1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: Stack overflow during MDS log replay
18 Details : ease stack pressure by using a thread dealing llog_process.
22 Description: fix build on ppc32
23 Details : compile code with -m64 flag produce wrong object file for ppc32.
27 Description: add message levels for liblustreapi
31 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
32 Details : Move the 'good_osts' check before the 'total_bavail' check. This
33 will result in an -EAGAIN and in the exit call path we call
34 alloc_rr() which will with increasing aggressiveness attempt to
35 aquire precreated objects on the minimum number of required OSCs.
39 Description: Use old size assignment to avoid deadlock
40 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
41 the scheduling while holding a spinlock. We do not need locking
42 for size in ll_update_inode() because size is only updated from
43 the MDS for directories or files without objects, so there is no
44 other place to do the update, and concurrent access to such inodes
45 are protected by the inode lock.
49 Description: resolve "_IOWR redefined" build error on SLES10
53 Description: dump the memory debugging after all modules are unloaded to
54 suppress false negative in conf_sanity test 39
56 Severity : enhancement
58 Description: build kernel-ib packages for OFED 1.3 in our release cycle
63 Description: fix SLES kernel versioning
64 Details : the kernel version for our SLES 10 kernel did not include a "-"
65 before the "smp" at the end. while this was not a problem in
66 general, it did mean that software trying to use the kernel
67 version to try to detect a vendor specific kernel would fail.
68 this was most evident by the OFED build scripts.
72 Description: Don't update lov_desc members until making sure they are valid
73 Details : When updating lov_desc members via proc fs, need fix their
74 validities before doing the real update.
78 Description: don't put request into delay list while invalidate in flight.
79 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
80 import in flight. this produce timeout for invalidate and sometimes
85 Frequency : on ppc only
86 Description: not convert ost objects for directory because it's not exist.
87 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
91 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
93 Severity : enhancement
95 Description: Update to support 2.6.22.14 vanilla kernel.
99 Frequency : rare, on recovery
100 Description: read procfs can produce deadlock in some situation
101 Details : Holding lprocfs lock which send rpc can produce block for destroy
102 obd objects and this also block reconnect with -EALREADY. This isn't
103 fix all lprocfs bugs - but make it rare.
105 Severity : enhancement
107 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
110 Frequency : frequent on X2 node
112 Description: mdc_set_open_replay_data LBUG
113 Details : Set replay data for requests that are eligible for replay.
117 Description: lustre_mgs: operation 101 on unconnected MGS
118 Details : When MGC is disconnected from MGS long enough, MGS will evict the
119 MGC, and late on MGC cannot successfully connect to MGS and a lot
120 of the error messages complaining that MGS is not connected.
123 Frequency : on start mds
125 Description: Implement get_info(last_id) in obdfilter.
128 Frequency : occasional
130 Description: Correctly check stale fid, not start epoch if ost not support SOM
131 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
132 overwrited with new generated fid, but mds can anwer with one of these
133 two fids and both is not stale. setattr incorectly start epoch and
134 assume will be called done_writeting, but without SOM done_writing
138 Frequency : rare, depends on device drivers and load
140 Description: MDS or OSS nodes crash due to stack overflow
141 Details : Code changes in 1.8.0 increased the stack usage of some functions.
142 In some cases, in conjunction with device drivers that use a lot
143 of stack the MDS (or possibly OSS) service threads could overflow
144 the stack. One change which was identified to consume additional
145 stack has been reworked to avoid the extra stack usage.
148 Frequency : occasional
150 Description: Do not fail import if osc_interpret_create gets -EAGAIN
151 Details : If osc_interpret_create got -EAGAIN it immediately exits and
152 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
153 oscc_has_objects and see OSC has no objests and call
154 oscc_internal_create to resend create request.
156 Severity : enhancement
158 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
160 Severity : enhancement
162 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
165 Frequency : very rare
167 Description: Fix replay if there is an un-replied request and open
168 Details : In some cases, older replay request will revert the
169 mcd->mcd_last_xid on MDS which is used to record the client's
172 Severity : enhancement
174 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
176 Severity : enhancement
178 Description: Add rhel5 support to HEAD.
180 Serverity : enhancement
182 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
187 Description: Don't allow skipping OSTs if index has been specified.
188 Details : Don't allow skipping OSTs if index has been specified, make locking
189 in internal create lots better.
193 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
194 Details : access to bitfield in structure is always rounded to long
195 and this produce problem with not atomic change any bit.
199 Description: Lustre make rpms failed.
200 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
201 builds Lustre rpms from tarball.
204 Frequency : rare on shutdown ost
206 Description: If llog cancel was not send before clean_exports phase, this can
207 produce deadlock in llog code.
208 Details : If llog thread has last reference to obd and call class_import_put
209 this produce deadlock because llog_cleanup_commit_master wait when
210 last llog_commit_thread exited, but this never success because was
211 called from llog_commit_thread.
215 Description: allow userland application know is lost one of stripes.
216 Details : fill lvb_blocks with error code on ost and return it to
217 application if error flag found.
221 Description: NULL lov_tgts causing MDS oops
222 Details : more safe checks for NULL lov_tgts for avoid oops.
224 Severity : enhancement
226 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
230 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
231 Details : If an OST is down the MDS will hang indefinitely in
232 obd_statfs_rqset() waiting for the statfs data. While for
233 MDS QOS usage of statfs, it should not stuck in waiting.
235 Severity : enhancement
237 Description: remote_acl support
238 Details : Support ACL-based permission check for remote user.
239 Support setfacl/getfacl for remote user with the utils
240 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
241 the system "{s,g}etfacl" utils.
243 Severity : enhancement
245 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
247 Severity : enhancement
249 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
253 Description: make mgs_setparam() handle fsname containing dash
254 Details : fsname containing a dash does not work with lctl conf_param
256 Severity : enhancement
258 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
260 Severity : enhancement
262 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
266 Frequency : rare, at shutdown
267 Description: access already free / zero obd_namespace.
268 Details : if client_disconnect_export was called without force flag set,
269 and exist connect request in flight, this can produce access to
270 NULL pointer (or already free pointer) when connect_interpret
271 store ocd flags in obd_namespace.
275 Frequency : only at startup
276 Description: not alloc memory with spinlock held.
277 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
278 if any spinlock held.
280 Severity : enhancement
282 Description: make lustre randomly fail allocating memory
283 Details : Make lustre randomly failed allocating memory for testing purpose.
285 Severity : enhancement
287 Description: lost problems with lov objid file
288 Details : Fixes some scability and access to not inited memory problems
289 in work with lov objdid file.
294 Description: lfs find does not continue on file error
295 Details : Continue other files processing when a file/dir is absent.
299 Description: Inconsistent usage of lustre_pack_reply()
300 Details : Standardize the usage of lustre_pack_reply() such that it
301 always generate a CERROR on failure.
304 Frequency : occasional
306 Description: MDS hang or stay in waiting lock
307 Details : If client receive lock with CBPENDING flag ldlm need send lock
308 cancel as separate rpc, to avoid situation when cancel request
309 can't processed due all i/o threads stay in wait lock.
313 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
317 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
321 Description: Update to SLES9 kernel-2.6.5-7.287.3.
325 Description: Update to RHEL4 latest kernel.
327 Severity : enhancement
329 Description: Build SLES10 patchless client fails
330 Details : The configure was broken by run ./configure with
331 --with-linux-obj=.... argument for patchless client. When the
332 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
333 can't search header adequately. Use absolute path such as
334 -I($LINUX)/include instead.
338 Description: interrupt oig_wait produce painc on resend.
339 Details : brw_redo_request can be used for resend requests from ptlrpcd and
340 private set, and this produce situation when rq_ptlrpcd_data not
341 copyed to new allocated request and triggered LBUG on assert
342 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
343 ptlrpcd set if request is changed and can be safety changed to use
348 Description: LASSERT_{REQ,REP}SWAB macros are buggy
349 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
350 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
351 these macros and replace them with their difinitions instead.
355 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
356 Details : Update which_patch & target file for SLES10 latest kernel.
359 Description: Remove client patches from SLES 10 kernel.
360 Details : This causes SLES 10 clients to behave as patchless clients
361 even on a Lustre-patched (server) kernel.
363 Severity : enhancement
365 Description: self-adjustable client's lru lists
366 Details : use adaptive algorithm for managing client cached locks lru
367 lists according to current server load, other client's work
368 pattern, memory activities, etc. Both, server and client
369 side namespaces provide number of proc tunables for controlling
372 Severity : enhancement
374 Description: light-weight GSS support
375 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
376 and reduce performance overhead.
378 Severity : enhancement
380 Description: Linux keyring support
381 Details : Support using service of Linux keyring for Lustre GSS internal
382 context refresh/cache mechanism.
386 Description: Fix errors in lfs documentation
387 Details : Fixes man pages
391 Description: don't use GFP_* in generic Lustre code.
392 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
396 Description: obdclass is limited by single OBD_ALLOC(idarray)
397 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
401 Description: warnings with build patchless client with vanila 2.6.19 and up
402 Details : change old ctl_table style and replace ctl_table/ctl_table_header
403 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
407 Frequency : Only for SLES
408 Description: sanity_quota fail test_1
409 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
410 file, which makes TSTID[2] inunique.
414 Frequency : Only for SLES9
415 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
416 Details : Update target/ChangeLog/which_patch .
420 Frequency : Only for test_10 in sanity-quota.sh
421 Description: a bug in quota test code
422 Details : checking a test flag in a wrong place
426 Frequency : Only for test_13 in sanity-quota.sh
427 Description: a bug in quota test script
428 Details : shouldn't check overfull allocated quota
432 Description: jbd statistics
433 Details : Port older jbd statistics patch for sles10
435 Severity : enhancement
437 Description: Add printing inode info into message about error in writepage.
442 Description: reply_lock_interpret crash due to race with it and lock cancel.
443 Details : Do not replay locks that are being cancelled. Do not reference
444 locks by their address during replay, just by their handle.
448 Frequency : When flocks are used.
449 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
450 Details : Flock locks might destroy just granted lock if it could be merged
451 with another existing flock, this is done in completion handler,
452 so teach ldlm_cli_enquque_fini that this is a valid case for
458 Description: Oops in read and write path when failing to allocate lock.
459 Details : Check if lock allocation failed and return error back.
463 Description: lstripe command fails for valid OST index
464 Details : The stripe offset is compared to lov->desc.ld_tgt_count
465 instead of lov->desc.ld_active_tgt_count.
469 Description: sanity.sh failed test 103
470 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
475 Description: df doesn't work properly if diskfs blocksize != 4K
476 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
480 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
481 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
485 Description: obd_health_check_timeout too short
486 Details : set obd_health_check_timeout as 1.5x of obd_timeout
490 Description: llapi_file_create() does not allow some changes
491 Details : add llapi_file_open() that allows specifying the mode and
492 open flags, and also returns an open file handle.
496 Description: merge and cleanup kernel patches.
497 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
501 Description: Add journal checksum support.(Kernel part)
502 Details : The journal checksum feature adds two new flags i.e
503 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
504 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
505 indicates that the commit block contains the checksum for
506 the blocks described by the descriptor blocks. Now commit
507 record can be sent to disk without waiting for descriptor
508 blocks to be written to disk. This behavior is controlled
509 using JBD2_FEATURE_ASYNC_COMMIT flag.
513 Description: OSS needs mutliple precreate threads
514 Details : Add ability to start more than one create thread per OSS.
518 Description: Sanity.sh test_65k failed: setstripe should have succeeded
519 Details : execute lfs setstripe on client
523 Description: mds_obd_create error creating tmp object
524 Details : When the user sets quota on root, llog will be affected and can't
525 create files and write files.
528 Frequency : Always on ia64 patchless client, and possibly others.
530 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
531 Details : This allows the patchless client to be loaded on architectures
536 Description: RedHat Update kernel for RHEL5
537 Details : Add the kernel config file for RHEL5.
541 Description: RedHat Update kernel for RHEL5
542 Details : Modify the kernel config file more closer RHEL5.
546 Description: Build failure against Centos5 (RHEL5)
547 Details : Define PAGE_SIZE when it isn't present.
551 Description: test_7 on sanity-quota.sh can't be used on separate machines
552 Details : add facet to handle it
556 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
557 Details : with NFS, the anon dentry's parent was set to itself in
558 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
559 corresponding dentry other than use rec->ur_name.
561 Severity : enhancement
563 Description: lfs setstripe enhancement
564 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
568 Description: enable data checksumming by default
569 Details : enable checksum by default, allow --disable-checksum
570 configure option and "-o nochecksum" mount option. Checksums
571 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
575 Description: lustre support for RHEL5
576 Details : Add support for RHEL5.
580 Description: Client eviction due to failover config
581 Details : after a connection loss, the lustre client should attempt to
582 reconnect to the last active server first before trying the
583 other potential connections.
587 Description: when mds and osts use different quota unit(32bit and 64bit),
588 quota will be released repeatly.
589 Details : void sending multiple quota reqs to mds, which will keep the
590 status between the reqs.
594 Description: osts not allocated evenly to files
595 Details : change the condition to increase offset_idx
598 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
600 Description: Data corruption for OSTs that are formatted larger than 2TB
602 Details : When generating the bio request for lustre file writes the
603 sector number would overflow a temporary variable before being
604 used for the IO. The data reads correctly from Lustre (which
605 will overflow in a similar manner) but other file data or
606 filesystem metadata may be corrupted in some cases.
610 Description: Tune the kernel for good SCSI performance.
611 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
612 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
617 Description: rewrite ext2-derived code in obdclass/uuid.c
618 Details : rewrite inherited code (uuid parsing code from ext2 utils)
619 from scratch preserving functionality.
623 Description: "lfs find -obd UUID" prints directories
624 Details : "lfs find -obd UUID" will return all directory names instead
625 of just file names. It is incorrect because the directories
626 do not reside on the OSTs.
630 Description: use i_size_read and i_size_write in 2.6 port
631 Details : replace inode->i_size access with i_size_read/write()
634 Frequency : when removing large files
636 Description: scheduling issue during removal of large Lustre files
637 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
638 It causes scheduling issues when removing large files (17TB in the
642 Frequency : only with liblustre clients on XT3
644 Description: evictions taking too long
645 Details : allow llrd to evict clients directly on OSTs
648 Frequency : only on ppc
650 Description: /proc/fs/lustre/devices broken on ppc
651 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
652 We should make sure the type of 'index' is loff_t and then cast
653 explicitly as needed below. Do not assign an explicitly cast
657 Frequency : only for rhel5
659 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
660 Details : Modify the target file & which_kernel.
662 Severity : enhancement
664 Description: omit set fsid for export NFS
665 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
666 as FS_REQUIRES_DEV for avoid problems with generate fsid.
670 Frequency : Always, for kernels after 2.6.16
671 Description: Fix warning idr_remove called for id=.. which is not allocated.
672 Details : Last kernels save old s_dev before kill super and not allow
673 to restore from callback - restore it before call kill_anon_super.
677 Description: buffer overruns could theoretically occur
678 Details : llapi_semantic_traverse() modifies the "path" argument by
679 appending values to the end of the origin string, and a
680 overrun may occur. Adding buffer overrun check in liblustreapi.
684 Description: Fix error on 'ls .' at the top of the Lustre mount.
685 Details : Don't revalidate dentry if it is a root dentry.
689 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
690 Details : Modify vm-tunables-rhel4.patch.
694 Description: Kernel config for 2.6.18-vanilla.
695 Details : Modify targets/2.6-vanilla.target.in.
696 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
697 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
698 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
699 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
703 Description: improve handling recoverable errors
704 Details : if request processig with error which can be recoverable on server
705 request should be resend, otherwise page released from cache and
710 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
711 Details : Modify target file & which_patch.
712 A flaw was found in the IA32 system call emulation provided
713 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
714 value could be stored in the %RAX register, which could trigger an
715 out-of-bounds system call table access. An untrusted local user
716 could exploit this flaw to run code in the kernel
717 (ie a root privilege escalation). (CVE-2007-4573).
721 Description: O_DIRECT bypasses client statistics.
722 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
723 read_ahead_stats were updated. copied the stats section from
724 osc_send_oap_rpc() into async_internal().
728 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
732 Description: change order of libsysio includes
733 Details : '#include sysio.h' should always come before '#include xtio.h'
737 Description: handle "serious error: objid * already exists" more gracefully
738 Details : If LAST_ID value on disk is smaller than the objects existing in
739 the O/0/d* directories, it indicates disk corruption and causes an
740 LBUG(). If the object is 0-length, then we should use the existing
741 object. This will help to avoid a full fsck in most cases.
743 Severity : enhancement
745 Description: adapt the lustre_config script to support the upgrade case
746 Details : Add "-u" option for lustre_config script to support upgrading 1.4
747 server targets to 1.6 in parallel.
751 Description: To avoid grant space > avaible space when the disk is almost
752 full. Without this patch you might see the error "grant XXXX >
753 available" or some LBUG about grant, when the disk is almost
755 Details : In filter_check_grant, for non_grant cache write, we should
756 check the left space by if (*left > ungranted + bytes), instead
757 of (*left > ungranted), because only we are sure the left space
758 is enough for another "bytes", then the ungrant space should be
759 increase. In client, we should update cl_avail_grant only there
760 is OBD_MD_FLGRANT in the reply.
764 Description: Update RHEL 4 kernel to fix local root privilege escalation.
765 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
766 described in CVE-2007-4573. This problem could allow untrusted
767 local users to gain root access.
770 Frequency : when using O_DIRECT and quotas
772 Description: Incorrect file ownership on O_DIRECT output files
773 Details : block usage reported by 'lfs quota' does not take into account
774 files that have been written with O_DIRECT.
779 Description: touch file failed when fs is not full
780 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
781 otherwise we can get ENOSP while fs is not full.
785 Description: parallel lock callbacks
786 Details : Instead of sending blocking and completion callbacks as separated
787 requests, adding them to a set and sending in parallel.
790 Frequency : only for Cray XT3
791 Bugzilla : 12829/13455
792 Description: Changing primary group doesn't change the group lustre assigns to
794 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
795 overridden with the primary group provided by the group upcall,
796 whereas the supplied fsgid can be trusted if it is in the list of
797 supplementary groups returned by the group upcall.
799 Severity : enhancement
801 Description: Allow masking D_WARNING, D_ERROR messages from console
802 Details : Console messages can now be disabled via lnet.printk.
806 Description: User code with malformed file open parameter crashes client node
807 Details : Before packing join_file req, all the related reference should be
808 checked carefully in case some malformed flags cause fake
809 join_file req on client.
813 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
814 Details : ldlm_completion_ast() assumes that a lock is granted when the req
815 mode is equal to the granted mode. However, it should also check
816 that LDLM_FL_CP_REQD is not set.
820 Description: Heavy nfs access might result in deadlocks
821 Details : After ELC code landed, it is now improper to enqueue any mds
822 locks under och_sem, because enqueue might want to decide to
823 cancel open locks for same inode we are holding och_sem for.
827 Description: Client eviction while running blogbench
828 Details : A lot of unlink operations with concurrent I/O can lead to a
829 deadlock causing evictions. To address the problem, the number of
830 oustanding OST_DESTROY requests is now throttled to
831 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
832 callbacks are processed in priority.
836 Description: enable ACLs on MDS by default
837 Details : ACLs must be enabled on MDS by default.
840 Frequency : PPC/PPC64 only
842 Description: conflicts between asm-ppc64/types.h and lustre_types.h
843 Details : fix duplicated definitions between asm-ppc64/types.h and
844 lustre_types.h on PPC.
847 Frequency : PPC/PPC64 only
849 Description: asm-ppc/segment.h does not exist
850 Details : fix compile issue on PPC.
854 Description: better handle error messages in extents code
857 Frequency : RHEL4 only
859 Description: mkfs is very slow on IA64/RHEL4
860 Details : A performance regression has been discovered in the MPT Fusion
861 driver between versions 3.02.73rh and 3.02.99.00rh. As a
862 consequence, we have downgraded the MPT Fusion driver in the RHEL4
863 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
865 Severity : enhancement
867 Description: SNMP support enhancement
868 Details : Adding total number of sampled request for an MDS node in snmp
871 Severity : enhancement
873 Description: Optimize ldlm waiting list processing for PR extent locks
874 Details : When processing waiting list for read extent lock and meeting read
875 lock that is same or wider to it that is not contended, skip
876 processing rest of the list and immediatelly return current
877 status of conflictness, since we are guaranteed there are no
878 conflicting locks in the rest of the list.
882 Description: Time out and refuse to reconnect
883 Details : When the failover node is the primary node, it is possible
884 to have two identical connections in imp_conn_list. We must
885 compare not conn's pointers but NIDs, otherwise we can defeat
886 connection throttling.
890 Description: port llog fixes from b1_6 into HEAD
891 Details : Port llog reference couting and some llog cleanups from b1_6
892 (bug 10800) into HEAD, for protect from panic and access to already
893 free llog structures.
897 Description: Detect stride IO mode in read-ahead
898 Details : When a client does stride read, read-ahead should detect that and
899 read-ahead pages according to the detected stride pattern.
903 Description: data checksumming impacts single node performance
904 Details : add support for several checksum algorithm. Currently, only CRC32
905 and Adler-32 are supported. The checksum type can be changed on
906 the fly via /proc/fs/lustre/osc/*/checksum_type.
910 Description: use adler32 for page checksums
911 Details : when available, use the Adler-32 algorithm instead of CRC32 for
916 Description: build for x2 fails
917 Details : fix compile issue on Cray systems.
921 Description: Properly match for duplicate locks
922 Details : Due to different lock order from skiplists code, we need to
923 traverse entire list for now
926 Frequency : only on PPC/SLES10
928 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
929 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
930 be defined. Add a hack in mkfs_lustre.c to work around this compile
935 Description: LASSERT on MDS when client holding flock lock dies
936 Details : ldlm pool logic depends on number of granted locks equal to
937 number of released locks which is not true for flock locks, so
938 just exclude such locks from consideration.
942 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
943 Details : Use fsfilt_sync() for both the whole filesystem sync and
944 individual file sync to eliminate dangerous inode locking
945 with I_LOCK that can lead to a deadlock.
949 Description: performance in 1.6.3
950 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
951 sure that 1MB requests are not fragmented by the block layer.
953 Severity : enhancement
955 Description: organize the server-side client stats on per-nid basis
956 Details : Change the structure of stats under obdfilter and mds to
964 The "uuid"s file would list the uuids of _active_ exports.
965 And the clear entry is to clear all stats and stale nids.
967 Severity : enhancement
969 Description: eliminate client locks in face of contention
970 Details : file contention detection and lockless i/o implementation
973 --------------------------------------------------------------------------------
975 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
977 * Support for kernels:
978 2.6.5-7.283 (SLES 9),
979 2.6.9-55.EL (RHEL 4),
980 2.6.16.46-0.14 (SLES 10),
981 2.6.18.8 vanilla (kernel.org)
982 * Client support for unpatched kernels:
983 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
984 2.6.16 - 2.6.22 vanilla (kernel.org)
985 * Due to recently discovered recovery problems, we do not recommend
986 using patchless RHEL 4 clients with this or any earlier release.
987 * Recommended e2fsprogs version: 1.39.cfs8
988 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
989 * Starting with this release, the ldiskfs backing filesystem required
990 by Lustre is now in its own package, lustre-ldiskfs. This package
991 should be installed. It is versioned separately from Lustre and
992 may be released separately in future.
996 Description: block reactivating mgc import until all deactivates complete
997 Details : Fix race when failing back MDT/MGS to itself (testing)
1000 Frequency : at statup only
1002 Description: mds_lov_synchronize race leads to various problems
1003 Details : simultaneous MDT->OST connections at startup can cause the
1004 sync to abort, leaving the OSC in a bad state.
1006 Severity : enhancement
1008 Description: add optional extra BUILD_VERSION info
1009 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1010 to override the lustre version.
1013 Frequency : 2.6.18 servers only
1015 Description: ll_kern_mount() doesn't release the module reference
1016 Details : The ldiskfs module reference count never drops down to 0
1017 because ll_kern_mount() doesn't release the module reference.
1022 Description: server LBUG when using old ost_num_threads parameter
1023 Details : Accept the old ost_num_threads parameter but warn that it
1024 is deprecated, and fix an off-by-one error that caused an LBUG.
1029 Description: Transient SCSI error results in persistent IO issue
1030 Details : iobuf->dr_error is not reinitialized to 0 between two
1034 Frequency : sometimes when underlying device returns I/O errors
1036 Description: OSTs not going read-only during write failures
1037 Details : OSTs are not remounted read-only when the journal commit threads
1038 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1039 instead of the ext3 wrappers.
1043 Description: poor connect scaling with increasing client count
1044 Details : Don't run filter_grant_sanity_check for more than 100 exports
1045 to improve scaling for large numbers of clients.
1048 Frequency : SLES10 only
1050 Description: sanity-quota.sh quotacheck failed: rc = -22
1051 Details : Quotas cannot be enabled on SLES10.
1054 Frequency : liblustre clients only
1056 Description: getdirentries does not give error when run on compute nodes
1057 Details : getdirentries does not fail when the size specified as an argument
1058 is too small to contain at least one entry
1060 Severity : enhancement
1062 Description: Add LNET router traceability for debug purposes
1063 Details : If a checksum failure occurs with a router as part of the
1064 IO path, the NID of the last router that forwarded the bulk data
1065 is printed so it can be identified.
1070 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1071 Details : Due to a race condition, liblustre clients were occasionally
1072 evicted incorrectly.
1074 Severity : enhancement
1076 Description: lfs setstripe use optional parameters instead of postional
1079 Severity : enhancement
1081 Description: Nanosecond timestamp support for ldiskfs
1082 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1083 resolution timestamps. There is not yet support for this at
1084 the Lustre filesystem level.
1087 Frequency : during server recovery
1089 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1090 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1094 Frequency : nfs export on patchless client
1096 Description: connectathon hang when test nfs export over patchless client
1097 Details : Disconnected dentry cannot be found with lookup, so we do not need
1098 to unhash it or make it invalid
1101 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1102 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1103 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1108 Description: open req refcounting wrong on reconnect
1109 Details : If reconnect happened between getting open reply from server and
1110 call to mdc_set_replay_data in ll_file_open, we will schedule
1111 replay for unreferenced request that we are about to free.
1112 Subsequent close will crash in variety of ways.
1113 Check that request is still eligible for replay in
1114 mdc_set_replay_data().
1119 Description: disable writes to filesystem when reading health_check file
1120 Details : the default for reading the health_check proc file has changed
1121 to NOT do a journal transaction and write to disk, because this
1122 can cause reads of the /proc file to hang and block HA state
1123 checking on a healthy but otherwise heavily loaded system. It
1124 is possible to return to the previous behaviour during configure
1125 with --enable-health-write.
1127 Severity : enhancement
1129 Description: 64-bit inode version
1130 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1131 to the inode. This will be required for version-based recovery.
1136 Description: MDS fails to start if a duplicate client export is detected
1137 Details : in some rare cases it was possible for a client to connect to
1138 an MDS multiple times. Upon recovery the MDS would detect this
1139 and fail during startup. Handle this more gracefully.
1141 Severity : enhancement
1143 Description: Add -o localflock option to simulate old noflock
1145 Details : This will achieve local-only flock/fcntl locks
1151 Description: log_commit_thread vs filter_destroy race leads to crash
1152 Details : Take import reference before releasing llog record semaphore
1157 Description: Wrong request locking in request set processing
1158 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1159 imp_delayed_list, in this place should be used imp_lock.
1162 Frequency : when reconnection
1164 Description: Grant Leak when osc reconnect to OST
1165 Details : When osc reconnect ost, OST(filter) should check whether it
1166 should grant more space to client by comparing fed_grant and
1167 cl_avail_grant, and return the granted space to client instead
1168 of "new granted" space, because client will call osc_init_grant
1169 to update the client grant space info.
1172 Frequency : when client reconnect to OST
1174 Description: Grant Leak when osc do resend and replay bulk write
1175 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1176 bulk write request, because the grant info while be sync between
1177 OSC and OST when reconnect, and we should ignore the grant info
1178 these of resend/replay write req.
1183 Description: Grant space more than avaiable left space sometimes.
1184 Details : When then OST is about to be full, if two bulk writing from
1185 different clients came to OST. Accord the avaliable space of the
1186 OST, the first req should be permitted, and the second one
1187 should be denied by ENOSPC. But if the seconde arrived before
1188 the first one is commited. The OST might wrongly permit second
1189 writing, which will cause grant space > avaiable space.
1192 Frequency : when client is evicted
1194 Description: Grant might be wrongly erased when osc is evicted by OST
1195 Details : when the import is evicted by server, it will fork another
1196 thread ptlrpc_invalidate_import_thread to invalidate the
1197 import, where the grant will be set to 0. While the original
1198 thread will update the grant it got when connecting. So if
1199 the former happened latter, the grant will be wrongly errased
1200 because of this race.
1205 Description: Checking Stale with correct fid
1206 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1207 to check whether it is stale, because sometimes, we want the
1208 enqueue happened anyway, and op_data.fid2 will not be initialized.
1210 Severity : enhancement
1212 Description: update patchless client
1213 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1216 Frequency : only with 2.4 kernel
1218 Description: random memory corruption
1219 Details : size of struct ll_inode_info is to big for union inode.u and this
1220 can be cause of random memory corruption.
1225 Description: Memory leak in recovery
1226 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1227 It should also check obd_fail before fsfilt_start, otherwise if
1228 fsfilt_start return -EROFS,(failover mds during mds recovery).
1229 then the req will return with repmsg->transno = 0 and rc = EROFS.
1230 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1231 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1232 be freed no matter whether fsfilt_commit success or not.
1235 Frequency : only with huge count clients
1237 Description: Prevents from taking the superblock lock in llap_from_page for
1239 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1240 for llap_from_page call in ll_removepage prevents from taking the
1241 superblock lock for a soon died page.
1246 Description: Not check open intent error before release open handle
1247 Details : in some rare cases, the open intent error is not checked before
1248 release open handle, which may cause
1249 ASSERTION(open_req->rq_transno != 0), because it tries to release
1250 the failed open handle.
1255 Description: Set cat log bitmap only after create log success.
1256 Details : in some rare cases, the cat log bitmap is set too early. and it
1257 should be set only after create log success.
1261 Description: Accessing a block bevice can re-enable I/O when Lustre is
1262 tearing down a device.
1263 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1268 Description: service threads may hog cpus when there are a lot of requests
1269 Details : Insert cond_resched to give other threads a chance to use some CPU
1274 Description: the cat log was not initialized in recovery
1275 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1276 unlink log on mds will not be initialized until mds post
1277 recovery. And also in mds post recovery, the unlink log will
1278 initialization will be done asynchronausly, so there will be race
1279 between add unlink log and unlink log initialization.
1283 Description: brw_stats were being printed incorrectly
1284 Details : brw_stats were being printed as log2 but all of them were not
1285 recorded as log2. Also remove some code duplication arising from
1286 filter_tally_{read,write}.
1290 Frequency : rare, only in recovery.
1291 Description: ASSERTION(req->rq_type != LI_POISON) failed
1292 Details : imp_lock should be held while iterating over imp_sending_list for
1293 prevent destroy request after get timeout in ptlrpc_queue_wait.
1297 Description: replay-single.sh test 52 fails
1298 Details : A lock's skiplist need to be cleanup when it being unlinked
1299 from its resource list.
1303 Description: Short directio read returns full requested size rather than
1305 Details : Direct I/O operations should return actual amount of bytes
1306 transferred rather than requested size.
1308 Severity : enhancement
1310 Description: metadata RPC reduction (e.g. for rm performance)
1311 Details : decrease the amount of synchronous RPC between clients and servers
1312 by canceling conflicing lock before the operation on the client
1313 and packing thier handles into the main operation RPC to server.
1315 Severity : enhancement
1317 Description: Async OSC create to avoid the blocking unnecessarily.
1318 Details : If a OST has no remain object, system will block on the creating
1319 when need to create a new object on this OST. Now, ways use
1320 pre-created objects when available, instead of blocking on an
1321 empty osc while others are not empty. If we must block, we block
1322 for the shortest possible period of time.
1324 Severity : enhancement
1326 Description: refine locking for avoid write wrong info into lov_objid file
1327 Details : fix possible races with add new target and write/update data in
1330 --------------------------------------------------------------------------------
1332 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1337 Frequency : on some architectures
1339 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1340 Details : Uninitialized flags sometimes cause configuration commands to
1344 Frequency : patchless clients only
1346 Description: missing __iget() symbol export
1347 Details : The __iget() symbol export is missing. To avoid the need for
1348 this on patchless clients the deathrow inode reaper is turned
1349 off, and we depend on the VM to clean up old inodes. This
1350 dependency was during via the fix for bug 12181.
1352 --------------------------------------------------------------------------------
1354 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1356 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1357 INTEROPERATE with older versions automatically. In many cases a
1358 special upgrade step is needed. Please read the
1359 user documentation before upgrading any part of a live system.
1360 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1361 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1362 * WARNING: Lustre configuration and startup changes are required with
1363 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1369 Severity : enhancement
1371 Description: MountConf
1372 Details : Lustre configuration is now managed via mkfs and mount
1373 commands instead of lmc and lconf. New obd types (MGS, MGC)
1374 are added for dynamic configuration management. See
1375 https://mail.clusterfs.com/wikis/lustre/MountConf for
1378 Severity : enhancement
1380 Description: dynamic OST addition
1381 Details : OSTs can now be added to a live filesystem
1383 Severity : enhancement
1385 Description: startup order invariance
1386 Details : MDTs and OSTs can be started in any order. Clients only
1387 require the MDT to complete startup.
1389 Severity : enhancement
1391 Description: parallel, asynchronous orphan cleanup
1392 Details : orphan cleanup is now performed in separate threads for each
1393 OST, allowing parallel non-blocking operation.
1395 Severity : enhancement
1397 Description: optimized stripe assignment
1398 Details : stripe assignments are now made based on ost space available,
1399 ost previous usage, and OSS previous usage, in order to try
1400 to optimize storage space and networking resources.
1402 Severity : enhancement
1404 Description: Permanently set tunables
1405 Details : All writable /proc/fs/lustre tunables can now be permanently
1406 set on a per-server basis, at mkfs time or on a live system.
1408 Severity : enhancement
1410 Description: Lustre message v2
1411 Details : Add lustre message format v2.
1413 Severity : enhancement
1415 Description: client OST exclusion list
1416 Details : Clients can be started with a list of OSTs that should be
1417 declared "inactive" for known non-responsive OSTs.
1421 Description: SPEC SFS validation failure on NFS v2 over lustre.
1422 Details : Changes the blocksize for regular files to be 2x RPC size,
1423 and not depend on stripe size.
1425 Severity : enhancement
1427 Description: Multiple MD RPCs in flight.
1428 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1429 To avoid overly-overloading MDS, introduce a limit on number of
1430 MDS RPCs in flight for a single client and add /proc controls
1431 to adjust this limit.
1433 Severity : enhancement
1435 Description: client read/write statistics
1436 Details : Add client read/write call usage stats for performance
1437 analysis of user processes.
1438 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1439 file access. extents_stats shows chunk size distribution.
1440 extents_stats_per_process show chunk size distribution per
1443 Severity : enhancement
1445 Description: mds statistics
1446 Details : Add detailed mds operations statistics in
1447 /proc/fs/lustre/mds/*/stats.
1451 Description: Failure of copying files with lustre special EAs.
1452 Details : Client side always return success for setxattr call for lustre
1453 special xattr (currently only "trusted.lov").
1458 Description: Refcount LNET uuids
1459 Details : The global LNET uuid list grew linearly with every startup;
1460 refcount repeated list entries instead of always adding to
1463 Severity : enhancement
1465 Description: Dynamic service threads
1466 Details : Within a small range, start extra service threads
1467 automatically when the request queue builds up.
1470 Frequency : mixed-endian client/server environments
1472 Description: mixed-endian crashes
1473 Details : The new msg_v2 system had some failures in mixed-endian
1476 Severity : enhancement
1478 Description: Easy OST removal
1479 Details : OSTs can be permanently deactivated with e.g. 'lctl
1480 conf_param lustre-OST0001.osc.active=0'
1482 Severity : enhancement
1484 Description: MGS proc entries
1485 Details : Added basic proc entries for the MGS showing what filesystems
1488 Severity : enhancement
1490 Description: provide MGS failover
1491 Details : Added config lock reacquisition after MGS server failover.
1493 Severity : enhancement
1495 Description: add Linux 2.4 support
1496 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1500 Description: a large application tries to do I/O to the same resource and dies
1501 in the middle of it.
1502 Details : Check the req->rq_arrival time after the call to
1503 ost_brw_lock_get(), but before we do anything about
1504 processing it & sending the BULK transfer request. This
1505 should help move old stale pending locks off the queue as
1506 quickly as obd_timeout.
1509 Frequency : when an incorrect nid is specified during startup
1511 Description: ptlrpc connect to non-existant node causes kernel crash
1512 Details : LNET can't be re-entered from an event callback, which
1513 happened when we expire a message after the export has been
1514 cleaned up. Instead, hand the zombie cleanup off to another
1517 Severity : enhancement
1519 Description: plain/inodebits lock performance improvement
1520 Details : Grouping plain/inodebits in granted list by their request modes
1521 and bits policy, thus improving the performance of search through
1525 Frequency : only if OST filesystem is corrupted
1527 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1528 Details : for a short time RPCs with bulk IO are in the replay list,
1529 but replay of bulk IOs is unimplemented. If the OST filesystem
1530 is corrupted due to disk cache incoherency and then replay is
1531 started it is possible to trip an assertion. Avoid putting
1532 committed RPCs into the replay list at all to avoid this issue.
1535 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1537 Description: System hang on startup
1538 Details : This bug allowed the liblustre (e.g. catamount) client to
1539 return to the app before handling all startup RPCs. This
1540 could leave the node unresponsive to lustre network traffic
1541 and manifested as a server ptllnd timeout.
1543 Severity : enhancement
1545 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1546 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1547 boolean to control whether to print peer debug info when a
1548 client's RPC times out.
1551 Frequency : only for kernels with patches from Lustre below 1.4.3
1553 Description: Remove old rdonly API
1554 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1557 Frequency : only for devices with external journals
1559 Description: Set external device read-only also
1560 Details : During a commanded failover stop, we set the disk device
1561 read-only while the server shuts down. We now also set any
1562 external journal device read-only at the same time.
1565 Frequency : when upgrading from 1.4 while trying to change parameters
1567 Description: The wrong (new) MDC name was used when setting parameters for
1568 upgraded MDT's. Also allows changing of OSC (and MDC)
1569 parameters if --writeconf is specified at tunefs upgrade time.
1572 Frequency : when setting specific ost indicies
1574 Description: QOS code breaks on skipped indicies
1575 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1576 created with --index need not be sequential.
1578 Severity : enhancement
1580 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1581 Details : The uninit_groups feature works in conjunction with the kernel
1582 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1583 up the pass1 processing of e2fsck. This is a read-only feature
1584 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1585 mount filesystems that have had this feature enabled.
1587 Severity : enhancement
1589 Description: Improve multi-block allocation algorithm to avoid fragmentation
1590 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1591 allocation locality and avoid filesystem fragmentation.
1593 ------------------------------------------------------------------------------
1595 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1597 * Support for kernels:
1598 2.4.21-47.0.1.EL (RHEL 3)
1599 2.6.5-7.283 (SLES 9)
1600 2.6.9-42.0.10.EL (RHEL 4)
1601 2.6.12.6 vanilla (kernel.org)
1602 2.6.16.27-0.9 (SLES 10)
1603 * Recommended e2fsprogs version: 1.39.cfs5
1605 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1609 Frequency : occasional, depends on client load and configuration
1610 Bugzilla : 12181, 12203
1611 Description: data loss for recently-modified files
1613 Details : In some cases it is possible that recently written or created
1614 files may not be written to disk in a timely manner (this should
1615 normally be within 30s unless client IO load is very high).
1616 The problem appears as zero-length files or files that are a
1617 multiple of 1MB in size after a client crash or client eviction
1618 that are missing data at the end of the file.
1620 This problem is more likely to be hit on clients where files are
1621 repeatedly created and unlinked in the same directory, clients
1622 have a large amount of RAM, have many CPUs, the filesystem has
1623 many OSTs, the clients are rebooted frequently, and/or the files
1624 are not accessed by other nodes after being written.
1626 The presence of the problem can be detected by looking at
1627 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1628 smaller than the second (nr_unused) then dirty files will not
1629 be flushed automatically to disk. "sync; sleep 10" should be
1630 run several times on the node before unmounting it to update
1631 Lustre (this is also safe to run on nodes without this problem).
1633 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1634 that can cause this same problem, so customers using that kernel
1635 also need to update the kernel in addition to Lustre. In order
1636 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1638 It is normal that files written just before a client crash (less
1639 than 30s) may not yet have been flushed to disk, even for local
1643 Frequency : frequent on thin XT3 nodes
1645 Description: UUID collision on thin XT3 Linux nodes
1646 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1647 randomly, since we relied on an insufficiently-seeded PRNG.
1652 Description: OSS hangs after "All ost request buffers busy"
1653 Details : A deadlock between quota and journal operations caused OSS
1654 hangs after printing "All ost request buffers busy."
1657 Frequency : always on liblustre builds
1659 Description: Cleanup compiler warnings on liblustre
1662 Frequency : always on liblustre builds on XT3
1664 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1668 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1669 > 32000 subdirectories
1670 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1671 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1672 subdirectory count crosses 32000. This will aid e2fsck to
1673 correctly handle more than 32000 subdirectories.
1676 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1678 Description: System hang on startup
1679 Details : This bug allowed the liblustre (e.g. catamount) client to
1680 return to the app before handling all startup RPCs. This
1681 could leave the node unresponsive to lustre network traffic
1682 and manifested as a server ptllnd timeout.
1684 Severity : enhancement
1686 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1687 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1688 boolean to control whether to print peer debug info when a
1689 client's RPC times out.
1694 Description: make O_SYNC working on 2.6 kernels
1695 Details : 2.6 kernels use different method for mark pages for write,
1696 so need add a code to lustre for O_SYNC work.
1701 Description: Failure to close file and release space on NFS
1702 Details : Put inode details into lock acquired in ll_intent_file_open.
1703 Use mdc_intent_lock in ll_intent_open to properly
1704 detect all kind of errors unhandled by mdc_enqueue.
1709 Description: proc file read during shutdown sometimes raced obd removal,
1711 Details : Add lock to prevent obd access after proc file removal.
1714 Frequency : Only for files larger than 4GB on 32-bit clients.
1716 Description: improperly doing page alignment of locks
1717 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1718 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1723 Description: RPCs being resent when they shouldn't be
1724 Details : Some RPCs that should not be resent are being resent. This
1725 can cause inconsistencies in the RPC state machine. Do not
1726 resend such requests.
1729 Frequency : rare, only with NFS export
1731 Description: Crash on NFS re-export node
1732 Details : under very unusual load conditions an assertion is hit in
1733 ll_intent_file_open()
1736 Frequency : only if OST filesystem is corrupted
1738 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1739 Details : for a short time RPCs with bulk IO are in the replay list,
1740 but replay of bulk IOs is unimplemented. If the OST filesystem
1741 is corrupted due to disk cache incoherency and then replay is
1742 started it is possible to trip an assertion. Avoid putting
1743 committed RPCs into the replay list at all to avoid this issue.
1748 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1749 Details : Large single O_DIRECT read and write calls can fail to allocate
1750 a sufficiently large buffer to process the request. In case of
1751 allocation failure the allocation is retried with a smaller
1752 buffer and broken into smaller requests.
1754 Severity : enhancement
1756 Description: Add -o localflock option to simulate old noflock behaviour.
1757 Details : This will achieve local-only flock/fcntl locks coherentness.
1762 Description: versioning check is incomplete
1763 Details : Checking the version difference of client vs. server, report
1764 error if the gap is too big.
1769 Description: add support PG_writeback bit
1770 Details : add support for PG_writeback bit for Lustre, for more carefull
1771 work with page cache in 2.6 kernel. This also fix some deadlocks
1772 and remove hack for work O_SYNC with 2.6 kernel.
1774 Severity : enhancement
1776 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1777 Details : The uninit_groups feature works in conjunction with the kernel
1778 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1779 up the pass1 processing of e2fsck. This is a read-only feature
1780 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1781 mount filesystems that have had this feature enabled.
1783 Severity : enhancement
1785 Description: Improve multi-block allocation algorithm to avoid fragmentation
1786 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1787 allocation locality and avoid filesystem fragmentation.
1789 ------------------------------------------------------------------------------
1791 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1793 * Support for kernels:
1794 2.6.9-42.0.3.EL (RHEL 4)
1795 2.6.5-7.276 (SLES 9)
1796 2.4.21-47.0.1.EL (RHEL 3)
1797 2.6.12.6 vanilla (kernel.org)
1798 2.6.16.21-0.8 (SLES10)
1799 * Recommended e2fsprogs version: 1.39.cfs2-0
1801 * The backwards-compatible /proc/sys/portals symlink has been removed
1802 in this release. Before upgrading, please ensure that you change
1803 any configuration scripts or /etc/sysctl.conf files that access
1804 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1805 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1806 in advance of the upgrade on any system running Lustre 1.4.6 or
1807 newer, since /proc/sys/lnet was added in that version.
1808 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1812 Frequency : only when quota is used
1814 Description: avoid scanning export list for quota master
1815 Details : Change the algorithms to avoid scanning export list in order
1816 to improve the efficiency.
1819 Frequency : MDS failover only, very rarely
1821 Description: "went back in time" messages on mds failover
1822 Details : The greatest transno may be lost when the current operation
1823 finishes with an error (transno==0) and the client's last_rcvd
1824 record is over-written. Save the greatest transno in the
1825 mds_last_transno for this case.
1828 Frequency : always for specific kernels and striping counts
1830 Description: client may get "Matching packet too big" without ACL support
1831 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1832 when trying to access files in certain configurations. The
1833 clients should in fact be denied when mounting because they do
1834 not understand ACLs.
1837 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1839 Description: many clients connecting with IO in progress causes connect timeouts
1840 Details : Avoid synchronous journal commits to avoid delays caused by many
1841 clients connecting/disconnecting when bulk IO is in progress.
1842 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1843 OST_IO_PORTAL to avoid delays behind potentially many pending
1847 Frequency : occasionally with multiple writers to a single file
1849 Description: shared writes to file may result in wrong size reported by stat()
1850 Details : Allow growing of kms when extent lock is cancelled
1853 Frequency : always with random mmap IO to multi-striped file
1855 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1856 Details : Only if the hole is at the end of OST object so that kms is too
1857 small. Fix is to increase kms accordingly in ll_nopage.
1860 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1862 Description: writes to a missing object would leak memory on the OST
1863 Details : If there is an inconsistency between the MDS and OST filesystems,
1864 such that the MDS references an object that doesn't exist, writes
1865 to that object will leak memory due to incorrect cleanup in the
1866 error handling path, eventually running out of memory on the OST.
1871 Description: Creating too long symlink causes lustre errors
1872 Details : Check symlink and name lengths before sending requests to MDS.
1875 Frequency : only if flock is enabled (not on by default)
1877 Description: posix locks not released on fd closure on 2.6.9+
1878 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1879 kernels, this caused such locks not to be released on fd close and
1880 then assertions on fs unmount about still used locks.
1883 Frequency : MDS failover only, very rarely
1885 Description: clients may get ASSERTION(granted_lock != NULL)
1886 Details : When request was taking a long time, and a client was resending
1887 a getattr by name lock request. The were multiple lock requests
1888 with the same client lock handle and
1889 mds_getattr_name->fixup_handle_for_resent_request found one of the
1890 lock handles but later failed with ASSERTION(granted_lock != NULL).
1895 Description: handle->h_buffer_credits > 0, assertion failure
1896 Details : h_buffer_credits is zero after truncate, causing assertion
1897 failure. This patch extends the transaction or creates a new
1901 Frequency : NFS re-export or patchless client
1902 Bugzilla : 11179, 10796
1903 Description: Crash on NFS re-export node (__d_move)
1904 Details : We do not want to hash the dentry if we don't have a lock.
1905 But if this dentry is later used in d_move, we'd hit uninitialised
1906 list head d_hash, so we just do this to init d_hash field but
1907 leave dentry unhashed.
1910 Frequency : NFS re-export or patchless client
1912 Description: NFS exports has problem with symbolic link
1913 Details : lustre client didn't properly install dentry when re-exported
1914 to NFS or running patchless client.
1917 Frequency : NFS re-export or patchless client
1919 Description: Various nfs/patchless fixes.
1920 Details : fixes reuse disconected alias for lookup process - this fixes
1921 warning "find_exported_dentry: npd != pd",
1922 fix permission error with open files at nfs.
1923 fix apply umask when do revalidate.
1926 Frequency : occasional
1928 Description: Crash on NFS re-export node
1929 Details : calling clear_page() on the wrong pointer triggered oops in
1930 generic_mapping_read().
1933 Frequency : rarely, using O_DIRECT IO
1935 Description: unaligned directio crashes client with LASSERT
1936 Details : check for unaligned buffers before trying any requests.
1939 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
1941 Description: stale data returned from RAID cache
1942 Details : If only a small amount of IO is done to the RAID device before
1943 reading it again it is possible to get stale data from the RAID
1944 cache instead of reading it from disk.
1947 Frequency : always for sles10 kernel
1949 Description: sles10 support
1950 Details : ll_follow_link: compile fixes and using of nd_set_link
1951 under newer kernels.
1954 Frequency : depends on arch, kernel and compiler version, always on sles10
1957 Description: recursive or deep enough symlinks cause stack overflow
1958 Details : getting rid of large stack-allocated variable in
1962 Frequency : depends on hardware
1964 Description: lustre write performance loss in the SLES10 kernel
1965 Details : the performance loss is caused by using of write barriers in the
1966 ext3 code. The SLES10 kernel turns barrier support on by
1967 default. The fix is to undo that change for ldiskfs.
1969 ------------------------------------------------------------------------------
1971 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
1973 * Support for kernels:
1974 2.6.9-42.0.3EL (RHEL 4)
1975 2.6.5-7.276 (SLES 9)
1976 2.4.21-47.0.1.EL (RHEL 3)
1977 2.6.12.6 vanilla (kernel.org)
1981 Frequency : quota enabled and large files being deleted
1983 Description: releasing more than 4GB of quota at once hangs OST
1984 Details : If a user deletes more than 4GB of files on a single OST it
1985 will cause the OST to spin in an infinite loop. Release
1986 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
1991 Description: statfs data retrieved from /proc may be stale or zero
1992 Details : When reading per-device statfs data from /proc, in the
1993 {kbytes,files}_{total,free,avail} files, it may appear
1994 as zero or be out of date.
1997 Frequency : systems with MD RAID1 external journal devices
1999 Description: lconf's call to blkid is confused by RAID1 journal devices
2000 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2001 of returning all block devices that match the journal UUID.
2004 Frequency : always, for aggregate stripe size over 4GB
2006 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2007 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2008 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2011 Frequency : always if "lfs find" used on a local file/directory
2013 Description: "lfs find" segfaults if used on a local file/directory
2014 Details : The case where a directory component was not specified wasn't
2015 handled correctly. Handle this properly.
2018 Frequency : always on ppc64
2020 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2021 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2022 perform endianness conversion.
2025 Frequency : rarely (truncate to non-zero file size after write under load)
2026 Bugzilla : 10730, 10687
2027 Description: Files padded with zeros to next 4K multiple
2028 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2029 it is possible that files that are truncated to a non-zero size
2030 immediately after being written are filled with zero bytes beyond
2031 the truncated size. No file data is lost.
2033 Severity : enhancement
2035 Description: Allow recovery/failover for liblustre clients.
2036 Details : liblustre clients were unaware of failover configurations until
2039 Severity : enhancement
2041 Description: user file locks should fail when not mounting with flock option
2042 Details : Set up an error-returning stub in ll_file_operations.lock field
2043 to prevent incorrect behaviour when client is mounted without
2044 flock option. Also, set up properly f_op->flock field for
2048 Frequency : always on ia64
2050 Description: "lfs df" loops on printing out MDS statfs information
2051 Details : The obd_ioctl_data was not initialized and in some systems
2052 this caused a failure during the ioctl that did not return
2053 an error. Initialize the struct and return an error on failure.
2056 Frequency : SLES 9 only
2058 Description: Error of copying files with lustre special EAs as root
2059 Details : Client side always return success for setxattr call for lustre
2060 special xattr (currently only "trusted.lov").
2063 Frequency : rarely on clusters with both ia64+i386 clients
2065 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2066 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2067 clients are doing concurrent writes to the same file it is
2068 possible that the ia64 clients may LASSERT because the OST
2069 extent locks are not PAGE_SIZE aligned. Ensure that grown
2070 locks are always aligned on the request boundary.
2073 Frequency : specific use, occasional
2075 Description: Overwriting in use executable truncates on-disk binary image
2076 Details : If one node attempts to overwrite an executable in use by
2077 another node, we now correctly return ETXTBSY instead of
2078 truncating the file.
2080 Severity : enhancement
2082 Description: Async OSC create to avoid the blocking unnecessarily.
2083 Details : If a OST has no remain object, system will block on the creating
2084 when need to create a new object on this OST. Now, ways use
2085 pre-created objects when available, instead of blocking on an
2086 empty osc while others are not empty. If we must block, we block
2087 for the shortest possible period of time.
2092 Description: chmod on Lustre root is propagated to other clients
2093 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2094 invalid by the follow_mount time.
2097 Frequency : liblustre clients only
2099 Description: Race in 'instant cancel' lock handling could lead to such locks
2100 never to be granted in case of SMP MDS
2101 Details : Do not destroy not yet granted but cbpending locks in
2105 Frequency : replay/resend of open
2107 Description: non null lock assetion failure in mds_intent_policy
2108 Details : Trying to replay/resend lockless open requests resulted in
2109 mds_open() returning 0 with no lock. Now it sets a flag if
2110 a lock is going to be returned.
2112 Severity : enhancement
2114 Description: Checksum enhancements
2115 Details : New checksum enhancements allow for resending RPCs that failed
2118 Severity : enhancement
2120 Description: Tunables on number of dirty pages in cacche
2121 Details : Allow to set limit on number of dirty pages cached.
2126 Description: client crash on unmount - lock still has references
2127 Details : In some error handling cases it was possible to leak a lock
2128 reference on a client while accessing a file. This was not
2129 harmful to the client during operation, but would cause the
2130 client to crash when the filesystem is unmounted.
2133 Frequency : specific case, rare
2135 Description: ETXTBSY on mds though file not in use by client
2136 Details : ETXTBSY is no longer incorrectly returned when attempting to
2137 chmod or chown a directory that the user previously tried to
2138 execute or a currently-executing binary.
2141 Frequency : extremely rare except on liblustre-based clients
2143 Description: Lustre space not freed when files are deleted
2144 Details : Clean up open-unlinked files after client eviction. Previously
2145 the unlink was skipped and the files remained as orphans.
2150 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2151 Details : In some cases it was possible to send a blocking callback to
2152 a client doing a glimpse, even though that client didn't get
2153 a lock granted. When the glimpse lock is cancelled on the OST
2154 the freed lock is left on the waiting list and corrupted the list.
2157 Frequency : all core dumps
2159 Description: Broke core dumps to lustre
2160 Details : Negative dentry may be unhashed if parent does not have UPDATE
2161 lock, but some callers, e.g. do_coredump, expect dentry to be
2162 hashed after successful create, hash it in ll_create_it.
2164 ------------------------------------------------------------------------------
2166 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2168 * Support for kernels:
2169 2.6.9-42.0.2.EL (RHEL 4)
2170 2.6.5-7.276 (SLES 9)
2171 2.4.21-40.EL (RHEL 3)
2172 2.6.12.6 vanilla (kernel.org)
2176 Frequency : always on RHEL 3
2178 Description: Number of open files grows over time
2179 Details : The number of open files grows over time, whether or not
2180 Lustre is started. This was due to a filp leak introduced
2181 by one of our kernel patches.
2183 ------------------------------------------------------------------------------
2185 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2187 * Support for kernels:
2188 2.6.9-42.EL (RHEL 4)
2189 2.6.5-7.276 (SLES 9)
2190 2.4.21-40.EL (RHEL 3)
2191 2.6.12.6 vanilla (kernel.org)
2196 Bugzilla : 5719, 9635, 9792, 9684
2197 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2198 Details : If a server is under heavy load and cannot reply to new
2199 connection requests before the client resends the (re)connect,
2200 the connection handling code can behave badly if two service
2201 threads are concurrently handing separate (re)connections from
2202 the same client. Add better locking to the connection handling
2203 code, and ensure that only a single connection will be processed
2204 for a given client UUID, even if the lock is dropped.
2206 Severity : enhancement
2208 Description: add TCP zero-copy support to kernel
2209 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2210 sends if the hardware supports scatter-gather and checksumming.
2211 This allows socklnd to do client-write and server-read more
2212 efficiently and reduce CPU utilization from skbuf copying.
2215 Frequency : only if NFS exporting from client
2217 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2218 Details : If setting attributes on a file created under NFS that had
2219 never been opened it would be possible to oops the client
2220 if the file had no objects.
2223 Frequency : always for liblustre
2225 Description: liblustre client does MDS+OSTs setattr RPC for each write
2226 Details : When doing a write from a liblustre client, the client
2227 incorrectly issued an RPC to the MDS and each OST the file was
2228 striped over in order to update the timestamps. When writing
2229 with small chunks and many clients this could overwhelm the MDS
2230 with RPCs. In all cases it would slow down the write because
2231 these RPCs are unnecessary.
2233 Severity : enhancement
2235 Description: allow number of MDS service threads to be changed at module load
2236 Details : It is now possible to change the number of MDS service threads
2237 running. Adding "options mds mds_num_threads={N}" to the MDS's
2238 /etc/modprobe.conf will set the number of threads for the next
2239 time Lustre is restarted (assuming the "mds" module is also
2240 reloaded at that time). The default number of threads will
2241 stay the same, 32 for most systems.
2246 Description: OST crash if filesystem is unformatted or corrupt
2247 Details : If an OST is started on a device that has never been formatted
2248 or if the filesystem is corrupt and cannot even mount then the
2249 error handling cleanup routines would dereference a NULL pointer.
2254 Description: NULL pointer deref in llap_from_page.
2255 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2256 incorrect) mapping if the page was truncated/reclaimed while it was
2257 searched for. Check for this condition and skip such pages when
2258 doing readahead. Introduce extra check to llap_from_page() to
2259 verify page->mapping->host is non-NULL (so page is not anonymous).
2262 Frequency : Sometimes when using sys_sendfile
2264 Description: "page not covered by a lock" warnings from ll_readpage
2265 Details : sendfile called ll_readpage without right page locks present.
2266 Now we introduced ll_file_sendfile that does necessary locking
2267 around call to generic_file_sendfile() much like we do in
2271 Frequency : with certain MDS communication failures at client mount time
2273 Description: NULL pointer deref after failed client mount
2274 Details : a client connection request may delayed by the network layer
2275 and not be sent until after the PTLRPC layer has timed out the
2276 request. If the client fails the mount immediately it will try
2277 to clean up before the network times out the request. Add a
2278 reference from the request import to the obd device and delay
2279 the cleanup until the network drops the request.
2282 Frequency : occasionally during client (re)connect
2284 Description: assertion failure during client (re)connect
2285 Details : processing a client connection request may be delayed by the
2286 client or server longer than the client connect timeout. This
2287 causes the client to resend the connection request. If the
2288 original connection request is replied in this interval, the
2289 client may trip an assertion failure in ptlrpc_connect_interpret()
2290 which thought it would be the only running connect process.
2293 Frequency : only with obd_echo servers and clients that are rebooted
2295 Description: kernel BUG accessing uninitialized data structure
2296 Details : When running an obd_echo server it did not start the ping_evictor
2297 thread, and when a client was evicted an uninitialized data
2298 structure was accessed. Start the ping_evictor in the RPC
2299 service startup instead of the OBD startup.
2301 Severity : enhancement
2302 Bugzilla : 10193 (patchless)
2303 Description: Remove dependency on various unexported kernel interfaces.
2304 Details : No longer need reparent_to_init, exit_mm, exit_files,
2305 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2308 Frequency : rare (only users of deprecated and unsupported LDAP config)
2310 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2311 Details : LDAP apparently contains 'lustreName' attributes instead of
2312 'name'. A simple remapping of the name is sufficient.
2315 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2317 Description: waiting_locks_callback trips kernel BUG if client is evicted
2318 Details : Running with the dump_on_timeout debug flag turned on makes
2319 it possible that the waiting_locks_callback() can try to dump
2320 the Lustre kernel debug logs from an interrupt handler. Defer
2321 this log dumping to the expired_lock_main() thread.
2323 Severity : enhancement
2325 Description: Support NFS exporting on 2.6 kernels.
2326 Details : Implement non-rawops metadata methods for NFS server to use without
2327 changing NFS server code.
2330 Frequency : very rare (synthetic metadata workload only)
2332 Description: two racing renames might cause an MDS thread to deadlock
2333 Details : Running the "racer" program may cause one MDS thread to rename
2334 a file from being the source of a rename to being the target of
2335 a rename at exactly the same time that another thread is doing
2336 so, and the second thread has already enqueued these locks after
2337 doing a lookup of the target and is trying to relock them in
2338 order. Ensure that we don't try to re-lock the same resource.
2341 Frequency : only very large systems with liblustre clients
2343 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2344 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2345 all OSTs in parallel. This allows the eviction of stale liblustre
2346 clients to proceed much faster than if they were done in series,
2347 and also offers similar improvements for other set_info RPCs.
2352 Description: excessive CPU usage during initial read phase on client
2353 Details : During the initial read phase on a client, it would agressively
2354 retry readahead on the file, consuming too much CPU and impacting
2355 performance (since 1.4.5.8). Improve the readahead algorithm
2356 to avoid this, and also improve some other common cases (read
2357 of small files in particular, where "small" is files smaller than
2358 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2363 Description: MDS crash when receiving packet with unknown intent.
2364 Details : Do not LBUG in unknown intent case, just return -EFAULT
2366 Severity : enhancement
2367 Bugzilla : 9293, 9385
2368 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2369 Details : Do not serialize getattr (non-intent version) and statfs.
2372 Frequency : occasional, when OST network is overloaded/intermittent
2374 Description: client evicted by OST after bulk IO timeout
2375 Details : If a client sends a bulk IO request (read or write) the OST
2376 may evict the client if it is unresposive to its data GET/PUT
2377 request. This is incorrect if the network is overloaded (takes
2378 too long to transfer the RPC data) or dropped the OST GET/PUT
2379 request. There is no need to evict the client at all, since
2380 the pinger and/or lock callbacks will handle this, and the
2381 client can restart the bulk request.
2384 Frequency : Always when mmapping file with no objects
2386 Description: client crashes when mmapping file with no objects
2387 Details : Check that we actually have objects in a file before doing any
2388 operations on objects in ll_vm_open, ll_vm_close and
2394 Description: Request leak when working with deleted CWD
2395 Details : Introduce advanced request refcount tracking for requests
2396 referenced from lustre intent.
2398 Severity : Enhancement
2400 Description: Cache open file handles on client.
2401 Details : MDS now will return special lock along with openhandle, if
2402 requested and client is allowed to hold openhandle, even if unused,
2403 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2404 closing files for every read/write openration.
2406 Severity : Enhancement
2408 Description: Cache open negative dentries on client when possible.
2409 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2410 negative dentries on lock revocation.
2415 Description: Remounting a client read-only wasn't possible with a zconf mount
2416 Details : It wasn't possible to remount a client read-only with llmount.
2418 Severity : enhancement
2419 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2420 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2421 MPICH is included to add Lustre-specific ADIO interfaces.
2422 This is based closely on the UFS ADIO layer and only differs
2423 in file creation, in order to allow the OST striping to be set.
2424 This is user-contributed code and not supported by CFS.
2429 Description: extended inode attributes (immutable, append-only) work improperly
2430 when 2.4 and 2.6 kernels are used on client/server or vice versa
2431 Details : Introduce kernel-independent values for these flags.
2433 Severity : enhancement
2436 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2437 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2438 granularity. Now, Fractional MB granularity is be supported,
2439 this is very useful for low memory system.
2441 Severity : enhancement
2443 Description: Getattr by fid
2444 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2445 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2446 have locks and avoids getattr rpc requests for it on every stat.
2449 Frequency : Always, for filesystems larger than 2TB
2451 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2452 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2453 filesystems larger than 2TB to be mounted with mballoc enabled.
2456 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2458 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2459 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2460 filesystems larger than 2TB is mounted then the kernel will
2461 silently corrupt the start of the filesystem. CONFIG_LBD is
2462 enabled for all CFS-supported kernels, but the possibility of
2463 this happening with a modified kernel config exists.
2465 Severity : enhancement
2467 Description: add client O_DIRECT support for 2.6 kernels
2468 Details : It is now possible to do O_DIRECT reads and writes to files
2469 in the Lustre client mountpoint on 2.6 kernel clients.
2471 Severity : enhancement
2473 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2474 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2475 parallel, not waiting for response from every OST before sending
2476 a rpc to the next OST.
2481 Description: setattr vs write race when updating file timestamps
2482 Details : Client processes that update a file timestamp into the past
2483 right after writing to the file (e.g. tar) it is possible that
2484 the updated file modification time can be reset to the current
2485 time due to a race between processing the setattr and write RPC.
2487 Severity : enhancement
2489 Description: Bring 'lfs find' closer in line with regular Linux find.
2490 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2491 -print0 options and obtains all the needed info through the lustre
2494 Severity : enhancement
2496 Description: support up to 1024 configured devices on one node
2497 Details : change obd_dev array from statically allocated to dynamically
2498 allocated structs as they are first used to reduce memory usage
2503 Description: Flush dirty partially truncated pages during truncate
2504 Details : Immediatelly flush partially truncated pages in filter_setattr,
2505 this way we completely avoid having any pages in page cache on OST
2506 and can retire ugly workarounds during writes to flush such pages.
2511 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2512 Details : protect inode from truncation within vfs_unlink() context
2513 just take a reference before calling vfs_unlink() and release it
2514 when parent's i_sem is free.
2519 Description: last_id value checked outside lock on OST caused LASSERT failure
2520 Details : If there were multiple MDS->OST object precreate requests in
2521 flight, it was possible that the OST's last object id was checked
2522 outside a lock and incorrectly tripped an assertion. Move checks
2523 inside locks, and discard old precreate requests.
2526 Frequency : always, if extents are used on OSTs
2528 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2529 Details : OSTs using the extents format would not zero the high 16 bits of
2530 the index physical block number. This is not a problem for any
2531 OST filesystems smaller than 16TB, and no kernels support ext3
2532 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2533 new/modified files) and can be fixed for existing filesystems
2534 with e2fsprogs-1.39-cfs1.
2539 Description: import connection selection may be incorrect if timer wraps
2540 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2541 import connections to be ignored if the 32-bit jiffies counter
2542 wraps. Use a 64-bit jiffies counter.
2545 Frequency : very large clusters immediately after boot
2547 Description: LNET request buffers exhausted under heavy short-term load
2548 Details : If a large number of client requests are generated on a service
2549 that has previously never seen so many requests it is possible
2550 that the request buffer growth cannot keep up with the spike in
2551 demand. Instead of dropping incoming requests, they are held in
2552 the LND until the RPC service can accept more requests.
2555 Frequency : Sometimes during replay
2557 Description: Assertion failure in ll_local_open after replay.
2558 Details : If replay happened on an open request reply before we were able
2559 to set replay handler, reply will become not swabbed tripping the
2560 assertion in ll_local_open. Now we set the handler right after
2561 recognising of open request
2564 Frequency : very rare
2566 Description: kernel reports "badness in vsnprintf"
2567 Details : Reading from the "recovery_status" /proc file in small chunks
2568 may cause a negative length in lprocfs_obd_rd_recovery_status()
2569 call to vsnprintf() (which is otherwise harmless). Exit early
2570 if there is no more space in the output buffer.
2572 Severity : enhancement
2574 Description: clear OBD RPC statistics by writing to them
2575 Details : It is now possible to clear the OBD RPC statistics by writing
2576 to the "stats" file.
2581 Description: Client mtime is not the same on different clients after utimes
2582 Details : In some cases, the client was using the utimes() syscall on
2583 a file cached on another node. The clients now validate the
2584 ctime from the MDS + OSTs to determine which one is right.
2589 Description: Inability to activate failout mode
2590 Details : lconf script incorrectly assumed that in pythong string's numeric
2591 value is used in comparisons.
2594 Frequency : always with multiple stripes per file
2596 Description: Inefficient object allocation for mutli-stripe files
2597 Details : When selecting which OSTs to stripe files over, for files with
2598 a stripe count that divides evenly into the number of OSTs,
2599 the MDS is always picking the same starting OST for each file.
2600 Return the OST selection heuristic to the original design.
2605 Description: mount failures may take full timeout to return an error
2606 Details : Under some heavy load conditions it is possible that a
2607 failed mount can wait for the full obd_timeout interval,
2608 possibly several minutes, before reporting an error.
2609 Instead return an error as soon as the status is known.
2611 Frequency : quota enabled and large files being deleted
2613 Description: releasing more than 4GB of quota at once hangs OST
2614 Details : If a user deletes more than 4GB of files on a single OST it
2615 will cause the OST to spin in an infinite loop. Release
2616 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2621 Description: statfs data retrieved from /proc may be stale or zero
2622 Details : When reading per-device statfs data from /proc, in the
2623 {kbytes,files}_{total,free,avail} files, it may appear
2624 as zero or be out of date.
2627 Frequency : systems with MD RAID1 external journal devices
2629 Description: lconf's call to blkid is confused by RAID1 journal devices
2630 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2631 of returning all block devices that match the journal UUID.
2634 Frequency : always, for aggregate stripe size over 4GB
2636 Description: assertion fails when trying to use 4GB stripe size
2637 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2638 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2641 Frequency : always on ppc64
2643 Description: the first write on an ext3 filesystem with mballoc got stuck
2644 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2645 perform endianness conversion.
2647 ------------------------------------------------------------------------------
2649 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2651 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2652 INTEROPERATE with older versions automatically. Please read the
2653 user documentation before upgrading any part of a live system.
2654 * WARNING: Lustre networking configuration changes are required with
2655 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2658 * Support for kernels:
2659 2.6.9-22.0.2.EL (RHEL 4)
2660 2.6.5-7.244 (SLES 9)
2661 2.6.12.6 vanilla (kernel.org)
2664 Severity : enhancement
2665 Bugzilla : 7981/8208
2666 Description: Introduced Lustre Networking (LNET)
2667 Details : LNET is new networking infrastructure for Lustre, it includes
2668 a reorganized network configuration mode (see the user
2669 documentation for full details) as well as support for routing
2670 between different network fabrics. Lustre Networking Devices
2671 (LNDs) for the supported network fabrics have also been
2672 created for this new infrastructure.
2674 Severity : enhancement
2675 Description: Introduced Access control lists
2676 Details : clients can set ACLs on files and directories in order to have
2677 more fine-grained permissions than the standard Unix UGO+RWX.
2678 The MDS must be started with the "-o acl" mount option.
2680 Severity : enhancement
2681 Description: Introduced filesystem quotas
2682 Details : Administrators may now establish per-user quotas on the
2685 Severity : enhancement
2687 Description: Configuration change for the XT3
2688 The PTLLND is now used to run Lustre over Portals on the XT3
2689 The configure option(s) --with-cray-portals are no longer used.
2690 Rather --with-portals=<path-to-portals-includes> is used to
2691 enable building on the XT3. In addition to enable XT3 specific
2692 features the option --enable-cray-xt3 must be used.
2697 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2698 Details : A race between allocating a new llcd and re-getting the llcd_lock
2699 allowed another thread to grab newly-allocated llcd.
2701 Severity : enhancement
2703 Description: 2.6 OST async journal commit and locking fix to improve performance
2704 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2705 2.6 kernels are now async as they already were in 2.4 kernels so
2706 that they can commit concurrently with the network bulk transfer.
2707 For block-allocated files the filter allocation semaphore is held
2708 to avoid filesystem fragmentation during allocation. BKL lock
2709 removed for 2.6 xattr operations where it is no longer needed.
2714 Description: lconf incorrectly determined whether two IP networks could talk
2715 Details : In some more complicated routing and multiple-network
2716 configurations, lconf will avoid trying to make a network
2717 connection to a disjoint part of the IP space. It was doing the
2718 math incorrectly for one set of cases.
2723 Description: Fix for potential infinite loop processing records in an llog.
2724 Details : If an llog record is corrupted/zeroed, it is possible to loop
2725 forever in llog_process(). Validate the llog record length
2726 and skip the remainder of the block on error.
2729 Frequency : occasional (liblustre only)
2731 Description: liblustre could not open files whose last component is a symlink
2732 Details : sysio_path_walk() would incorrectly pass the open intent to
2733 intermediate path components.
2736 Frequency : rare (liblustre only with non-standard tuning)
2737 Bugzilla : 7201 (7350)
2738 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2739 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2740 after it has been released, only do this for referenced locks
2742 Severity : enhancement
2744 Description: specifying an (invalid) directory default stripe_size of -1
2745 would reset the directory default striping
2746 Details : stripe_size -1 was used internally to signal directory stripe
2747 removal, now use "all default" to signal dir stripe removal
2748 as a directory striping of "all default" is not useful
2751 Frequency : common for large clusters running liblustre clients
2753 Description: doing an ls when liblustre clients are running is slow
2754 Details : sending a glimpse AST to a liblustre client waits for every AST
2755 to time out, as liblustre clients will not respond. Since they
2756 cannot cache data we refresh the OST lock LVB from disk instead.
2758 Severity : enhancement
2760 Description: doing an ls at the same time as file IO can be slow
2761 Details : enqueue and other "small" requests can be blocked behind many
2762 large IO requests. Create a new OST IO portal for non-IO
2763 requests so they can be processed faster.
2766 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2768 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2769 created file with mode 000
2770 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2771 allow an ftruncate() as a non-root user to a file with mode 000.
2772 Lustre now respects this flag to disable mode checks when
2773 truncating a file owned by the user
2776 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2779 Description: Revoking locks from clients that went dead or catatonic might take
2781 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2782 cancellation of such locks instant on servers without waiting for
2783 any reply from clients. Clients drops these locks when cancel
2784 notification from server is received without replying.
2787 Frequency : liblustre-only, when liblustre client dies or becomes busy
2789 Description: Doing ls on Linux clients can take a long time with active
2791 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2792 granting such locks to it in the first place if possible. Locks
2793 are taken by proxy on the OST during the read or write and
2794 dropped immediately afterward. Add connect flags handling, do
2795 not grant locks to liblustre clients for glimpse ASTs.
2797 Severity : enhancement
2799 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2800 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2801 sequential reads and behaves badly with applications doing random
2802 reads. Improve it by reading ahead at least read region, and
2803 avoiding excessive large RPC for small reads.
2805 Severity : enhancement
2807 Description: Creating more than 1000 files for a single job may cause a load
2808 imbalance on the OSTs if there are also a large number of OSTs.
2809 Details : qos_prep_create() uses an OST index reseed value that is an
2810 even multiple of the number of available OSTs so that if the
2811 reseed happens in the middle of the object allocation it will
2812 still utilize the OSTs as uniformly as possible.
2817 Description: OST or MDS may oops in ping_evictor_main()
2818 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2819 but doesn't restart at beginning of obd_exports_timed list
2822 Severity : enhancement
2824 Description: improve by-nid export eviction on the MDS and OST
2825 Details : allow multiple exports with the same NID to be evicted at one
2826 time without re-searching the exports list.
2829 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2831 Description: MDS may oops in groups_free()
2832 Details : in rare race conditions a newly allocated group_info struct is
2833 freed again, and this can be NULL. The 2.4 compatibility code
2834 for groups_free() checked for a NULL pointer, but 2.6 did not.
2837 Frequency : common for liblustre clients doing little filesystem IO
2838 Bugzilla : 9352, 7313
2839 Description: server may evict liblustre clients accessing contended locks
2840 Details : if a client is granted a lock or receives a completion AST
2841 with a blocking AST already set it would not reply to the AST
2842 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2845 Frequency : lfs setstripe, only systems with more than 160 OSTs
2847 Description: unable to set striping with a starting offset beyond OST 160
2848 Details : llapi_create_file() incorrectly limited the starting stripe
2849 index to the maximum single-file stripe count.
2852 Frequency : LDAP users only
2854 Description: lconf did not handle in-kernel recovery with LDAP properly
2855 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2857 Severity : enhancement
2859 Description: bind OST threads to NUMA nodes to improve performance
2860 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2861 node and do their allocations there to localize memory access
2863 Severity : enhancement
2865 Description: llmount can determine client NID directly from Myrinet (GM)
2866 Details : the client NID code from gmnalnid was moved directly into
2867 llmount, removing the need to use this or specifying the
2868 client NID explicitly when mounting GM clients with zeroconf
2871 Frequency : if client is started with down MDS
2873 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2874 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2875 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2876 ptlrpc_import_delayed_req() because we didn't check if the
2877 request was interrupted, and we also didn't break out of the
2878 event loop if there was no timeout
2883 Description: data loss during non-page-aligned writes to a single file from
2884 both multiple nodes and multiple threads on one node at same time
2885 Details : updates to KMS and lsm weren't protected by common lock. Resulting
2886 inconsistency led to false short-reads, that were cached and later
2887 used by ->prepare_write() to fill in partially written page,
2888 leading to data loss.
2891 Frequency : always, if lconf --abort_recovery used
2893 Description: lconf --abort_recovery fails with 'Operation not supported'
2894 Details : lconf was attempting to abort recovery on the MDT device and not
2897 Severity : enhancement
2899 Description: remove cleanup logs
2900 Details : replace lconf-generated cleanup logs with lustre internal
2901 cleanup routines. Eliminates the need for client-cleanup and
2904 Severity : enhancement
2906 Description: add support for EAs (user and system) on lustre filesystems
2907 Details : it is now possible to store extended attributes in the Lustre
2908 client filesystem, and with the user_xattr mount option it
2909 is possible to allow users to store EAs on their files also
2911 Severity : enhancement
2913 Description: Add possibility (config option) to show minimal available OST free
2915 Details : When compiled with --enable-mindf configure option, statfs(2)
2916 (and so, df) will return least minimal free space available from
2917 all OSTs as amount of free space on FS, instead of summary of
2918 free spaces of all OSTs.
2920 Severity : enhancement
2922 Description: do not expand extent locks acquired on OST-side
2923 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
2924 by server: they are not cached anyway.
2927 Frequency : when mmap is used/binaries executed from Lustre
2929 Description: Unmmap pages before throwing them away from read cache.
2930 Details : llap_shrink cache now attempts to unmap pages before discarding
2931 them (if unmapping failed - do not discard). SLES9 kernel has
2932 extra checks that trigger if this unmapping is not done first.
2937 Description: lconf didn't resolve symlinks before checking to see whether a
2938 given mountpoint was already in use
2941 Frequency : when migrating failover services
2942 Bugzilla : 6395, 9514
2943 Description: When migrating a subset of services from a node (e.g. failback
2944 from a failover service node) the remaining services would
2945 time out and evict clients.
2946 Details : lconf --force (implied by --failover) sets the global obd_timeout
2947 to 5 seconds in order to quickly disconnect, but this caused
2948 other RPCs to time out too quickly. Do not change the global
2949 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
2951 Severity : enhancement
2952 Frequency : if MDS is started with down OST
2953 Bugzilla : 9439,5706
2954 Description: Allow startup/shutdown of an MDS without depending on the
2955 availability of the OSTs.
2956 Details : Asynchronously call mds_lov_synchronize during MDS startup.
2957 Add appropriate locking and lov-osc refcounts for safe
2958 cleaning. Add osc abort_inflight calls in case the
2959 synchronize never started.
2962 Frequency : occasional (Cray XT3 only)
2964 Description: root not authorized to access files in CRAY_PORTALS environment
2965 Details : The client process capabilities were not honoured on the MDS in
2966 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
2967 been accessed by an authorized user then root was able to access
2968 the file on the local client also. The root user capabilities
2969 are now allowed on the MDS, as this environment has secure UID.
2972 Frequency : occasional
2974 Description: ldiskfs "too long searching" message happens too often
2975 Details : A debugging message (otherwise harmless) prints too often on
2976 the OST console. This has been reduced to only happen when
2977 there are fragmentation problems on the filesystem.
2982 Description: Division by zero in statfs when all OSCs are inactive
2983 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
2984 lov_statfs divides by value returned by lov_get_stripecnt().
2988 Bugzilla : 9489, 3273
2989 Description: First write from each client to each OST was only 4kB in size,
2990 to initialize client writeback cache, which caused sub-optimal
2991 RPCs and poor layout on disk for the first writen file.
2992 Details : Clients now request an initial cache grant at (re)connect time
2993 and so that they can start streaming writes to the cache right
2994 away and always do full-sized RPCs if there is enough data.
2995 If the OST is rebooted the client also re-establishes its grant
2996 so that client cached writes will be honoured under the grant.
3001 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3002 Details : Now I/O RPCs go to different portal number and (presumably) fast
3003 lock requests (and glimses) and other RPCs get their own service
3004 threads pool that should be able to service those RPCs
3007 Severity : enhancement
3009 Description: Ability to exchange lustre version between client and servers and
3010 issue warnings at client side if client is too old. Also for
3011 liblustre clients there is ability to refuse connection of too old
3013 Details : New 'version' field is added to connect data structure that is
3014 filled with version info. That info is later checked by server and
3018 Frequency : rare, liblustre only.
3019 Bugzilla : 9296, 9581
3020 Description: Two simultaneous writes from liblustre at offset within same page
3021 might proceed at the same time overwriting eachother with stale
3023 Details : I/O lock withing llu_file_prwv was released too early, before data
3024 actually was hitting the wire. Extended lock-holding time until
3025 server acknowledges receiving data.
3028 Frequency : extremely rare. Never observed in practice.
3030 Description: avoid generating lustre_handle cookie of 0.
3031 Details : class_handle_hash() generates handle cookies by incrementing
3032 global counter, and can hit 0 occasionaly (this is unlikely, but
3033 not impossible, because initial value of cookie counter is
3034 selected randonly). Value of 0 is used as a sentinel meaning
3035 "unassigned handle" --- avoid it. Also coalesce two critical
3036 sections in this function into one.
3038 Severity : enhancement
3040 Description: allow liblustre clients to delegate truncate locking to OST
3041 Details : To avoid overhead of locking, liblustre client instructs OST to
3042 take extent lock in ost_punch() on client's behalf. New connection
3043 flag is added to handle backward compatibility.
3045 Severity : enhancement
3046 Bugzilla : 4928, 7341, 9758
3047 Description: allow number of OST service threads to be specified
3048 Details : a module parameter allows the number of OST service threads
3049 to be specified via "options ost ost_num_threads={N}" in the
3050 OSS's /etc/modules.conf or /etc/modprobe.conf.
3054 Bugzilla : 6146, 9635, 9895
3055 Description: servers crash with bad pointer in target_handle_connect()
3056 Details : In rare cases when a client is reconnecting it was possible that
3057 the connection request was the last reference for that export.
3058 We would temporarily drop the export reference and get a new
3059 one, but this may have been the last reference and the export
3060 was just destroyed. Get new reference before dropping old one.
3062 Severity : enhancement
3063 Frequency : if client is started with failover MDS
3065 Description: Allow multiple MDS hostnames in the mount command
3066 Details : Try to read the configuration from all specified MDS
3067 hostnames during a client mount in case the "primary"
3070 Severity : enhancement
3072 Description: Stop sending data to evicted clients as soon as possible.
3073 Details : Check if the client we are about to send or are sending data to
3074 was evicted already. (Check is done every second of waiting,
3075 for which l_wait_event interface was extended to allow checking
3076 of exit condition at specified intervals).
3079 Frequency : rare, normally only when NFS exporting is done from client
3081 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3083 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3086 Frequency : always, if srand() is called before liblustre initialization
3088 Description: Liblustre uses system PRNG disturbing its usage by user application
3089 Details : Introduce internal to lustre fast and high-quality PRNG for
3090 lustre usage and make liblustre and some other places in generic
3091 lustre code to use it.
3093 Severity : enhancement
3094 Bugzilla : 9477, 9557, 9870
3095 Description: Verify that the MDS configuration logs are updated when xml is
3096 Details : Check if the .xml configuration logs are newer than the config
3097 logs stored on the MDS and report an error if this is the case.
3098 Request --write-conf, or allow starting with --old_conf.
3100 Severity : enhancement
3102 Description: Handle symlinks in the path when checking if Lustre is mounted.
3103 Details : Resolve intermediate symlinks when checking if a client has
3104 mounted a filesystem to avoid duplicate client mounts.
3109 Description: lconf can hit an error exception but still return success.
3110 Details : The lconf command catches the Command error exception at the top
3111 level script context and will exit with the associated exit
3112 status, but doesn't ensure that this exit status is non-zero.
3117 Description: failure of ptlrpc thread startup can cause oops
3118 Details : Starting a ptlrpc service thread can fail if there are a large
3119 number of threads or the server memory is very fragmented.
3120 Handle this without oopsing.
3123 Frequency : always, only if liblustre and non-default acceptor port was used
3125 Description: liblustre cannot connect to servers with non-default acceptor port
3126 Details : tcpnal_set_default_params() was not called and was therefore
3127 ignoring the environment varaible TCPNAL_PORT, as well as other
3128 TCPNAL_ environment variables
3133 Description: two objects could be created on the same OST for a single file
3134 Details : If an OST is down, in some cases it was possible to create two
3135 objects on a single OST for a single file. No problems other
3136 than potential performance impact and spurious error messages.
3140 Bugzilla : 5681, 9562
3141 Description: Client may oops in ll_unhash_aliases
3142 Details : Client dcache may become inconsistent in race condition.
3143 In some cases "getcwd" can fail if the current directory is
3149 Description: Inode refcounting problems in NFS export code
3150 Details : link_raw functions used to call d_instantiate without obtaining
3151 extra inode reference first.
3155 Bugzilla : 9942, 9903
3156 Description: Referencing freed requests leading to crash, memleaks with NFS.
3157 Details : We used to require that call to ll_revalidate_it was always
3158 followed by ll_lookup_it. Also with revalidate_special() it is
3159 possible to call ll_revalidate_it() twice for the same dentry
3160 even if first occurence returned success. This fix changes semantic
3161 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3162 reference on a request referred from the intent.
3163 ll_intent_release() then releases such a request.
3166 Frequency : rare, normally benchmark loads only
3168 Description: unlinked inodes were kept in memory on the client
3169 Details : If a client is repeatedly creating and unlinking files it
3170 can accumulate a lot of stale inodes in the inode slab cache.
3171 If there is no other client load running this can cause the
3172 client node to run out of memory. Instead flush old inodes
3173 from client cache that have the same inode number as a new inode.
3176 Frequency : SLES9 2.6.5 kernel and long filenames only
3177 Bugzilla : 9969, 10379
3178 Description: utime reports stale NFS file handle
3179 Details : SLES9 uses out-of-dentry names in some cases, which confused
3180 the lustre dentry revalidation. Change it to always use the
3184 Frequency : rare, unless heavy write-truncate concurrency is continuous
3185 Bugzilla : 4180, 6984, 7171, 9963, 9331
3186 Description: OST becomes very slow and/or deadlocked during object unlink
3187 Details : filter_destroy() was holding onto the parent directory lock
3188 while truncating+unlinking objects. For very large objects this
3189 may block other threads for a long time and slow overall OST
3190 responsiveness. It may also be possible to get a lock ordering
3191 deadlock in this case, or run out of journal credits because of
3192 the combined truncate+unlink. Solution is to do object truncate
3193 first in one transaction without parent lock, and then do the
3194 final unlink in a new transaction with the parent lock. This
3195 reduces the lock hold time dramatically.
3198 Frequency : rare, 2.4 kernels only
3200 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3201 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3202 before unmounting their filesystem in order to avoid deadlock.
3203 The kernel_locked() function in 2.4 kernels only checks whether
3204 the kernel lock is held, not whether it is this process that is
3205 holding it as 2.6 kernels do.
3210 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3211 Details : The client ptlrpc code may be trying to reconnect to a down
3212 server before a previous connection attempt has timed out.
3213 Increase the reconnect interval to be longer than the connection
3214 timeout interval to avoid sending duplicate connections to
3218 Frequency : echo_client brw_test command
3220 Description: fix echo_client to work with OST preallocated code
3221 Details : OST preallocation code (5137) didn't take echo_client IO path
3222 into account: echo_client calls filter methods outside of any
3223 OST thread and, hence, there is no per-thread preallocated
3224 pages and buffers to use. Solution: hijack pga pages for IO. As
3225 a byproduct, this avoids unnecessary data copying.
3229 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3230 Description: Client can oops in mdc_commit_close() after open replay
3231 Details : It was possible for the MDS to return an open request with no
3232 transaction number in mds_finish_transno() if the client was
3233 evicted, but without actually returning an error. Clients
3234 would later try to replay that open and may trip an assertion
3235 Simplify the client close codepath, and always return an error
3236 from the MDS in case the open is not successful.
3239 Frequency : rare, 2.6 OSTs only
3241 Description: OST may deadlock under high load on fragmented files
3242 Details : If there was a heavy load and highly-fragmented OST filesystems
3243 it was possible to have all the OST threads deadlock waiting on
3244 allocation of biovecs, because the biovecs were not released
3245 until the entire RPC IO was completed. Instead, release biovecs
3246 as soon as they are complete to ensure forward IO progress.
3248 Severity : enhancement
3250 Description: Support for specifying external journal device at mount
3251 Details : If an OST or MDS device is formatted with an external journal
3252 device, this device major/minor is stored in the ext3 superblock
3253 and may not be valid for failover. Allow detecting and
3254 specifying the external journal at mount time.
3259 Description: Mounting an MDS with pending unlinked files may cause oops
3260 Details : target_finish_recovery() calls mds_postrecov() which returned
3261 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3262 considers this an error and immediately begins cleaning up the
3263 lov, just after starting the mds_lov process
3265 Severity : enhancement
3267 Description: Implement 'lfs df' to report actual free space on per-OST basis
3268 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3269 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3270 usage of objects; '-h' to report in human readable format.
3272 ------------------------------------------------------------------------------
3274 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3281 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3282 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3283 at startup during orphan unlinking. Instead, skip buddy update
3284 before setup, as it will be regenerated after recovery is complete.
3289 Description: If an OST is inactive, its locks might reference stale inodes.
3290 Details : lov_change_cbdata() must iterate over all namespaces, even if
3291 they are inactive to clear inode references from the lock.
3293 Severity : enhancement
3294 Frequency : occasional, if non-standard max_dirty_mb used
3296 Description: Client will block write RPCs if not enough grant
3297 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3298 then the client will block writes while waiting for another RPC
3299 to complete instead of consuming its dirty limit. With change
3300 we get improved performance when max_dirty_mb is small.
3302 Severity : enhancement
3303 Bugzilla : 3389, 6253
3304 Description: Add support for supplementary groups on the MDS.
3305 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3306 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3307 lookups for user supplementary groups into a cache.
3311 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3312 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3313 for liblustre O_EXCL is enforced by MDS.
3317 Description: readdir never returns NULL in liblustre.
3318 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3319 dentry in struct dirent.
3323 Frequency : liblustre only, depends on application IO pattern
3324 Description: liblustre clients evicted if not contacting servers
3325 Details : Don't put liblustre clients into the ping_evictor list, so
3326 they will not be evicted by the pinger ever.
3328 Severity : enhancement
3330 Description: Add ability to evict clients by NID from MDS.
3331 Details : By echoing "nid:$NID" string into
3332 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3333 $NID would be instantly evicted from this MDS and from all active
3334 OSTs connected to it.
3338 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3339 Details : lookup_it_finish() used to query file size from OSTs that was not
3344 Description: service threads change working directory to that of init
3345 Details : Starting lustre service threads may pin the working directory
3346 of the parent thread, making that filesystem busy. Threads
3347 now change to the working directory of init to avoid this.
3351 Frequency : during shutdown only
3352 Description: shutdown with a failed MDS or OST can cause unmount to hang
3353 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3358 Frequency : 2.6 only
3359 Description: chmod/chown may include an extra supplementary group
3360 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3361 supplementary group and if none is specified this is used.
3364 Bugzilla : 5479 (6816)
3366 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3367 Details : If lookup is in progress on a file that is unlinked we might try
3368 to revalidate the inode and fail in revalidate after lookup is
3369 complete and ll_file_open() enqueues the open again but
3370 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3373 Frequency : always, if lconf --abort_recovery used
3375 Description: lconf --abort_recovery fails with 'Operation not supported'
3376 Details : lconf was attempting to abort recovery on the MDT device and not
3379 ------------------------------------------------------------------------------
3381 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3386 Frequency : rare (only unsupported configurations with a node running as an
3388 Bugzilla : 6514, 5137
3389 Description: Mounting a Lustre file system on a node running as an OST could
3391 Details : OSTs now preallocates memory needed to write out data at
3392 startup, instead of when needed, to avoid having to
3393 allocate memory in possibly low memory situations.
3394 Specifically, if the file system is mounted on on OST,
3395 memory pressure could force it to try to write out data,
3396 which it needed to allocate memory to do. Due to the low
3397 memory, it would be unable to do so and the node would
3398 become unresponsive.
3400 Severity : enhancement
3402 Description: Addition of lconf --service command line option
3403 Details : lconf now accepts a '--service <arg>' option, which is
3404 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3406 Severity : enhancement
3408 Description: Failover mode is now the default for OSTs.
3409 Details : By default, OSTs will now run in failover mode. To return to
3410 the old behaviour, add '--failout' to the lmc line for OSTs.
3412 Severity : enhancement
3414 Description: Health checks are now provided for MDS and OSTs
3415 Details : Additional detailed health check information on MSD and OSTs
3416 is now provided through the procfs health_check value.
3419 Frequency : occasional, depends on IO load
3421 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3422 after numerous create/delete cycles
3423 Details : The ext3 inode allocation policy would not allocate new inodes
3424 very well on the OSTs because there are no new directories
3425 being created. Instead we look for groups with free space if
3426 the parent directories are nearly full.
3431 Description: Network or server problems during mount may cause partially
3432 mounted clients instead of returning an error.
3433 Details : The config llog parsing code may overwrite the error return
3434 code during mount error handling, returning success instead
3440 Description: MDS can fail to allocate large reply buffers
3441 Details : After long uptimes the MDS can fail to allocate large reply
3442 buffers (e.g. zconf client mount config records) due to memory
3443 fragmentation or consumption by the buffer cache. Preallocate
3444 some large reply buffers so that these replies can be sent even
3445 under memory pressure.
3449 Frequency : rare (liblustre)
3450 Description: fsx running with liblustre complained that using truncate() to
3451 extend the file doesn't work. This patch corrects that issue.
3452 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3453 ATTR_SIZE and lsm use in llu_setattr_raw.
3457 Frequency : rare, only 2.6 kernels
3458 Description: Unusual file access patterns on the MDS may result in inode
3459 data being lost in very rare circumstances.
3460 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3461 code in the 2.6 kernel caused the inode and/or EA data not to
3462 be read from disk, causing single-file corruption.
3466 Frequency : rare, only 2.6 filesystems using extents
3467 Description: Heavy concurrent write and delete load may cause data corruption.
3468 Details : It was possible under high-load situations to have an extent
3469 metadata block in the block device cache from a just-unlinked
3470 file overwrite a newly-allocated data block. We now unmap any
3471 metadata buffers that alias just-allocated data blocks.
3475 Frequency : filesystems with default stripe_count larger than 77
3476 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3477 Details : lconf specifies an inode size of 4096 bytes when the default
3478 stripe_count is larger than 77. This conflicts with the default
3479 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3480 this case to avoid pinning too much memory for large EAs.
3482 ------------------------------------------------------------------------------
3484 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3489 Frequency : rare (extremely heavy IO load with hundreds of clients)
3491 Description: Client is evicted, gets IO error writing to file
3492 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3493 caused clients to be evicted for AST timeouts. The fixes in
3494 bug 5192 mean we no longer need to have such short AST timeouts
3495 so ldlm_timeout has been increased.
3498 Frequency : occasional during --force or --failover shutdown under load
3499 Bugzilla : 5949, 4834
3500 Description: Server oops/LBUG if stopped with --force or --failover under load
3501 Details : a collection of import/export refcount and cleanup ordering
3502 issues fixed for safer force cleanup
3505 Frequency : only filesystems larger than 120 OSTs
3506 Bugzilla : 5990, 6223
3507 Description: lfs getstripe would oops on a very large filesystem
3508 Details : lov_getconfig used kfree on vmalloc'd memory
3511 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3512 Bugzilla : 6242, 6243
3513 Description: reading from files that had been truncated to a non-zero size
3514 but never opened returned no data
3515 Details : ll_file_read() reads zeros from no-object files to EOF
3520 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3521 Details : There's little harm in aborting MDS/OSS recovery and letting it
3522 try again, so I removed the LASSERT and return an error instead.
3524 Severity : enhancement
3526 Description: New debugging infrastructure for tracking down data corruption
3527 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3528 (b) cover more of the client-side code path, and (c) try to narrow
3529 down where problems occurred
3533 Bugzilla : 3819, 4364, 4397, 6313
3534 Description: Racing close and eviction MDS could cause assertion in mds_close
3535 Details : It was possible to get multiple mfd references during close and
3536 client eviction, leading to one thread referencing a freed mfd.
3538 Severity: : enhancement
3539 Bugzilla : 3262, 6359
3540 Description: Attempts to reconnect to servers are now more aggressive.
3541 Details : This builds on the enhanced upcall-less recovery that was added
3542 in 1.4.2. When trying to reconnect to servers, clients will
3543 now try each server in the failover group every 10 seconds. By
3544 default, clients would previously try one server every 25 seconds.
3549 Description: After recovery, certain operations trigger a failed
3550 assertion on a client.
3551 Details : Failing over an mds, using lconf -d --failover, while a
3552 client was doing a readdir() call would cause the client to
3553 LBUG after recovery completed and the readdir() was resent.
3555 Severity : enhancement
3557 Description: Default groups are now added by lconf
3558 Details : You can now run lconf --group <servicename> without having to
3559 manually add groups with lmc.
3562 Frequency : occasional
3564 Description: Nodes with an elan id of 0 trigger a failed assertion
3567 Frequency : always when accessing e.g. tty/console device nodes
3569 Description: tty and some other devices nodes cannot be used on lustre
3570 Details : file's private_data field is used by device data and lustre
3571 values in there got lost. New field was added to struct file to
3572 store fs-specific private data.
3575 Frequency : when exporting Lustre via NFS
3577 Description: NFSD failed occasionally when looking up a path component
3578 Details : NFSD is looking up ".." which was broken in ext3 directories
3579 that had grown large enough to become hashed.
3582 Frequency : Clusters with multiple interfaces not on the same subnet
3584 Description: Nodes will repeatedly try to reconnect to an interface which it
3585 cannot reach and report an error to the log.
3586 Details : Extra peer list entries will be created by lconf with some peers
3587 unreachable. lconf now validates the peer before adding it.
3590 Frequency : Only if a default stripe is set on the filesystem root.
3592 Description: Setting a default stripe on the filesystem root prevented the
3593 filesystem from being remounted.
3594 Details : The client was sending extra request flags in the root getattr
3595 request and did not allocate a reply buffer for the dir EA.
3598 Frequency : occasional, higher if lots of files are accessed by one client
3599 Bugzilla : 6159, 6097
3600 Description: Client trips assertion regarding lsm mismatch/magic
3601 Details : While revalidating inodes the VFS looks up inodes with ifind()
3602 and in rare cases can find an inode that is being freed.
3603 The ll_test_inode() code will free the lsm during ifind()
3604 when it finds an existing inode and then the VFS later attaches
3605 this free lsm to a new inode.
3609 Bugzilla : 6422, 7030
3610 Description: MDS deadlock between mkdir and client eviction
3611 Details : Creating a new file via mkdir or mknod (starting a transaction
3612 and getting the ns lock) can deadlock with client eviction
3613 (gets ns lock and trying to finish a synchronous transaction).
3616 Frequency : occasional
3617 Description: While starting a server, the fsfilt_ext3 module could not be
3619 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3620 kernels. Previously, lconf would still use the ext3 name
3621 when trying to load modules. Now, it will correctly use
3622 ext3 on 2.4 and ldiskfs on 2.6.
3624 Severity : enhancement
3625 Description: The default stripe count has been changed to 1
3626 Details : The interpretation of the default stripe count (0, to lfs
3627 or lmc) has been changed to mean striping across a single
3628 OST, rather than all available. For general usage we have
3629 found a stripe count of 1 or 2 works best.
3631 Severity : enhancement
3632 Description: Add support for compiling against Cray portals.
3633 Details : Conditional compiling for some areas that are different
3637 Frequency : occasional
3638 Bugzilla : 6409, 6834
3639 Description: Creating files with an explicit stripe count may lead to
3640 a failed assertion on the MDS
3641 Details : If some OSTs are full or unavailable, creating files may
3642 trigger a failed assertion on the MDS. Now, Lustre will
3643 try to use other servers or return an error to the
3647 Frequency : occasional
3649 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3650 caused lock timeout during readahead (since 1.4.2).
3651 Details : Processes doing readahead might match a lock that hasn't been
3652 granted yet if there are overlapping and conflicting lock
3653 requests. The readahead process waits on ungranted lock
3654 (original lock is CBPENDING), while OST waits for that process
3655 to cancel CBPENDING read lock and eventually evicts client.
3657 Severity : enhancement
3659 Description: Initial enabling of flock support for clients
3660 Details : Implements fcntl advisory locking and file status functions.
3661 This feature is provided as an optional mount flag (default
3662 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3663 locking are implemented yet, and those that are are not guaranteed
3664 to be completely correct in production environments.
3665 mount -t lustre -o [flock|noflock] ...
3668 Frequency : occasional
3670 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3671 assertion in the ext3 JBD (journaling) layer.
3672 Details : The b_committed_data struct is protected by the big kernel lock
3673 in 2.4 kernels, serializing journal_commit_transaction() and
3674 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3675 access to this struct. In 2.6 kernels there is finer grained
3676 locking to improve SMP performance of the JBD layer.
3680 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3685 Description: merge and cleanup kernel patches.
3688 -----------------------------------------------------------------------------
3690 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3692 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3693 versions of Lustre. Please update all servers and clients to
3694 version 1.4.2 or later at the same time. You must also run
3695 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3696 to update the configuration logs.
3698 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3699 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3700 - swab llog records of type '0' so we get proper header size/idx (5861)
3701 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3702 - fix rename of one directory over another leaking an inode (5953)
3703 - avoid SetPageDirty on 2.6 (5981)
3704 - don't re-add just-being-destroyed locks to the waiting list (5653)
3705 - when creating new directories, inherit the parent's custom
3706 striping settings if present parent (3048)
3707 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3708 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3709 - don't hold client locks on temporary worklist from l_lru (5666)
3710 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3711 - automatically evict dead clients (5921)
3712 - Update file size properly in create+truncate+fstat case (6196)
3713 - Do not unhash mountpoint dentries, do not allow removal of
3715 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3716 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3717 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3718 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3719 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3720 - Correctly handle reads of files with no objects (6243)
3721 - lctl recover will also mark a device active if deactivate used (5933)
3723 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3724 - allow --write-conf on an MDS with different nettype than client (5619)
3725 - don't write config llogs to MDS for mounts not from that MDS (5617)
3726 - lconf should create multiple TCP connections from a client (5201)
3727 - init scripts are now turned off by default; run chkconfig --on
3728 lustre and chkconfig --on lustrefs to use them
3729 - upcalls are no longer needed for clients to recover to failover
3731 - add --abort-recovery option to lconf to abort recovery on device
3733 - add support for an arbitrary number of OSTs (3026)
3734 - Quota support protocol changes.
3735 - forward compatibility changes to wire structs (6007)
3736 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3737 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3738 - improved "lustre status" script
3739 - initialize blocksize for non-regular files (6062)
3740 - added --disable-server and --disable-client configure options (5782)
3741 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3742 - Vanilla 2.4.29 support
3743 - increase maximum number of obd devices to 520 (6242)
3744 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3745 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3746 SLES 9 2.6.5 kernels
3747 - limit stripes per file to 160 (the maximum EA size) (6093)
3749 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3752 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3753 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3754 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3755 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3756 - small ext3 extents cleanups and fixes (5733)
3757 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3758 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3759 - avoid needless client->OST connect, fix handle mismatch (5317)
3760 - fix DLM error path that led to out-of-sync client, long delays (5779)
3761 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3762 - fix several locking issues related to i_size (5492,5624,5654,5672)
3763 - don't move pending lock onto export if it is already evicted (5683)
3764 - fix kernel oops when creating .foo in unlinked directory (5548)
3765 - fix deadlock in obdfilter statistics vs. object create (5811)
3766 - use time_{before,after} to avoid timer jiffies wrap (5882)
3767 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3768 - Do not leak request if server was not able to process it (5154)
3769 - If mds_open unable to find parent dir, make that negative lookup(5154)
3770 - don't create new directories with extent-mapping (5909, 5936)
3772 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3773 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3774 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3775 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3776 - support for RHEL4 kernel on i686 (5773)
3777 - provide error messages when incompatible logs are encountered (5898)
3779 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3780 * version 1.4.0.10 (1.4.1 release candidate 1)
3782 - don't keep a lock reference when lock is not granted (4238)
3783 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3784 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3785 - drop import inflight refcount on signal_completed_replay error (5255)
3786 - unlock page after async write error during send (3677)
3787 - handle missing objects in filter_preprw_read properly (5265)
3788 - no transno return for symlink open, don't save no-trasno open (3440)
3789 - don't try to complete elan receive that already failed (4012)
3790 - free RPC server reply state on error (5406)
3791 - clean up thread from ptlrpc_start_thread() on error (5160)
3792 - readahead could read extra page into cache that wasn't ejected (5388)
3793 - prevent races in class_attach/setup/cleanup/detach (5260)
3794 - don't dereference de->d_inode after l_dput of de (5458)
3795 - use "int" for stripe value returned from lock_to_stripe (5544)
3796 - mballoc allocation and error-checking fixes in 2.6 (5504)
3797 - block device patches to fix I/O request sizes in 2.6 (5482)
3798 - look up hostnames for IB nals (5602)
3799 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3800 - don't start multiple acceptors for the same port (5277)
3801 - fix incorrect LASSERT in mds_getattr_name (5635)
3802 - export a proc file for general "ping" checking (5628)
3803 - fix "lfs check" to not block when the MDS is down (5628)
3805 - service request history (4965)
3806 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3807 - create an "evict_client" /proc entry on OSTs, like the MDS has
3808 - fix mount usage message, return errors per mount(8) (5168)
3809 - change grep [] to grep "[]" in tests so they work in more UMLs
3810 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3811 - remove ancient LOV_MAGIC_V0 EA support (5047)
3812 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3813 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3814 - allow readahead allocations to fail when low on memory (5383)
3815 - mmap locking landed again, after considerable improvement (2828)
3816 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3818 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3821 - send OST transaction number in read/write reply to free req (4966)
3822 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3823 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3824 - the watchdog thread now runs as interruptible (5246)
3825 - flock/lockf fixes (but it's still disabled, pending 5135)
3826 - don't use EXT3 constants in llite code (5094)
3827 - memory shortage at startup could cause assertion (5176)
3829 - reorganization of lov code
3830 - single portals codebase
3832 - add extents/mballoc support (5025)
3833 - direct I/O reads in the obdfilter (4048)
3834 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3836 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3839 - send OST transaction number in read/write reply to free req (4966)
3840 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3841 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3842 - the watchdog thread now runs as interruptible (5246)
3843 - handle missing objects in filter_preprw_read properly (5265)
3844 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3845 - drop import inflight refcount on signal_completed_replay error (5255)
3846 - unlock page after async write error during send (3677)
3847 - return original error code on reconstructed replies (3761)
3848 - no transno return for symlink open, don't save no-trasno open (3440)
3850 - add pid to ldlm debugging output (4922)
3851 - bump the watchdog timeouts -- we can't handle 30sec yet
3852 - extra debugging for orphan dentry/inode bug (5259)
3854 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3857 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3858 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3859 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3860 - fix NULL dereference in /proc/sys/portals/routes (4827)
3861 - allow failed mdc_close() operations to be interrupted (4561)
3862 - stop precreate on OST before MDS would time out on it (4778)
3863 - don't send partial-page writes before EOF from client (4410)
3864 - discard client grant for sub-page writes on large-page clients (4520)
3865 - don't free dentries not owned by NFS code, check generation (4806)
3866 - fix lsm leak if mds_create_objects() fails (4801)
3867 - limit debug_daemon file size, always print CERROR messages (4789)
3868 - use transno after validating reply (3892)
3869 - process timed out requests if import state changes (3754)
3870 - update mtime on OST during writes, return in glimpse (4829)
3871 - add mkfsoptions to LDAP (4679)
3872 - use ->max_readahead method instead of zapping global ra (5039)
3873 - don't interrupt __l_wait_event() during strace
3875 - add software watchdogs to catch hung threads quickly (4941)
3876 - make lustrefs init script start after nfs is mounted
3877 - fix CWARN/ERROR duplication (4930)
3878 - return async write errors to application if possible (2248)
3879 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3880 - print NAL number in %x format (4645)
3881 - update barely-supported suse-2.4.21-171 series (4842)
3882 - support for sles 9 %post scripts
3883 - support for building 2.6 kernel-source packages
3884 - support for sles km_* packages
3886 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
3889 - ignore -ENOENT errors in osc_destroy (3639)
3890 - notify osc create thread that OSC is being cleaned up (4600)
3891 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
3892 - reconstruct ost_handle() like mds_handle() (4657)
3893 - create a new thread to do import eviction to avoid deadlock (3969)
3894 - let lconf resolve symlinked-to devices (4629)
3895 - don't unlink "objects" from directory with default EA (4554)
3896 - hold socknal file ref over connect in case target is down (4394)
3897 - allow more than 32000 subdirectories in a single directory (3244)
3898 - fix blocks count for O_DIRECT writes (3751)
3899 - OST returns ENOSPC from object create when no space left (4539)
3900 - don't send truncate RPC if file size isn't changing (4410)
3901 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
3902 - bind to privileged port in socknal and tcpnal (3689)
3904 - rate limit CERROR/CWARN console message to avoid overload (4519)
3905 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
3906 - basic mmap support (3918)
3907 - kernel patch series update from b1_4 (4711)
3909 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
3912 - avoid crash during MDS cleanup with OST shut down (2775)
3913 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
3914 - don't use bad inodes on the MDS (3744)
3915 - dynamic object preallocation to improve recovery speed (4236)
3916 - don't hold spinlock over lock dumping or change debug flags (4401)
3917 - don't zero obd_dev when it is force cleaned (3651)
3918 - print grants to console if they go negative (4431)
3919 - "lctl deactivate" will stop automatic recovery attempts (3406)
3920 - look for existing locks in ldlm_handle_enqueue() (3764)
3921 - don't resolve lock handle twice in recovery avoiding race (4401)
3922 - revalidate should check working dir is a directory (4134)
3924 - don't always mark "slow" obdfilter messages as errors (4418)
3926 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
3929 - don't close LustreDB during write_conf until it is done (3860)
3930 - fix typo in lconf for_each_profile (3821)
3931 - allow dumping logs from multiple threads at one time (3820)
3932 - don't allow multiple threads in OSC recovery (3812)
3933 - fix debug_size parameters (3864)
3934 - fix mds_postrecov to initialize import for llog ctxt (3121)
3935 - replace config semaphore with spinlock (3306)
3936 - be sure to send a reply for a CANCEL rpc with bad export (3863)
3937 - don't allow enqueue to complete on a destroyed export (3822)
3938 - down write_lock before checking llog header bitmap (3825)
3939 - recover from lock replay timeout (3764)
3940 - up llog sem before sending rpc (3652)
3941 - reduce ns lock hold times when setting kms (3267)
3942 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
3943 - fix NULL deref and obd_dev leak on setup error (3312)
3944 - replace some LBUG about llog ops with error handling (3841)
3945 - don't match INVALID dentries from d_lookup and spin (3784)
3946 - hold dcache_lock while marking dentries INVALID and hashing (4255)
3947 - fix invalid assertion in ptlrpc_set_wait (3880)
3949 - add libwrap support for the TCP acceptor (3996)
3950 - add /proc/sys/portals/routes for non-root route listing (3994)
3951 - allow setting MDS UUID in .xml (2580)
3952 - print the stack of a process that LBUGs (4228)
3954 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
3957 - don't cleanup request in ll_file_open() on failed MDS open (3430)
3958 - make sure to unset replay flag from failed open requests (3440)
3959 - if default stripe count is 0, use OST count for inode size (3636)
3960 - update parent mtime/ctime on client for create/unlink (2611)
3961 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
3962 - free recovery state on server during a forced cleanup (3571)
3963 - unregister_reply for resent reqs (3063)
3964 - loop back devices mounting and status check on 2.6 (3563)
3965 - fix resource-creation race that can provoke i_size == 0 (3513)
3966 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
3967 - more debugging for page-accounting assertion (3746)
3968 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
3969 - avoid deadlock after precreation failure (3758)
3970 - fix race and lock order deadlock in orphan handling (3450, 3750)
3971 - add validity checks when grabbing inodes from l_ast_data (3599)
3973 - add /proc/.../recovery_status to obdfilter (3428)
3974 - lightweight CDEBUG infrastructure, debug daemon (3668)
3975 - change default OSC RPC parameters to be better on small clusters
3976 - turn off OST read cache for files smaller than 32MB
3977 - install man pages and include them in rpms (3100)
3978 - add new init script for (un)mounting lustre filesystems (2593)
3979 - run chkconfig in %post for init scripts (3701)
3980 - drop scimac NAL (unmaintained)
3982 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
3985 - clean kiobufs before and after use (3485)
3986 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
3987 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
3988 - change init script to fail more clearly if not run as root (1528)
3989 - allow clients to reconnect during replay (1742)
3990 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
3991 - don't do DNS lookups on NIDs too small for IP addresses (3442)
3992 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
3993 - fix cond_resched (3554)
3994 - only evict unfinished clients after recovery (3515)
3995 - allow bulk resend, prevent data loss (3570)
3996 - dynamic ptlrpc request buffer allocation (2102)
3997 - don't allow unlinking open directory if it isn't empty (2904)
3998 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
3999 - remove extraneous obd dereference causing LASSERT failure (3334)
4000 - don't use get_cycles() when creating temp. files on the mds (3156)
4001 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4002 - handle EEXIST for set-stripe, set proper directory name (3336)
4004 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4005 - fix ksocknal_fmb_callback() error messages (2918)
4007 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4010 - don't copy lvb into (possibly NULL) reply on error (2983)
4011 - don't deref dentry after dput, don't free lvb on error (2922)
4012 - use the kms to determine writeback rpc length (2947)
4013 - increment oti_logcookies when osc is inactive (2948)
4014 - update client's i_blocks count via lvb messages (2543)
4015 - handle intent open/close of special files properly (1557)
4016 - mount MDS with errors=remount-ro, like obdfilter (2009)
4017 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4018 - don't use cancelling-locks' kms values (2947)
4019 - use highest lock extent for kms, not last one (2925)
4020 - don't dereference ERR_PTR() dentry in error handling path (3107)
4021 - fix thread race in portals_debug_dumplog() (3122)
4022 - create lprocfs device entries at setup instead of at attach (1519)
4023 - common AST error handler, don't evict client on completion race (3145)
4024 - zero nameidata in detach_mnt in 2.6 (3118)
4025 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4026 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4027 - initialize RPC timeout timer earlier for 2.6 (3219)
4028 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4029 - print nal/nid for unknown nid (3258)
4030 - additional checks for oscc recovery before doing precreate (3284)
4031 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4032 - don't crash in mdc_close for bad permissions on open (3285)
4033 - zero i_rdev for non-device files (3147)
4034 - clear page->private before handing to FS, better assertion (3119)
4035 - tune the read pipeline (3236)
4036 - fix incorrect decref of invalidated dentry (2350)
4037 - provide read-ahead stats and refine rpc in flight stats (3328)
4038 - don't hold journal transaction open across create RPC (3313)
4039 - update atime on MDS at close time (3265)
4040 - close LDAP connection when recovering to avoid server load (3315)
4041 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4042 - don't leak open file on MDS after open resend (3325)
4043 - serialize filter_precreate and filter_destroy_precreated (3329)
4044 - loop device shouldn't call sync_dev() for nul device (3092)
4045 - clear page cache after eviction (2766)
4046 - resynchronize MDS->OST in background (2824)
4047 - refuse to mount the same filesystem twice on same mountpoint (3394)
4048 - allow llmount to create routes for mounting behind routers (3320)
4049 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4050 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4051 - fix rare problem with rename on htree directories (3417)
4053 - allow default OST striping configuration per directory (1414)
4054 - fix compilation for qswnal for 2.6 kernels (3125)
4055 - increase maximum number of MDS request buffers for large systems
4056 - change liblustreapi to be useful for external progs like lfsck (3098)
4057 - increase local configuration timeout for slow disks (3353)
4058 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4060 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4063 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4064 - don't overwrite extent policy data in reply if lock was blocked (2901)
4065 - drop filter export grants atomically with removal from device (2663)
4066 - del obd_self_export from work_list in class_disconnect_exports (2908)
4067 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4068 - swab reply message in mdc_close, other PPC fixes (2464)
4069 - fix destroying of named logs (2325)
4070 - overwrite old logs when running lconf --write_conf (2264)
4071 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4072 - fix race in target_handle_connect (2898)
4073 - mds_reint_create() should take same inode create lock (2926)
4074 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4075 - don't close files for self_export to avoid uninitialized obd (2936)
4076 - allow MDS with the same name as client node (2939)
4077 - hold dentry reference for closed log files for unlink (2325)
4078 - reserve space for all logs during transactions (2059)
4079 - don't evict page beyond end of stripe extent (2925)
4080 - don't oops on a deleted current working directory (2399)
4081 - handle hard links to targets without a parent properly (2517)
4082 - don't dereference NULL lock when racing during eviction (2867)
4083 - don't grow lock extents when lots of conflicting locks (2919)
4085 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4088 - account for cache space usage on clients to avoid data loss (974)
4089 - lfsck support in lustre kernel code (2349)
4090 - reduce journal credits needed for BRW writes (2370)
4091 - orphan handling to avoid losing space on client/server crashes
4092 - ptlrpcd can be blocked, stopping ALL progress (2477)
4093 - use lock value blocks to assist in proper KMS, faster stat (1021)
4094 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4095 - recovery for initial connections (2355)
4096 - fixes for mds_cleanup_orphans (1934)
4097 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4098 - block all file creations until orphan recovery completes (1901)
4099 - client remove rq_connection from request struct (2423)
4100 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4101 - recovery timer race (2670)
4102 - mdc_close recovey bug (2532)
4103 - ptlrpc cleanup bug (2710)
4104 - mds timeout on local locks (2588)
4105 - namespace lock held during RPCs (2431)
4106 - handle interrupted sync write properly (2503)
4107 - don't try to handle a message that hasn't been replied to (2699)
4108 - client assert failure during cleanup after abort recovery (2701)
4109 - leak mdc device after failed mount (2712)
4110 - ptlrpc_check_set allows timedout requests to complete (2714)
4111 - wait for inflight reqs when ptlrpcd finishes (2710)
4112 - make sure unregistered services are removed from the srv_list
4113 - reset bulk XID's when resending them (caught by 1138 test)
4114 - unregister_bulk after timeout
4115 - fix lconf error (2694)
4116 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4117 - readahead locks pages, leaves pending causing memory pressure (2673)
4118 - increase OST request buffers to 4096 on large machines (2729)
4119 - fix up permission of existing directories in simple_mkdir (2661)
4120 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4121 - don't assign transno to errored transactions (2742)
4122 - don't delete objects on OST if given a bogus objid from MDS (2751)
4123 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4124 - if rq_no_resend, then timeout request after recovery (2432)
4125 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4126 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4127 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4128 - MDS recovery shouldn't delete objects that it has given out (2730)
4129 - if enqueue arrives after completion, don't clobber LVB (2819)
4130 - don't unlock pages twice when trigger_group_io returns error (2814)
4131 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4132 - don't write pages to disk if there was an error (1450)
4133 - don't ping imports that have recovery disabled (2676)
4134 - take buffered bytes into account when balancing socknal conn (2817)
4135 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4136 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4137 - remove little-endian swabbing of llog records (1987)
4138 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4139 - retry reposting request buffers if they fail (1191)
4140 - grow extent at grant time to avoid granting a revoked lock (2809)
4141 - lock revoke doesn't evict page if covered by a second lock (2765)
4142 - disable VM readahead to avoid reading outside lock extents (2805)
4144 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4145 - updated kernel patches for hp-2.4.20 kernel (2681)
4147 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4150 - fix truncated write corruption (2366)
4151 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4153 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4154 - ensure len(uuid) < 37 in lmc (1171)
4155 - fix ia64 OOPS in llog_test (2255)
4156 - zero end of page at obdfilter for partial page writes (2648)
4157 - don't leave stale dentries around after renames (bug 2428)
4158 - fix timeouts when evicting a client with a single lock held (2642)
4159 - set deadline for the initial HELLO message to drain (2634)
4160 - print out dotted-quad IP addresses in the socknal (2302)
4162 - additional debugging for MDS client eviction problem (2443)
4163 - fix mkfsoptions support for osts (2603, 2604)
4165 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4168 - add series for the vanilla 2.6.0 kernel
4169 - add series for the vanilla 2.4.24 kernel
4170 - add series for a cray x86/64 UL kernel drop
4171 - fix xattr patches for the vanilla 2.4.19 series
4173 - generate true UUIDs in lmc (1171)
4174 - have portals stack dumping break in UML (2466)
4175 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4176 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4177 - wake callback waiting threads on client eviction (2460)
4178 - Add --ptldebug and --subsystem to lmc (1719)
4179 - update assertion to allow safe interrupt allocation
4180 - set rq_no_resend for cancel requests (2432)
4181 - recalculate ptlrpcd timeout after resend (2494)
4182 - call vfs_rmdir when removing pending directories (2368)
4183 - fix renaming a file to itself (2429)
4184 - lmc creates a default one-stripe lov (2454)
4185 - expand procfs space to handle large clusters (2326)
4186 - increase UML stack to avoid overflow
4187 - update lconf's list of debug and subsystem masks
4188 - fix lfs find --obd (2510)
4189 - /proc tunable for disabling filter read caching (2591)
4190 - stop rpm packages from altering slapd.conf (2301)
4191 - disable nagle in the socknal under 0conf (2578)
4192 - choose mds inode size based on stripe count (2572)
4193 - fix kernel-source rpm problems (2516)
4195 - add --disable-doc to avoid pdf generation (2421)
4196 - update documentation, tests, type-os, comments
4197 - avoid format warnings on ia64
4198 - remove the TOE NAL
4199 - tiny code cleanups by removing unused fields
4201 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4204 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4205 - remove the most blatant lies from BUILDING file (2371)
4206 - change default debug level to reasonable production setting
4207 - reduce client side cache size to reduce cache flush time
4208 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4209 - make TCP zerocopy and pinger support enabled by default (2476)
4210 - sync writes completed after process exits caused crashes (2319)
4211 - maintain correct mount count on the MDS (2356)
4212 - backout 1557, because 2316 wasn't really fixed
4213 - better file I/O statistics gathering in /proc
4214 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4215 - another kernel patch to fix zero-copy TCP function export
4216 - don't take duplicate lock when processing re-sent getattr (2420)
4217 - lctl uses obd_self_export instead of creating new conn (2353)
4218 - MDS/OST recovery case which requires object creation asserted (2425)
4219 - move lfs from /usr/sbin to /usr/bin in packages
4220 - fix race between mds_client_add and mds_client_free (2417)
4221 - use kmalloc instead of slabs in portals (2430)
4222 - don't create duplicate records when a failover MDS is present (2442)
4223 - remove unnecessary mount age check (2332)
4224 - don't remove directory inodes from locks prematurely (2451)
4225 - don't break if MDS service name is the same as hostname (2103)
4226 - fix races in client write RPC generation when cache full (2482)
4228 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4231 - remove now-unused request->rq_obd (278)
4232 - if an allocation fails, print out how much memory we've used (1933)
4233 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4234 - add the "configurable stack size" patch to most series files (1256)
4235 - ability to write large log records, for 100+ OST configs (2306)
4236 - fix NULL deref when filter_prep fails (2314)
4237 - fix operator precedence error in filter_sync
4238 - dynamic allocation of socknal TX descriptors (2315)
4239 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4240 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4241 - ensure that CWARN messages in Portals always get to the syslog
4242 - __init/__exit are not for prototype decls (ldlm_init/exit)
4243 - x86-64 compile warning fixes
4244 - fix gateway LMC keyword conflict (2318)
4245 - fix MDS lock inversions in getattr/reint paths (1844)
4246 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4247 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4248 - DLM race condition prevented some lock evictions (2328)
4249 - ENOMEM detection and retry on socknal sends (2230)
4250 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4251 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4252 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4253 - better error messages when a client is rejected during recovery (1505)
4254 - avoid cancelling locks which were never granted, after failure (2330)
4255 - fix i_sem/journal inversion in mds_client_add (2333)
4256 - fix truncate/getattr lock cycle deadlock (2334)
4257 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4258 - fix two rare exit paths which could leak an l_lock() ref (2321)
4259 - fencepost error in MDS/OST orphan recovery (2226)
4260 - make log record alignment 8 bytes (1988)
4261 - lstripe now fails when requested offset > ost_count (2237)
4262 - ensure that all kernel series have a complete list.h (1607)
4263 - fix crashes in special-file operations (2316)
4264 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4266 - allow configurable automake binary, for testing new versions
4267 - small update to the lfs documentation
4269 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4271 * fix negative export reference count in fsfilt_sync (2312)
4273 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4274 * release candidate 0.9.1
4276 - orphans are moved into the PENDING directory for possible recovery
4277 - replayed opens now open by fid for orphan/rename safety (1042)
4278 - last close of an orphan inode generates a transno (683)
4279 - chdir() and mount() now pin the directory entry (1020)
4280 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4281 - discard very old requests without processing them (1502)
4282 - remove some common, well-understood CERRORs (1505)
4283 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4284 - clear "grant" flags in OST replies until OST grant code lands (1644)
4285 - fix read performance by not clobbering i_blksize on client (1598)
4286 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4287 - make LRU size a /proc tunable, clears locks when reduced (707)
4288 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4289 - server locks take references on exports now (1558)
4290 - build fixes for 2.4.20-rh trees (1663)
4291 - return an error from lov_create if all OSCs are inactive (1751)
4292 - fix import levels when a reconnect happens without a timeout (1597)
4293 - exit early from mds_open if we get a lookup error (1749)
4294 - partial page read at EOF wouldn't wait for disk before sending (1642)
4295 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4296 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4297 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4298 - write/truncate lock inversion (1639)
4299 - Don't auto-load obdclass, portals modules during cleanup (1495)
4300 - fix timestamps from jumping to "now" (1763)
4301 - extra journal assertions (1648)
4302 - add an extra multiunlink test (1771)
4303 - fix read_record/write_record API (1776)
4304 - fix leak of offset_extent, possible incorrect i_size later (1772)
4305 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4306 - Debugging for the kqswnal_get_idle_tx problems (1820)
4307 - Allow recovery to be attempted multiple times (1536)
4308 - Write out MDS last_rcvd file after it is first created (1600)
4309 - Fix tx_descriptor leak in failed transmit situations (1827)
4310 - ext3 journaling fixes for assertion failure after IO error (1871)
4311 - class_export_put() on freed export after completion AST error (1896)
4312 - Fix revalidate looping in VFS (1322)
4313 - Don't access a freed export during MDS_REINT timeout (1521)
4314 - Add open-unlink recovery support on the MDS (1673,1764)
4315 - Return an error if no MDS data was read from last_rcvd (1946)
4316 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4317 - Don't setup a disk device that doesn't match exported UUID (317)
4318 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4319 - avoid committing NULL handle in force close
4320 - local.sh is now a one-stripe LOV configuration
4321 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4322 - fix ext3 htree duplicate directory entry corruption (1516)
4323 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4324 - update to new LOV EA format (2097)
4325 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4327 2003-06-15 Phil Schwan <phil@clusterfs.com>
4330 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4331 - per-import/export recovery handling (958, 931, 959)
4332 - multiple last-rcvd slots, for serving multiple FSes (949)
4333 - connections are again shared between multiple imp/exports (963, 964)
4334 - "umount -f" would hang if any requests needed to be sent (393, 978)
4335 - avoid pinning large req buffer by copying for queued messages (989)
4336 - add "uuid" to "lctl device" command to help upcalls (991)
4337 - "open" RPCs with transnos would confuse recovery counters (1037)
4338 - do proper endian conversion of all wire messages (288, 340, 891)
4339 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4340 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4341 - fix ptlrpc_connection leak in target_handle_connect (1174)
4342 - fix import refcounting bug in OST and MDS cleanup (1134)
4343 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4344 - fix very unlikely obd_types race condition (501)
4345 - remove osc_open hack for echo_client (1187)
4346 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4347 - a failure in read_inode2 leads to deadlock (1139)
4348 - cancel ack-locks as soon as transaction is committed (1072)
4349 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4350 - make sure to commitrw after any preprw to avoid deadlock (1162)
4351 - failing to execute a file in a lustre FS would lock inode (1203)
4352 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4353 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4354 - obd_uuid2tgt would walk off the end of the list (1255)
4355 - on IA64 the osc would give portals incorrect bulk size (1258)
4356 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4357 - fix lock inversion caused by new llite matching code (1282)
4358 - limit the number of dirty pages on a client to 10MB (1286)
4359 - timed out locks were not being corrected cancelled (1289)
4360 - fix O_DIRECT above 4GB on IA-32 (1292)
4361 * major user-visible changes
4362 - fail out/fail over policy now controlled by the upcall (993)
4364 - add OBD_PING to check server availability and failure (954)
4365 - lustre messages are now sent in sending host order (288, 340, 891)
4366 - add eadatalen to MDS getattr reply (340)
4367 - OST read replies may contain second buffer, with per-page status (593)
4369 2003-03-11 Phil Schwan <phil@clusterfs.com>
4372 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4373 - failed open()s could cause deadlock; fixed (867, 869)
4374 - stop cancelling OST locks when files are closed (481)
4375 - overlapping XID spaces caused network corruption (851, 853)
4376 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4377 - setattr_raw added, to do single-RPC, server-side setattrs
4378 - lmc/lconf syntax change for OST UUIDs
4379 - fix crashy race condition between ptlrpc_free_req and osc_close
4380 - don't use request in mdc_enqueue if we hit a timeout (889)
4381 - don't set the inode i_size for regular files from the MDS (896)
4382 - handle out of order completion AST (842)
4383 - don't LBUG if a lock request times out after receiving AST (913)
4384 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4385 - if a bad lock AST arrives, send an error instead of dropping entirely
4386 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4387 - fix leak in bulk IO when only partially completed (899, 900, 926)
4388 - fix O_DIRECT for ia64 (55)
4389 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4390 - C-z after timeout could hang a process forever; fixed (977)
4392 - client-side I/O cache (678, 924, 929, 941, 970)
4394 - READPAGE and SETATTRs which don't take server-side locks get
4397 2003-02-11 Phil Schwan <phil@clusterfs.com>
4400 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4401 - Fix fsfilt_extN_readpage() to read a full page of directory
4402 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4403 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4404 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4405 - Most intent operations were removed, in favour of a new RPC mode
4406 that does a single RPC to the server and bypasses most of the VFS
4407 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4408 - Aggressively cancel local locks on DLM servers
4409 - mds_reint_unlink sends EA to the client if it's the last nlink.
4410 client uses that EA to unlink OST objects.
4411 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4412 - recursive symlinks were fixed (439)
4413 - fixed NULL deref in DEBUG_REQ
4414 - filter_update_lastobjid no longer calls sync, which annoyed extN
4415 - fixed multi-client small-writes to a single file problem (445)
4416 - fixed mtime updates during file writes (607)
4417 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4418 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4419 - fixed Linux OST size reporting problem (444, 656)
4420 - OST now updates object mtime with writes or setattr (607, 619)
4421 - client verifies file size before zeroing page past EOF (445)
4422 - OST now writes last allocated objid to disk with allocation (108)
4423 - LOV on echo now works (409)
4425 - mds_reint_unlink sends a new buffer, with the EA included. this
4426 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4427 set if a regular file was being unlinked, and it was the last link
4428 - use PtlGet from the target for bulk writes (315)
4429 - OST now updates object mtime with writes or setattr (607, 619)
4430 - LDLM now has a grant-time callback to revalidate locked items, if
4432 - Many MDS operations were reorganized to combat race conditions
4434 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4435 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4437 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4440 - Fully reactivate OST imports after reconnection (512, others)
4441 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4442 - More graceful error handling for truncating on dead OST (515)
4443 - Don't error out unless we're actually accessing dead stripes (474)
4444 - Fix garbage sizes when stripes are missing (410)
4445 - LRU counters were broken, causing constant lock purge (433, 432)
4446 - garbage on read from stripes with failed OSTs (441)
4447 - mark OSCs as active before reconnecting during recovery (438)
4448 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4449 - lfind did not preserve OST order in output (443)
4450 - symlinks cause hung clients, incorrect data (439)
4451 - stop dereferencing request after dropping refcount (457)
4452 - don't LASSERT(spin_is_locked) on non-SMP (455)
4453 - fixes for many rename() bugs
4454 - fstat didn't correctly synchronize attributes (399)
4455 - server must handle lock cancellation during blocking AST prep (487)
4456 - bulk descriptors were free()d too soon (511)
4457 - fix paths in lconf, which would load incorrect modules (451, 507)
4458 - fix confusing lconf 'host not found' error message (386)
4459 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4460 - fix race condition in mdc_blocking_ast() for inode access (526)
4461 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4462 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4463 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4464 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4465 - fix lstripe in several ways (kernel side) (527)
4466 - fix request leak in ldlm_cli_enqueue (262)
4467 - incorrect OSC was marked inactive after OST failure
4468 - call mds_fs_cleanup before unmounting filesystem (524)
4469 - fix races between taking ns_lock and ldlm_lock_change_resource
4470 - fix races updating LOV export open file list
4471 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4472 - fix recovery NULL deref in ldlm_cli_cancel_unused
4473 - fix some DLM races by using new hash table for lock handles (419)
4474 - permit the client to specify desired inodes, at replay
4475 - duplicate requests when we queue them for replay reintegration
4476 - fix last_rcvd offset calculation
4477 - sync after each recovered transaction, so we always make progress
4478 - never, not always, ERESTART requests without transnos
4479 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4481 - skip replay if the MDS says that the client is already connected
4482 - don't check for a recovery-enabled export to match lctl's UUID
4483 - don't INC_USE_COUNT for phantom exports
4484 - don't crash when cleaning up phantom exports (567)
4485 - don't double-finish or set replay data for errored mdc_open requests
4486 - abort requests when they time out, so we don't get old replies
4487 - send/receive replies for AST messages again
4488 - if the client says that it doesn't have the lock, cancel it on the
4490 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4491 mark it as destroyed, it will all work out when decref is called
4492 - fix module use counts (22, 581)
4494 - ASTs now expect a reply (server cancels lock on error reply)
4496 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4499 - fix many simultaneous client startup (392)
4500 - fix dentry->d_it clobbering
4501 - credentials weren't being shipped for readdir/getattr operations
4502 - remove invalid assertions triggered during some concurrent MD
4504 - proper Lustre versions added (336, 389)
4505 - fix memory leak for create error case (398)
4506 - fix LOV locking bug that would get cli/srv out of sync
4507 - fix echo client over LOV (409)
4508 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4509 - fix double-O_EXCL intent crash (424)
4510 - avoid sending multiple lock CANCELs (352)
4512 - MDS can do multi-client recovery (modulo bugs in new code)
4514 - many updates, edits, cleanups
4516 2002-11-18 Phil Schwan <phil@clusterfs.com>
4519 - fix null d_it dereference (346)
4520 - fix full OST/dbench hang (333)
4521 - fix permission problem with file removal (286)
4522 - fix removal of OSCs from LOV when they fail
4523 - fix NULL deref during bulk timeout (214)
4524 - fix problems related to multiple filesystems on one MDS (241)
4525 - fixed serious subtle metadata locking bugs
4526 - free locks on clients when inodes are removed due to memory
4528 - fix inode pointer in lock data (285)
4529 - partial support for multiple MDS on a single host (241)
4530 - data locks weren't cancelled at clear_inode time (290, 311)
4531 - intent locks could lead to unbounded lock growth (205)
4532 - added a maximum lock count, an LRU list, and a flusher
4533 - fix multiple rename (365)
4534 - properly abstracted the echo client
4535 - OSC locked 1 byte too many; fixed
4536 - rewrote brw callback code:
4537 - fixed recovery bugs related to LOVs (306)
4538 - fixed too-many-pages-in-one-write crash (191)
4539 - fixed (again) crash in sync_io_timeout (214)
4540 - probably fixed callback-related race (385)
4542 - Add capability to MDS protocol
4543 - LDLM cancellations and callbacks on different portals
4545 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4548 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4549 - defer open object destruction to close time (601981, 138)
4550 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4551 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4552 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4553 - don't call obd_setattr in ll_file_release for destroyed objects
4554 * protocol change to lustre_msg: move |version| and add |flags|
4555 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4556 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4557 * added replay of create, unlink, link and rename operations during
4558 MDS failover; recovery should be much more robust now
4559 * remove failed OSCs from LOVs (only lov_create uses this so far)
4560 * the lustre-HOWTO was brought (more) up to date (582544)
4562 2002-10-23 Phil Schwan <phil@clusterfs.com>
4565 - in-use dentries weren't being reused properly (617851)
4566 - prevent multiple LDLM setup (599178)
4567 - fix LOV size calculations for truncate (617853)
4568 - fix client handling of MDS intent errors (POSIX)
4569 - fix permission bug in lovstripe.c test (624321)
4570 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4571 - truncate past end of file could corrupt data
4572 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4573 - a race in recovery could return ETIMEDOUT to apps (623947)
4574 - building outside the source directory was fixed
4575 * the lustre-HOWTO was brought (more) up to date (582544)
4576 * major progress was made on recovery functionality
4578 2002-10-10 Phil Schwan <phil@clusterfs.com>
4581 - recovery deadlock fix
4582 - rm -rf causes LBUG fix (617817)
4583 - file open by multiple tasks fix (618962)
4584 - directory permissions bugs (602707 and 620007)
4585 - journal_stop fixed with locking (611313)
4586 - O_APPEND failures resolved (618273, perhaps 614459)
4587 - lconf PATH fix (619770)
4588 - IA64 build fix (621450)
4589 - RPC buffer sizes scale with amount of memory
4591 2002-10-01 Phil Schwan <phil@clusterfs.com>
4594 - locks would be cancelled without throwing away data pages,
4595 resulting in inconsistent data (605627)
4596 - inode attributes were not always being refreshed (605627, 612449)
4597 - lconf now continues to cleanup after lctl reports an error
4598 - MDS now enforces user permissions (602707)
4599 - lprocfs cleanup fixed, but not yet enabled (614157)
4600 - fixed infinite server hang, should a client not respond to an AST
4601 - avoid going into recovery if user calls readlink() with a buffer
4602 that's too small (613941)
4603 - AST RPCs no longer require replies (614867) -- this may be changed
4604 - don't crash server if client sends an IOV that's too big (611336)
4605 - fixed lock conversion deadlock (611892)
4606 - fixed the following of symlinks (614622)
4607 * recovery: the server can remove locks from a client that dies, other
4608 clients can make progress
4609 * more extN patch fixes
4610 * compile-time configurable ptlrpc buffer allocations
4612 - collaborative read cache document
4613 - Lustre Lite Performance CDR document-in-progress
4615 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4618 - fix typo in patch-2.4.18
4620 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4623 - clear ptlrpc request each time in handle_incoming_request()
4624 - unlink of files now destroys the object on the OST
4626 2002-09-19 Peter Braam <braam@clusterfs.com>
4628 * add hard link support
4629 * change obdfile creation method
4630 * kernel patch changed
4632 2002-09-19 Peter Braam <braam@clusterfs.com>
4635 - stack overflow bug in extN fixed
4637 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4639 * documentation updates
4640 - add man pages for config tools
4641 - update tests/README to describe testing with new config tools
4642 - finish metadata API descriptions
4643 * bug fixes and cleanups
4644 - statfs workaround for 16TB limit
4645 - LOV stripe allocation improved, can stripe on subset of OSTs
4646 - LOV file size/IO offset was wrong for files > 4GB in size
4647 - object EA data was being dropped, caused files to be unreadable
4648 - memory overflow with non-LOV OST caused memory corruption
4649 - fixed regression tests to work with new config tools, obdfilter
4650 - fixed bug when directory size became larger than 1 block
4651 - fixed bug (for single client case) when PWD was deleted
4652 - invalidate local directory pages when doing intent-based ops
4653 - avoid LDLM oops when lock callback contained bad data
4655 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4657 * documentation updates
4658 * bug fixes and cleanups
4659 - configuration tools
4662 - 64-bit compile warnings
4663 - 64-bit internal statfs data
4665 * test_brw on persistent OST devices
4667 * lprocfs (disabled)
4669 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4671 * documentation updates
4672 * bug fixes and cleanups
4673 * configuration tools
4675 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4678 * many small fixes to 0_5_4
4679 * io/network handling
4680 * thinkos in MDS operations
4682 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4685 * crucial basic fixes to 0.5.3
4686 * IOR, Iozone work over Elan
4689 2002-08-07 Phil Schwan <phil@clusterfs.com>
4690 * version 0_5_3, our first alpha
4691 * we use the new Portals iovs
4692 * documentation updates
4693 * bug fixes and cleanups
4694 * small changes in the DLM wire protocol
4696 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4697 * version 0_5_1 with some initial stability,
4698 * locking on MD and file I/O.
4699 * documentation updates
4700 * several bug fixes since 0.5.0
4701 * small changes in wire protocol
4703 2002-07-18 Phil Schwan <phil@clusterfs.com>
4705 * delivered as Lustre Light Alpha
4706 * fixed a crash after handling invalid MDS requests
4707 * fixed directory pages for architectures with non-4k pages sizes
4709 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4710 * release version v0_4_4
4711 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4713 * Fixes a number of interruption problems with OST operations.
4714 * Update documentation for portals header changes
4715 * Move all wire protocol structs/defines to lustre_idl.h
4716 * Fixes symlink length bug.
4717 * Add tcpdump to repository.
4719 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4720 * release version v0_4_3
4721 * Fixes statfs for inodes on extN.
4722 * Fixes bug in runtests which would delete /etc/hosts.
4723 * Use 64-bit object IDs wherever possible (not into VFS though)
4724 Remove ost_get_info, which is unused by lustre, and out of date.
4726 2002-07-03 Peter Braam <braam@clusterfs.com>
4727 * release version v0_4_2 Fixes a lookup error (type not passed)
4728 * move forward to head of Portals
4729 * move forward to latest Lustre kernel
4731 2002-06-25 Peter Braam <braam@clusterfs.com>
4732 * release version v0_4_1. Hopefully stable on single node use.