them in rpms (3100)
4000 - add new init script for (un)mounting lustre filesystems (2593)
4001 - run chkconfig in %post for init scripts (3701)
4002 - drop scimac NAL (unmaintained)
4004 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4007 - clean kiobufs before and after use (3485)
4008 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4009 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4010 - change init script to fail more clearly if not run as root (1528)
4011 - allow clients to reconnect during replay (1742)
4012 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4013 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4014 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4015 - fix cond_resched (3554)
4016 - only evict unfinished clients after recovery (3515)
4017 - allow bulk resend, prevent data loss (3570)
4018 - dynamic ptlrpc request buffer allocation (2102)
4019 - don't allow unlinking open directory if it isn't empty (2904)
4020 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4021 - remove extraneous obd dereference causing LASSERT failure (3334)
4022 - don't use get_cycles() when creating temp. files on the mds (3156)
4023 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4024 - handle EEXIST for set-stripe, set proper directory name (3336)
4026 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4027 - fix ksocknal_fmb_callback() error messages (2918)
4029 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4032 - don't copy lvb into (possibly NULL) reply on error (2983)
4033 - don't deref dentry after dput, don't free lvb on error (2922)
4034 - use the kms to determine writeback rpc length (2947)
4035 - increment oti_logcookies when osc is inactive (2948)
4036 - update client's i_blocks count via lvb messages (2543)
4037 - handle intent open/close of special files properly (1557)
4038 - mount MDS with errors=remount-ro, like obdfilter (2009)
4039 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4040 - don't use cancelling-locks' kms values (2947)
4041 - use highest lock extent for kms, not last one (2925)
4042 - don't dereference ERR_PTR() dentry in error handling path (3107)
4043 - fix thread race in portals_debug_dumplog() (3122)
4044 - create lprocfs device entries at setup instead of at attach (1519)
4045 - common AST error handler, don't evict client on completion race (3145)
4046 - zero nameidata in detach_mnt in 2.6 (3118)
4047 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4048 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4049 - initialize RPC timeout timer earlier for 2.6 (3219)
4050 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4051 - print nal/nid for unknown nid (3258)
4052 - additional checks for oscc recovery before doing precreate (3284)
4053 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4054 - don't crash in mdc_close for bad permissions on open (3285)
4055 - zero i_rdev for non-device files (3147)
4056 - clear page->private before handing to FS, better assertion (3119)
4057 - tune the read pipeline (3236)
4058 - fix incorrect decref of invalidated dentry (2350)
4059 - provide read-ahead stats and refine rpc in flight stats (3328)
4060 - don't hold journal transaction open across create RPC (3313)
4061 - update atime on MDS at close time (3265)
4062 - close LDAP connection when recovering to avoid server load (3315)
4063 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4064 - don't leak open file on MDS after open resend (3325)
4065 - serialize filter_precreate and filter_destroy_precreated (3329)
4066 - loop device shouldn't call sync_dev() for nul device (3092)
4067 - clear page cache after eviction (2766)
4068 - resynchronize MDS->OST in background (2824)
4069 - refuse to mount the same filesystem twice on same mountpoint (3394)
4070 - allow llmount to create routes for mounting behind routers (3320)
4071 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4072 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4073 - fix rare problem with rename on htree directories (3417)
4075 - allow default OST striping configuration per directory (1414)
4076 - fix compilation for qswnal for 2.6 kernels (3125)
4077 - increase maximum number of MDS request buffers for large systems
4078 - change liblustreapi to be useful for external progs like lfsck (3098)
4079 - increase local configuration timeout for slow disks (3353)
4080 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4082 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4085 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4086 - don't overwrite extent policy data in reply if lock was blocked (2901)
4087 - drop filter export grants atomically with removal from device (2663)
4088 - del obd_self_export from work_list in class_disconnect_exports (2908)
4089 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4090 - swab reply message in mdc_close, other PPC fixes (2464)
4091 - fix destroying of named logs (2325)
4092 - overwrite old logs when running lconf --write_conf (2264)
4093 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4094 - fix race in target_handle_connect (2898)
4095 - mds_reint_create() should take same inode create lock (2926)
4096 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4097 - don't close files for self_export to avoid uninitialized obd (2936)
4098 - allow MDS with the same name as client node (2939)
4099 - hold dentry reference for closed log files for unlink (2325)
4100 - reserve space for all logs during transactions (2059)
4101 - don't evict page beyond end of stripe extent (2925)
4102 - don't oops on a deleted current working directory (2399)
4103 - handle hard links to targets without a parent properly (2517)
4104 - don't dereference NULL lock when racing during eviction (2867)
4105 - don't grow lock extents when lots of conflicting locks (2919)
4107 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4110 - account for cache space usage on clients to avoid data loss (974)
4111 - lfsck support in lustre kernel code (2349)
4112 - reduce journal credits needed for BRW writes (2370)
4113 - orphan handling to avoid losing space on client/server crashes
4114 - ptlrpcd can be blocked, stopping ALL progress (2477)
4115 - use lock value blocks to assist in proper KMS, faster stat (1021)
4116 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4117 - recovery for initial connections (2355)
4118 - fixes for mds_cleanup_orphans (1934)
4119 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4120 - block all file creations until orphan recovery completes (1901)
4121 - client remove rq_connection from request struct (2423)
4122 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4123 - recovery timer race (2670)
4124 - mdc_close recovey bug (2532)
4125 - ptlrpc cleanup bug (2710)
4126 - mds timeout on local locks (2588)
4127 - namespace lock held during RPCs (2431)
4128 - handle interrupted sync write properly (2503)
4129 - don't try to handle a message that hasn't been replied to (2699)
4130 - client assert failure during cleanup after abort recovery (2701)
4131 - leak mdc device after failed mount (2712)
4132 - ptlrpc_check_set allows timedout requests to complete (2714)
4133 - wait for inflight reqs when ptlrpcd finishes (2710)
4134 - make sure unregistered services are removed from the srv_list
4135 - reset bulk XID's when resending them (caught by 1138 test)
4136 - unregister_bulk after timeout
4137 - fix lconf error (2694)
4138 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4139 - readahead locks pages, leaves pending causing memory pressure (2673)
4140 - increase OST request buffers to 4096 on large machines (2729)
4141 - fix up permission of existing directories in simple_mkdir (2661)
4142 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4143 - don't assign transno to errored transactions (2742)
4144 - don't delete objects on OST if given a bogus objid from MDS (2751)
4145 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4146 - if rq_no_resend, then timeout request after recovery (2432)
4147 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4148 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4149 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4150 - MDS recovery shouldn't delete objects that it has given out (2730)
4151 - if enqueue arrives after completion, don't clobber LVB (2819)
4152 - don't unlock pages twice when trigger_group_io returns error (2814)
4153 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4154 - don't write pages to disk if there was an error (1450)
4155 - don't ping imports that have recovery disabled (2676)
4156 - take buffered bytes into account when balancing socknal conn (2817)
4157 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4158 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4159 - remove little-endian swabbing of llog records (1987)
4160 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4161 - retry reposting request buffers if they fail (1191)
4162 - grow extent at grant time to avoid granting a revoked lock (2809)
4163 - lock revoke doesn't evict page if covered by a second lock (2765)
4164 - disable VM readahead to avoid reading outside lock extents (2805)
4166 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4167 - updated kernel patches for hp-2.4.20 kernel (2681)
4169 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4172 - fix truncated write corruption (2366)
4173 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4175 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4176 - ensure len(uuid) < 37 in lmc (1171)
4177 - fix ia64 OOPS in llog_test (2255)
4178 - zero end of page at obdfilter for partial page writes (2648)
4179 - don't leave stale dentries around after renames (bug 2428)
4180 - fix timeouts when evicting a client with a single lock held (2642)
4181 - set deadline for the initial HELLO message to drain (2634)
4182 - print out dotted-quad IP addresses in the socknal (2302)
4184 - additional debugging for MDS client eviction problem (2443)
4185 - fix mkfsoptions support for osts (2603, 2604)
4187 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4190 - add series for the vanilla 2.6.0 kernel
4191 - add series for the vanilla 2.4.24 kernel
4192 - add series for a cray x86/64 UL kernel drop
4193 - fix xattr patches for the vanilla 2.4.19 series
4195 - generate true UUIDs in lmc (1171)
4196 - have portals stack dumping break in UML (2466)
4197 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4198 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4199 - wake callback waiting threads on client eviction (2460)
4200 - Add --ptldebug and --subsystem to lmc (1719)
4201 - update assertion to allow safe interrupt allocation
4202 - set rq_no_resend for cancel requests (2432)
4203 - recalculate ptlrpcd timeout after resend (2494)
4204 - call vfs_rmdir when removing pending directories (2368)
4205 - fix renaming a file to itself (2429)
4206 - lmc creates a default one-stripe lov (2454)
4207 - expand procfs space to handle large clusters (2326)
4208 - increase UML stack to avoid overflow
4209 - update lconf's list of debug and subsystem masks
4210 - fix lfs find --obd (2510)
4211 - /proc tunable for disabling filter read caching (2591)
4212 - stop rpm packages from altering slapd.conf (2301)
4213 - disable nagle in the socknal under 0conf (2578)
4214 - choose mds inode size based on stripe count (2572)
4215 - fix kernel-source rpm problems (2516)
4217 - add --disable-doc to avoid pdf generation (2421)
4218 - update documentation, tests, type-os, comments
4219 - avoid format warnings on ia64
4220 - remove the TOE NAL
4221 - tiny code cleanups by removing unused fields
4223 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4226 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4227 - remove the most blatant lies from BUILDING file (2371)
4228 - change default debug level to reasonable production setting
4229 - reduce client side cache size to reduce cache flush time
4230 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4231 - make TCP zerocopy and pinger support enabled by default (2476)
4232 - sync writes completed after process exits caused crashes (2319)
4233 - maintain correct mount count on the MDS (2356)
4234 - backout 1557, because 2316 wasn't really fixed
4235 - better file I/O statistics gathering in /proc
4236 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4237 - another kernel patch to fix zero-copy TCP function export
4238 - don't take duplicate lock when processing re-sent getattr (2420)
4239 - lctl uses obd_self_export instead of creating new conn (2353)
4240 - MDS/OST recovery case which requires object creation asserted (2425)
4241 - move lfs from /usr/sbin to /usr/bin in packages
4242 - fix race between mds_client_add and mds_client_free (2417)
4243 - use kmalloc instead of slabs in portals (2430)
4244 - don't create duplicate records when a failover MDS is present (2442)
4245 - remove unnecessary mount age check (2332)
4246 - don't remove directory inodes from locks prematurely (2451)
4247 - don't break if MDS service name is the same as hostname (2103)
4248 - fix races in client write RPC generation when cache full (2482)
4250 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4253 - remove now-unused request->rq_obd (278)
4254 - if an allocation fails, print out how much memory we've used (1933)
4255 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4256 - add the "configurable stack size" patch to most series files (1256)
4257 - ability to write large log records, for 100+ OST configs (2306)
4258 - fix NULL deref when filter_prep fails (2314)
4259 - fix operator precedence error in filter_sync
4260 - dynamic allocation of socknal TX descriptors (2315)
4261 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4262 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4263 - ensure that CWARN messages in Portals always get to the syslog
4264 - __init/__exit are not for prototype decls (ldlm_init/exit)
4265 - x86-64 compile warning fixes
4266 - fix gateway LMC keyword conflict (2318)
4267 - fix MDS lock inversions in getattr/reint paths (1844)
4268 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4269 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4270 - DLM race condition prevented some lock evictions (2328)
4271 - ENOMEM detection and retry on socknal sends (2230)
4272 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4273 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4274 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4275 - better error messages when a client is rejected during recovery (1505)
4276 - avoid cancelling locks which were never granted, after failure (2330)
4277 - fix i_sem/journal inversion in mds_client_add (2333)
4278 - fix truncate/getattr lock cycle deadlock (2334)
4279 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4280 - fix two rare exit paths which could leak an l_lock() ref (2321)
4281 - fencepost error in MDS/OST orphan recovery (2226)
4282 - make log record alignment 8 bytes (1988)
4283 - lstripe now fails when requested offset > ost_count (2237)
4284 - ensure that all kernel series have a complete list.h (1607)
4285 - fix crashes in special-file operations (2316)
4286 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4288 - allow configurable automake binary, for testing new versions
4289 - small update to the lfs documentation
4291 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4293 * fix negative export reference count in fsfilt_sync (2312)
4295 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4296 * release candidate 0.9.1
4298 - orphans are moved into the PENDING directory for possible recovery
4299 - replayed opens now open by fid for orphan/rename safety (1042)
4300 - last close of an orphan inode generates a transno (683)
4301 - chdir() and mount() now pin the directory entry (1020)
4302 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4303 - discard very old requests without processing them (1502)
4304 - remove some common, well-understood CERRORs (1505)
4305 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4306 - clear "grant" flags in OST replies until OST grant code lands (1644)
4307 - fix read performance by not clobbering i_blksize on client (1598)
4308 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4309 - make LRU size a /proc tunable, clears locks when reduced (707)
4310 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4311 - server locks take references on exports now (1558)
4312 - build fixes for 2.4.20-rh trees (1663)
4313 - return an error from lov_create if all OSCs are inactive (1751)
4314 - fix import levels when a reconnect happens without a timeout (1597)
4315 - exit early from mds_open if we get a lookup error (1749)
4316 - partial page read at EOF wouldn't wait for disk before sending (1642)
4317 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4318 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4319 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4320 - write/truncate lock inversion (1639)
4321 - Don't auto-load obdclass, portals modules during cleanup (1495)
4322 - fix timestamps from jumping to "now" (1763)
4323 - extra journal assertions (1648)
4324 - add an extra multiunlink test (1771)
4325 - fix read_record/write_record API (1776)
4326 - fix leak of offset_extent, possible incorrect i_size later (1772)
4327 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4328 - Debugging for the kqswnal_get_idle_tx problems (1820)
4329 - Allow recovery to be attempted multiple times (1536)
4330 - Write out MDS last_rcvd file after it is first created (1600)
4331 - Fix tx_descriptor leak in failed transmit situations (1827)
4332 - ext3 journaling fixes for assertion failure after IO error (1871)
4333 - class_export_put() on freed export after completion AST error (1896)
4334 - Fix revalidate looping in VFS (1322)
4335 - Don't access a freed export during MDS_REINT timeout (1521)
4336 - Add open-unlink recovery support on the MDS (1673,1764)
4337 - Return an error if no MDS data was read from last_rcvd (1946)
4338 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4339 - Don't setup a disk device that doesn't match exported UUID (317)
4340 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4341 - avoid committing NULL handle in force close
4342 - local.sh is now a one-stripe LOV configuration
4343 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4344 - fix ext3 htree duplicate directory entry corruption (1516)
4345 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4346 - update to new LOV EA format (2097)
4347 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4349 2003-06-15 Phil Schwan <phil@clusterfs.com>
4352 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4353 - per-import/export recovery handling (958, 931, 959)
4354 - multiple last-rcvd slots, for serving multiple FSes (949)
4355 - connections are again shared between multiple imp/exports (963, 964)
4356 - "umount -f" would hang if any requests needed to be sent (393, 978)
4357 - avoid pinning large req buffer by copying for queued messages (989)
4358 - add "uuid" to "lctl device" command to help upcalls (991)
4359 - "open" RPCs with transnos would confuse recovery counters (1037)
4360 - do proper endian conversion of all wire messages (288, 340, 891)
4361 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4362 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4363 - fix ptlrpc_connection leak in target_handle_connect (1174)
4364 - fix import refcounting bug in OST and MDS cleanup (1134)
4365 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4366 - fix very unlikely obd_types race condition (501)
4367 - remove osc_open hack for echo_client (1187)
4368 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4369 - a failure in read_inode2 leads to deadlock (1139)
4370 - cancel ack-locks as soon as transaction is committed (1072)
4371 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4372 - make sure to commitrw after any preprw to avoid deadlock (1162)
4373 - failing to execute a file in a lustre FS would lock inode (1203)
4374 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4375 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4376 - obd_uuid2tgt would walk off the end of the list (1255)
4377 - on IA64 the osc would give portals incorrect bulk size (1258)
4378 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4379 - fix lock inversion caused by new llite matching code (1282)
4380 - limit the number of dirty pages on a client to 10MB (1286)
4381 - timed out locks were not being corrected cancelled (1289)
4382 - fix O_DIRECT above 4GB on IA-32 (1292)
4383 * major user-visible changes
4384 - fail out/fail over policy now controlled by the upcall (993)
4386 - add OBD_PING to check server availability and failure (954)
4387 - lustre messages are now sent in sending host order (288, 340, 891)
4388 - add eadatalen to MDS getattr reply (340)
4389 - OST read replies may contain second buffer, with per-page status (593)
4391 2003-03-11 Phil Schwan <phil@clusterfs.com>
4394 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4395 - failed open()s could cause deadlock; fixed (867, 869)
4396 - stop cancelling OST locks when files are closed (481)
4397 - overlapping XID spaces caused network corruption (851, 853)
4398 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4399 - setattr_raw added, to do single-RPC, server-side setattrs
4400 - lmc/lconf syntax change for OST UUIDs
4401 - fix crashy race condition between ptlrpc_free_req and osc_close
4402 - don't use request in mdc_enqueue if we hit a timeout (889)
4403 - don't set the inode i_size for regular files from the MDS (896)
4404 - handle out of order completion AST (842)
4405 - don't LBUG if a lock request times out after receiving AST (913)
4406 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4407 - if a bad lock AST arrives, send an error instead of dropping entirely
4408 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4409 - fix leak in bulk IO when only partially completed (899, 900, 926)
4410 - fix O_DIRECT for ia64 (55)
4411 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4412 - C-z after timeout could hang a process forever; fixed (977)
4414 - client-side I/O cache (678, 924, 929, 941, 970)
4416 - READPAGE and SETATTRs which don't take server-side locks get
4419 2003-02-11 Phil Schwan <phil@clusterfs.com>
4422 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4423 - Fix fsfilt_extN_readpage() to read a full page of directory
4424 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4425 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4426 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4427 - Most intent operations were removed, in favour of a new RPC mode
4428 that does a single RPC to the server and bypasses most of the VFS
4429 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4430 - Aggressively cancel local locks on DLM servers
4431 - mds_reint_unlink sends EA to the client if it's the last nlink.
4432 client uses that EA to unlink OST objects.
4433 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4434 - recursive symlinks were fixed (439)
4435 - fixed NULL deref in DEBUG_REQ
4436 - filter_update_lastobjid no longer calls sync, which annoyed extN
4437 - fixed multi-client small-writes to a single file problem (445)
4438 - fixed mtime updates during file writes (607)
4439 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4440 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4441 - fixed Linux OST size reporting problem (444, 656)
4442 - OST now updates object mtime with writes or setattr (607, 619)
4443 - client verifies file size before zeroing page past EOF (445)
4444 - OST now writes last allocated objid to disk with allocation (108)
4445 - LOV on echo now works (409)
4447 - mds_reint_unlink sends a new buffer, with the EA included. this
4448 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4449 set if a regular file was being unlinked, and it was the last link
4450 - use PtlGet from the target for bulk writes (315)
4451 - OST now updates object mtime with writes or setattr (607, 619)
4452 - LDLM now has a grant-time callback to revalidate locked items, if
4454 - Many MDS operations were reorganized to combat race conditions
4456 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4457 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4459 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4462 - Fully reactivate OST imports after reconnection (512, others)
4463 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4464 - More graceful error handling for truncating on dead OST (515)
4465 - Don't error out unless we're actually accessing dead stripes (474)
4466 - Fix garbage sizes when stripes are missing (410)
4467 - LRU counters were broken, causing constant lock purge (433, 432)
4468 - garbage on read from stripes with failed OSTs (441)
4469 - mark OSCs as active before reconnecting during recovery (438)
4470 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4471 - lfind did not preserve OST order in output (443)
4472 - symlinks cause hung clients, incorrect data (439)
4473 - stop dereferencing request after dropping refcount (457)
4474 - don't LASSERT(spin_is_locked) on non-SMP (455)
4475 - fixes for many rename() bugs
4476 - fstat didn't correctly synchronize attributes (399)
4477 - server must handle lock cancellation during blocking AST prep (487)
4478 - bulk descriptors were free()d too soon (511)
4479 - fix paths in lconf, which would load incorrect modules (451, 507)
4480 - fix confusing lconf 'host not found' error message (386)
4481 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4482 - fix race condition in mdc_blocking_ast() for inode access (526)
4483 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4484 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4485 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4486 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4487 - fix lstripe in several ways (kernel side) (527)
4488 - fix request leak in ldlm_cli_enqueue (262)
4489 - incorrect OSC was marked inactive after OST failure
4490 - call mds_fs_cleanup before unmounting filesystem (524)
4491 - fix races between taking ns_lock and ldlm_lock_change_resource
4492 - fix races updating LOV export open file list
4493 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4494 - fix recovery NULL deref in ldlm_cli_cancel_unused
4495 - fix some DLM races by using new hash table for lock handles (419)
4496 - permit the client to specify desired inodes, at replay
4497 - duplicate requests when we queue them for replay reintegration
4498 - fix last_rcvd offset calculation
4499 - sync after each recovered transaction, so we always make progress
4500 - never, not always, ERESTART requests without transnos
4501 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4503 - skip replay if the MDS says that the client is already connected
4504 - don't check for a recovery-enabled export to match lctl's UUID
4505 - don't INC_USE_COUNT for phantom exports
4506 - don't crash when cleaning up phantom exports (567)
4507 - don't double-finish or set replay data for errored mdc_open requests
4508 - abort requests when they time out, so we don't get old replies
4509 - send/receive replies for AST messages again
4510 - if the client says that it doesn't have the lock, cancel it on the
4512 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4513 mark it as destroyed, it will all work out when decref is called
4514 - fix module use counts (22, 581)
4516 - ASTs now expect a reply (server cancels lock on error reply)
4518 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4521 - fix many simultaneous client startup (392)
4522 - fix dentry->d_it clobbering
4523 - credentials weren't being shipped for readdir/getattr operations
4524 - remove invalid assertions triggered during some concurrent MD
4526 - proper Lustre versions added (336, 389)
4527 - fix memory leak for create error case (398)
4528 - fix LOV locking bug that would get cli/srv out of sync
4529 - fix echo client over LOV (409)
4530 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4531 - fix double-O_EXCL intent crash (424)
4532 - avoid sending multiple lock CANCELs (352)
4534 - MDS can do multi-client recovery (modulo bugs in new code)
4536 - many updates, edits, cleanups
4538 2002-11-18 Phil Schwan <phil@clusterfs.com>
4541 - fix null d_it dereference (346)
4542 - fix full OST/dbench hang (333)
4543 - fix permission problem with file removal (286)
4544 - fix removal of OSCs from LOV when they fail
4545 - fix NULL deref during bulk timeout (214)
4546 - fix problems related to multiple filesystems on one MDS (241)
4547 - fixed serious subtle metadata locking bugs
4548 - free locks on clients when inodes are removed due to memory
4550 - fix inode pointer in lock data (285)
4551 - partial support for multiple MDS on a single host (241)
4552 - data locks weren't cancelled at clear_inode time (290, 311)
4553 - intent locks could lead to unbounded lock growth (205)
4554 - added a maximum lock count, an LRU list, and a flusher
4555 - fix multiple rename (365)
4556 - properly abstracted the echo client
4557 - OSC locked 1 byte too many; fixed
4558 - rewrote brw callback code:
4559 - fixed recovery bugs related to LOVs (306)
4560 - fixed too-many-pages-in-one-write crash (191)
4561 - fixed (again) crash in sync_io_timeout (214)
4562 - probably fixed callback-related race (385)
4564 - Add capability to MDS protocol
4565 - LDLM cancellations and callbacks on different portals
4567 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4570 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4571 - defer open object destruction to close time (601981, 138)
4572 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4573 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4574 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4575 - don't call obd_setattr in ll_file_release for destroyed objects
4576 * protocol change to lustre_msg: move |version| and add |flags|
4577 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4578 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4579 * added replay of create, unlink, link and rename operations during
4580 MDS failover; recovery should be much more robust now
4581 * remove failed OSCs from LOVs (only lov_create uses this so far)
4582 * the lustre-HOWTO was brought (more) up to date (582544)
4584 2002-10-23 Phil Schwan <phil@clusterfs.com>
4587 - in-use dentries weren't being reused properly (617851)
4588 - prevent multiple LDLM setup (599178)
4589 - fix LOV size calculations for truncate (617853)
4590 - fix client handling of MDS intent errors (POSIX)
4591 - fix permission bug in lovstripe.c test (624321)
4592 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4593 - truncate past end of file could corrupt data
4594 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4595 - a race in recovery could return ETIMEDOUT to apps (623947)
4596 - building outside the source directory was fixed
4597 * the lustre-HOWTO was brought (more) up to date (582544)
4598 * major progress was made on recovery functionality
4600 2002-10-10 Phil Schwan <phil@clusterfs.com>
4603 - recovery deadlock fix
4604 - rm -rf causes LBUG fix (617817)
4605 - file open by multiple tasks fix (618962)
4606 - directory permissions bugs (602707 and 620007)
4607 - journal_stop fixed with locking (611313)
4608 - O_APPEND failures resolved (618273, perhaps 614459)
4609 - lconf PATH fix (619770)
4610 - IA64 build fix (621450)
4611 - RPC buffer sizes scale with amount of memory
4613 2002-10-01 Phil Schwan <phil@clusterfs.com>
4616 - locks would be cancelled without throwing away data pages,
4617 resulting in inconsistent data (605627)
4618 - inode attributes were not always being refreshed (605627, 612449)
4619 - lconf now continues to cleanup after lctl reports an error
4620 - MDS now enforces user permissions (602707)
4621 - lprocfs cleanup fixed, but not yet enabled (614157)
4622 - fixed infinite server hang, should a client not respond to an AST
4623 - avoid going into recovery if user calls readlink() with a buffer
4624 that's too small (613941)
4625 - AST RPCs no longer require replies (614867) -- this may be changed
4626 - don't crash server if client sends an IOV that's too big (611336)
4627 - fixed lock conversion deadlock (611892)
4628 - fixed the following of symlinks (614622)
4629 * recovery: the server can remove locks from a client that dies, other
4630 clients can make progress
4631 * more extN patch fixes
4632 * compile-time configurable ptlrpc buffer allocations
4634 - collaborative read cache document
4635 - Lustre Lite Performance CDR document-in-progress
4637 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4640 - fix typo in patch-2.4.18
4642 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4645 - clear ptlrpc request each time in handle_incoming_request()
4646 - unlink of files now destroys the object on the OST
4648 2002-09-19 Peter Braam <braam@clusterfs.com>
4650 * add hard link support
4651 * change obdfile creation method
4652 * kernel patch changed
4654 2002-09-19 Peter Braam <braam@clusterfs.com>
4657 - stack overflow bug in extN fixed
4659 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4661 * documentation updates
4662 - add man pages for config tools
4663 - update tests/README to describe testing with new config tools
4664 - finish metadata API descriptions
4665 * bug fixes and cleanups
4666 - statfs workaround for 16TB limit
4667 - LOV stripe allocation improved, can stripe on subset of OSTs
4668 - LOV file size/IO offset was wrong for files > 4GB in size
4669 - object EA data was being dropped, caused files to be unreadable
4670 - memory overflow with non-LOV OST caused memory corruption
4671 - fixed regression tests to work with new config tools, obdfilter
4672 - fixed bug when directory size became larger than 1 block
4673 - fixed bug (for single client case) when PWD was deleted
4674 - invalidate local directory pages when doing intent-based ops
4675 - avoid LDLM oops when lock callback contained bad data
4677 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4679 * documentation updates
4680 * bug fixes and cleanups
4681 - configuration tools
4684 - 64-bit compile warnings
4685 - 64-bit internal statfs data
4687 * test_brw on persistent OST devices
4689 * lprocfs (disabled)
4691 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4693 * documentation updates
4694 * bug fixes and cleanups
4695 * configuration tools
4697 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4700 * many small fixes to 0_5_4
4701 * io/network handling
4702 * thinkos in MDS operations
4704 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4707 * crucial basic fixes to 0.5.3
4708 * IOR, Iozone work over Elan
4711 2002-08-07 Phil Schwan <phil@clusterfs.com>
4712 * version 0_5_3, our first alpha
4713 * we use the new Portals iovs
4714 * documentation updates
4715 * bug fixes and cleanups
4716 * small changes in the DLM wire protocol
4718 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4719 * version 0_5_1 with some initial stability,
4720 * locking on MD and file I/O.
4721 * documentation updates
4722 * several bug fixes since 0.5.0
4723 * small changes in wire protocol
4725 2002-07-18 Phil Schwan <phil@clusterfs.com>
4727 * delivered as Lustre Light Alpha
4728 * fixed a crash after handling invalid MDS requests
4729 * fixed directory pages for architectures with non-4k pages sizes
4731 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4732 * release version v0_4_4
4733 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4735 * Fixes a number of interruption problems with OST operations.
4736 * Update documentation for portals header changes
4737 * Move all wire protocol structs/defines to lustre_idl.h
4738 * Fixes symlink length bug.
4739 * Add tcpdump to repository.
4741 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4742 * release version v0_4_3
4743 * Fixes statfs for inodes on extN.
4744 * Fixes bug in runtests which would delete /etc/hosts.
4745 * Use 64-bit object IDs wherever possible (not into VFS though)
4746 Remove ost_get_info, which is unused by lustre, and out of date.
4748 2002-07-03 Peter Braam <braam@clusterfs.com>
4749 * release version v0_4_2 Fixes a lookup error (type not passed)
4750 * move forward to head of Portals
4751 * move forward to latest Lustre kernel
4753 2002-06-25 Peter Braam <braam@clusterfs.com>
4754 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob