1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.19.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-sun3
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).
15 Severity : enhancement
17 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
21 Description: enable MGS and MDT services start separately
22 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
23 a co-located MGS without starting the MGS, which is a complement
24 to 'nosvc' mount option.
27 Frequency : always with o2ib 1.3 and sles10
29 Description: fix build with SLES10 and o2ib v3.
30 Details : sles10 uses diffrent name for Module.symver file but configure
31 assume this file has same name on RHEL/SLES/vanila kernels.
35 Frequency : after recovery
36 Description: precreate to many object's after del orphan.
37 Details : del orphan st in oscc last_id == next_id and this triger growing
38 count of precreated objects. Set flag LOW to skip increase count
39 of precreated objects.
43 Frequency : rare, on clear nid stats
44 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
45 Details : when clean nid stats sometimes try destroy live entry,
46 and this produce panic in free.
50 Description: Stack overflow during MDS log replay
51 ease stack pressure by using a thread dealing llog_process.
55 Description: wait until IO finished before start new when do lock cancel.
56 Details : VM protocol want old IO finished before start new, in this case
57 need wait until PG_writeback is cleared until check dirty flag and
58 call writepages in lock cancel callback.
60 Severity : enhancement
62 Description: using special macro for print time and cleanup in includes.
66 Description: mds_mfd_close() ASSERTION(rc == 0)
67 Details : In mds_mfd_close(), we need protect inode's writecount change
68 within its orphan write semaphore to prevent possible races.
72 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
77 Description: don't hit live lock with umount ost.
78 Details : shrink_dcache_parent can be in long loop with destroy dentries,
79 use shrink_dcache_sb instead.
83 Description: don't panic with use echo client
84 Details : echo client pass NULL as client nid pointer and this produce null
89 Description: fix build on ppc32
90 Details : compile code with -m64 flag produce wrong object file for ppc32.
94 Description: add message levels for liblustreapi
98 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
99 Details : Move the 'good_osts' check before the 'total_bavail' check. This
100 will result in an -EAGAIN and in the exit call path we call
101 alloc_rr() which will with increasing aggressiveness attempt to
102 aquire precreated objects on the minimum number of required OSCs.
106 Description: Use old size assignment to avoid deadlock
107 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
108 the scheduling while holding a spinlock. We do not need locking
109 for size in ll_update_inode() because size is only updated from
110 the MDS for directories or files without objects, so there is no
111 other place to do the update, and concurrent access to such inodes
112 are protected by the inode lock.
116 Description: resolve "_IOWR redefined" build error on SLES10
120 Description: dump the memory debugging after all modules are unloaded to
121 suppress false negative in conf_sanity test 39
123 Severity : enhancement
125 Description: build kernel-ib packages for OFED 1.3 in our release cycle
130 Description: fix SLES kernel versioning
131 Details : the kernel version for our SLES 10 kernel did not include a "-"
132 before the "smp" at the end. while this was not a problem in
133 general, it did mean that software trying to use the kernel
134 version to try to detect a vendor specific kernel would fail.
135 this was most evident by the OFED build scripts.
139 Description: Don't update lov_desc members until making sure they are valid
140 Details : When updating lov_desc members via proc fs, need fix their
141 validities before doing the real update.
145 Description: don't put request into delay list while invalidate in flight.
146 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
147 import in flight. this produce timeout for invalidate and sometimes
148 can cause stale data.
152 Frequency : on ppc only
153 Description: not convert ost objects for directory because it's not exist.
154 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
158 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
160 Severity : enhancement
162 Description: Update to support 2.6.22.14 vanilla kernel.
166 Frequency : rare, on recovery
167 Description: read procfs can produce deadlock in some situation
168 Details : Holding lprocfs lock which send rpc can produce block for destroy
169 obd objects and this also block reconnect with -EALREADY. This isn't
170 fix all lprocfs bugs - but make it rare.
172 Severity : enhancement
174 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
177 Frequency : frequent on X2 node
179 Description: mdc_set_open_replay_data LBUG
180 Details : Set replay data for requests that are eligible for replay.
184 Description: lustre_mgs: operation 101 on unconnected MGS
185 Details : When MGC is disconnected from MGS long enough, MGS will evict the
186 MGC, and late on MGC cannot successfully connect to MGS and a lot
187 of the error messages complaining that MGS is not connected.
191 Frequency : on network error
192 Description: panic with double free request if network error
193 Details : mdc_finish_enqueue is finish request if any network error ocuring,
194 but it's true only for synchronus enqueue, for async enqueue
195 (via ptlrpcd) this incorrect and ptlrpcd want finish request
198 Severity : enhancement
200 Description: client-side metadata stat-ahead during readdir(directory readahead)
201 Details : perform client-side metadata stat-ahead when the client detects
202 readdir and sequential stat of dir entries therein
205 Frequency : on start mds
207 Description: Implement get_info(last_id) in obdfilter.
210 Frequency : occasional
212 Description: Correctly check stale fid, not start epoch if ost not support SOM
213 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
214 overwrited with new generated fid, but mds can anwer with one of these
215 two fids and both is not stale. setattr incorectly start epoch and
216 assume will be called done_writeting, but without SOM done_writing
220 Frequency : rare, depends on device drivers and load
222 Description: MDS or OSS nodes crash due to stack overflow
223 Details : Code changes in 1.8.0 increased the stack usage of some functions.
224 In some cases, in conjunction with device drivers that use a lot
225 of stack the MDS (or possibly OSS) service threads could overflow
226 the stack. One change which was identified to consume additional
227 stack has been reworked to avoid the extra stack usage.
230 Frequency : occasional
232 Description: Do not fail import if osc_interpret_create gets -EAGAIN
233 Details : If osc_interpret_create got -EAGAIN it immediately exits and
234 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
235 oscc_has_objects and see OSC has no objests and call
236 oscc_internal_create to resend create request.
238 Severity : enhancement
240 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
242 Severity : enhancement
244 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
247 Frequency : very rare
249 Description: Fix replay if there is an un-replied request and open
250 Details : In some cases, older replay request will revert the
251 mcd->mcd_last_xid on MDS which is used to record the client's
254 Severity : enhancement
256 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
258 Severity : enhancement
260 Description: Add rhel5 support to HEAD.
262 Serverity : enhancement
264 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
269 Description: Don't allow skipping OSTs if index has been specified.
270 Details : Don't allow skipping OSTs if index has been specified, make locking
271 in internal create lots better.
275 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
276 Details : access to bitfield in structure is always rounded to long
277 and this produce problem with not atomic change any bit.
281 Description: Lustre make rpms failed.
282 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
283 builds Lustre rpms from tarball.
286 Frequency : rare on shutdown ost
288 Description: If llog cancel was not send before clean_exports phase, this can
289 produce deadlock in llog code.
290 Details : If llog thread has last reference to obd and call class_import_put
291 this produce deadlock because llog_cleanup_commit_master wait when
292 last llog_commit_thread exited, but this never success because was
293 called from llog_commit_thread.
297 Description: allow userland application know is lost one of stripes.
298 Details : fill lvb_blocks with error code on ost and return it to
299 application if error flag found.
303 Description: NULL lov_tgts causing MDS oops
304 Details : more safe checks for NULL lov_tgts for avoid oops.
306 Severity : enhancement
308 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
312 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
313 Details : If an OST is down the MDS will hang indefinitely in
314 obd_statfs_rqset() waiting for the statfs data. While for
315 MDS QOS usage of statfs, it should not stuck in waiting.
317 Severity : enhancement
319 Description: remote_acl support
320 Details : Support ACL-based permission check for remote user.
321 Support setfacl/getfacl for remote user with the utils
322 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
323 the system "{s,g}etfacl" utils.
325 Severity : enhancement
327 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
329 Severity : enhancement
331 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
335 Description: make mgs_setparam() handle fsname containing dash
336 Details : fsname containing a dash does not work with lctl conf_param
338 Severity : enhancement
340 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
342 Severity : enhancement
344 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
348 Frequency : rare, at shutdown
349 Description: access already free / zero obd_namespace.
350 Details : if client_disconnect_export was called without force flag set,
351 and exist connect request in flight, this can produce access to
352 NULL pointer (or already free pointer) when connect_interpret
353 store ocd flags in obd_namespace.
357 Frequency : only at startup
358 Description: not alloc memory with spinlock held.
359 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
360 if any spinlock held.
362 Severity : enhancement
364 Description: make lustre randomly fail allocating memory
365 Details : Make lustre randomly failed allocating memory for testing purpose.
367 Severity : enhancement
369 Description: lost problems with lov objid file
370 Details : Fixes some scability and access to not inited memory problems
371 in work with lov objdid file.
376 Description: lfs find does not continue on file error
377 Details : Continue other files processing when a file/dir is absent.
381 Description: Inconsistent usage of lustre_pack_reply()
382 Details : Standardize the usage of lustre_pack_reply() such that it
383 always generate a CERROR on failure.
386 Frequency : occasional
388 Description: MDS hang or stay in waiting lock
389 Details : If client receive lock with CBPENDING flag ldlm need send lock
390 cancel as separate rpc, to avoid situation when cancel request
391 can't processed due all i/o threads stay in wait lock.
395 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
399 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
403 Description: Update to SLES9 kernel-2.6.5-7.287.3.
407 Description: Update to RHEL4 latest kernel.
409 Severity : enhancement
411 Description: Build SLES10 patchless client fails
412 Details : The configure was broken by run ./configure with
413 --with-linux-obj=.... argument for patchless client. When the
414 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
415 can't search header adequately. Use absolute path such as
416 -I($LINUX)/include instead.
420 Description: interrupt oig_wait produce painc on resend.
421 Details : brw_redo_request can be used for resend requests from ptlrpcd and
422 private set, and this produce situation when rq_ptlrpcd_data not
423 copyed to new allocated request and triggered LBUG on assert
424 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
425 ptlrpcd set if request is changed and can be safety changed to use
430 Description: LASSERT_{REQ,REP}SWAB macros are buggy
431 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
432 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
433 these macros and replace them with their difinitions instead.
437 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
438 Details : Update which_patch & target file for SLES10 latest kernel.
441 Description: Remove client patches from SLES 10 kernel.
442 Details : This causes SLES 10 clients to behave as patchless clients
443 even on a Lustre-patched (server) kernel.
445 Severity : enhancement
447 Description: self-adjustable client's lru lists
448 Details : use adaptive algorithm for managing client cached locks lru
449 lists according to current server load, other client's work
450 pattern, memory activities, etc. Both, server and client
451 side namespaces provide number of proc tunables for controlling
454 Severity : enhancement
456 Description: light-weight GSS support
457 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
458 and reduce performance overhead.
460 Severity : enhancement
462 Description: Linux keyring support
463 Details : Support using service of Linux keyring for Lustre GSS internal
464 context refresh/cache mechanism.
468 Description: Fix errors in lfs documentation
469 Details : Fixes man pages
473 Description: don't use GFP_* in generic Lustre code.
474 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
478 Description: obdclass is limited by single OBD_ALLOC(idarray)
479 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
483 Description: warnings with build patchless client with vanila 2.6.19 and up
484 Details : change old ctl_table style and replace ctl_table/ctl_table_header
485 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
489 Frequency : Only for SLES
490 Description: sanity_quota fail test_1
491 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
492 file, which makes TSTID[2] inunique.
496 Frequency : Only for SLES9
497 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
498 Details : Update target/ChangeLog/which_patch .
502 Frequency : Only for test_10 in sanity-quota.sh
503 Description: a bug in quota test code
504 Details : checking a test flag in a wrong place
508 Frequency : Only for test_13 in sanity-quota.sh
509 Description: a bug in quota test script
510 Details : shouldn't check overfull allocated quota
514 Description: jbd statistics
515 Details : Port older jbd statistics patch for sles10
517 Severity : enhancement
519 Description: Add printing inode info into message about error in writepage.
524 Description: reply_lock_interpret crash due to race with it and lock cancel.
525 Details : Do not replay locks that are being cancelled. Do not reference
526 locks by their address during replay, just by their handle.
530 Frequency : When flocks are used.
531 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
532 Details : Flock locks might destroy just granted lock if it could be merged
533 with another existing flock, this is done in completion handler,
534 so teach ldlm_cli_enquque_fini that this is a valid case for
540 Description: Oops in read and write path when failing to allocate lock.
541 Details : Check if lock allocation failed and return error back.
545 Description: lstripe command fails for valid OST index
546 Details : The stripe offset is compared to lov->desc.ld_tgt_count
547 instead of lov->desc.ld_active_tgt_count.
551 Description: sanity.sh failed test 103
552 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
557 Description: df doesn't work properly if diskfs blocksize != 4K
558 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
562 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
563 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
567 Description: obd_health_check_timeout too short
568 Details : set obd_health_check_timeout as 1.5x of obd_timeout
572 Description: llapi_file_create() does not allow some changes
573 Details : add llapi_file_open() that allows specifying the mode and
574 open flags, and also returns an open file handle.
578 Description: merge and cleanup kernel patches.
579 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
583 Description: Add journal checksum support.(Kernel part)
584 Details : The journal checksum feature adds two new flags i.e
585 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
586 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
587 indicates that the commit block contains the checksum for
588 the blocks described by the descriptor blocks. Now commit
589 record can be sent to disk without waiting for descriptor
590 blocks to be written to disk. This behavior is controlled
591 using JBD2_FEATURE_ASYNC_COMMIT flag.
595 Description: OSS needs mutliple precreate threads
596 Details : Add ability to start more than one create thread per OSS.
600 Description: Sanity.sh test_65k failed: setstripe should have succeeded
601 Details : execute lfs setstripe on client
605 Description: mds_obd_create error creating tmp object
606 Details : When the user sets quota on root, llog will be affected and can't
607 create files and write files.
610 Frequency : Always on ia64 patchless client, and possibly others.
612 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
613 Details : This allows the patchless client to be loaded on architectures
618 Description: RedHat Update kernel for RHEL5
619 Details : Add the kernel config file for RHEL5.
623 Description: RedHat Update kernel for RHEL5
624 Details : Modify the kernel config file more closer RHEL5.
628 Description: Build failure against Centos5 (RHEL5)
629 Details : Define PAGE_SIZE when it isn't present.
633 Description: test_7 on sanity-quota.sh can't be used on separate machines
634 Details : add facet to handle it
638 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
639 Details : with NFS, the anon dentry's parent was set to itself in
640 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
641 corresponding dentry other than use rec->ur_name.
643 Severity : enhancement
645 Description: lfs setstripe enhancement
646 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
650 Description: enable data checksumming by default
651 Details : enable checksum by default, allow --disable-checksum
652 configure option and "-o nochecksum" mount option. Checksums
653 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
657 Description: lustre support for RHEL5
658 Details : Add support for RHEL5.
662 Description: Client eviction due to failover config
663 Details : after a connection loss, the lustre client should attempt to
664 reconnect to the last active server first before trying the
665 other potential connections.
669 Description: when mds and osts use different quota unit(32bit and 64bit),
670 quota will be released repeatly.
671 Details : void sending multiple quota reqs to mds, which will keep the
672 status between the reqs.
676 Description: osts not allocated evenly to files
677 Details : change the condition to increase offset_idx
680 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
682 Description: Data corruption for OSTs that are formatted larger than 2TB
684 Details : When generating the bio request for lustre file writes the
685 sector number would overflow a temporary variable before being
686 used for the IO. The data reads correctly from Lustre (which
687 will overflow in a similar manner) but other file data or
688 filesystem metadata may be corrupted in some cases.
692 Description: Tune the kernel for good SCSI performance.
693 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
694 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
699 Description: rewrite ext2-derived code in obdclass/uuid.c
700 Details : rewrite inherited code (uuid parsing code from ext2 utils)
701 from scratch preserving functionality.
705 Description: "lfs find -obd UUID" prints directories
706 Details : "lfs find -obd UUID" will return all directory names instead
707 of just file names. It is incorrect because the directories
708 do not reside on the OSTs.
712 Description: use i_size_read and i_size_write in 2.6 port
713 Details : replace inode->i_size access with i_size_read/write()
716 Frequency : when removing large files
718 Description: scheduling issue during removal of large Lustre files
719 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
720 It causes scheduling issues when removing large files (17TB in the
724 Frequency : only with liblustre clients on XT3
726 Description: evictions taking too long
727 Details : allow llrd to evict clients directly on OSTs
730 Frequency : only on ppc
732 Description: /proc/fs/lustre/devices broken on ppc
733 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
734 We should make sure the type of 'index' is loff_t and then cast
735 explicitly as needed below. Do not assign an explicitly cast
739 Frequency : only for rhel5
741 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
742 Details : Modify the target file & which_kernel.
744 Severity : enhancement
746 Description: omit set fsid for export NFS
747 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
748 as FS_REQUIRES_DEV for avoid problems with generate fsid.
752 Frequency : Always, for kernels after 2.6.16
753 Description: Fix warning idr_remove called for id=.. which is not allocated.
754 Details : Last kernels save old s_dev before kill super and not allow
755 to restore from callback - restore it before call kill_anon_super.
759 Description: buffer overruns could theoretically occur
760 Details : llapi_semantic_traverse() modifies the "path" argument by
761 appending values to the end of the origin string, and a
762 overrun may occur. Adding buffer overrun check in liblustreapi.
766 Description: Fix error on 'ls .' at the top of the Lustre mount.
767 Details : Don't revalidate dentry if it is a root dentry.
771 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
772 Details : Modify vm-tunables-rhel4.patch.
776 Description: Kernel config for 2.6.18-vanilla.
777 Details : Modify targets/2.6-vanilla.target.in.
778 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
779 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
780 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
781 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
785 Description: improve handling recoverable errors
786 Details : if request processig with error which can be recoverable on server
787 request should be resend, otherwise page released from cache and
792 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
793 Details : Modify target file & which_patch.
794 A flaw was found in the IA32 system call emulation provided
795 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
796 value could be stored in the %RAX register, which could trigger an
797 out-of-bounds system call table access. An untrusted local user
798 could exploit this flaw to run code in the kernel
799 (ie a root privilege escalation). (CVE-2007-4573).
803 Description: O_DIRECT bypasses client statistics.
804 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
805 read_ahead_stats were updated. copied the stats section from
806 osc_send_oap_rpc() into async_internal().
810 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
814 Description: change order of libsysio includes
815 Details : '#include sysio.h' should always come before '#include xtio.h'
819 Description: handle "serious error: objid * already exists" more gracefully
820 Details : If LAST_ID value on disk is smaller than the objects existing in
821 the O/0/d* directories, it indicates disk corruption and causes an
822 LBUG(). If the object is 0-length, then we should use the existing
823 object. This will help to avoid a full fsck in most cases.
825 Severity : enhancement
827 Description: adapt the lustre_config script to support the upgrade case
828 Details : Add "-u" option for lustre_config script to support upgrading 1.4
829 server targets to 1.6 in parallel.
833 Description: To avoid grant space > avaible space when the disk is almost
834 full. Without this patch you might see the error "grant XXXX >
835 available" or some LBUG about grant, when the disk is almost
837 Details : In filter_check_grant, for non_grant cache write, we should
838 check the left space by if (*left > ungranted + bytes), instead
839 of (*left > ungranted), because only we are sure the left space
840 is enough for another "bytes", then the ungrant space should be
841 increase. In client, we should update cl_avail_grant only there
842 is OBD_MD_FLGRANT in the reply.
846 Description: Update RHEL 4 kernel to fix local root privilege escalation.
847 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
848 described in CVE-2007-4573. This problem could allow untrusted
849 local users to gain root access.
852 Frequency : when using O_DIRECT and quotas
854 Description: Incorrect file ownership on O_DIRECT output files
855 Details : block usage reported by 'lfs quota' does not take into account
856 files that have been written with O_DIRECT.
861 Description: touch file failed when fs is not full
862 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
863 otherwise we can get ENOSP while fs is not full.
867 Description: parallel lock callbacks
868 Details : Instead of sending blocking and completion callbacks as separated
869 requests, adding them to a set and sending in parallel.
872 Frequency : only for Cray XT3
873 Bugzilla : 12829/13455
874 Description: Changing primary group doesn't change the group lustre assigns to
876 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
877 overridden with the primary group provided by the group upcall,
878 whereas the supplied fsgid can be trusted if it is in the list of
879 supplementary groups returned by the group upcall.
881 Severity : enhancement
883 Description: Allow masking D_WARNING, D_ERROR messages from console
884 Details : Console messages can now be disabled via lnet.printk.
888 Description: User code with malformed file open parameter crashes client node
889 Details : Before packing join_file req, all the related reference should be
890 checked carefully in case some malformed flags cause fake
891 join_file req on client.
895 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
896 Details : ldlm_completion_ast() assumes that a lock is granted when the req
897 mode is equal to the granted mode. However, it should also check
898 that LDLM_FL_CP_REQD is not set.
902 Description: Heavy nfs access might result in deadlocks
903 Details : After ELC code landed, it is now improper to enqueue any mds
904 locks under och_sem, because enqueue might want to decide to
905 cancel open locks for same inode we are holding och_sem for.
909 Description: Client eviction while running blogbench
910 Details : A lot of unlink operations with concurrent I/O can lead to a
911 deadlock causing evictions. To address the problem, the number of
912 oustanding OST_DESTROY requests is now throttled to
913 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
914 callbacks are processed in priority.
918 Description: enable ACLs on MDS by default
919 Details : ACLs must be enabled on MDS by default.
922 Frequency : PPC/PPC64 only
924 Description: conflicts between asm-ppc64/types.h and lustre_types.h
925 Details : fix duplicated definitions between asm-ppc64/types.h and
926 lustre_types.h on PPC.
929 Frequency : PPC/PPC64 only
931 Description: asm-ppc/segment.h does not exist
932 Details : fix compile issue on PPC.
936 Description: better handle error messages in extents code
939 Frequency : RHEL4 only
941 Description: mkfs is very slow on IA64/RHEL4
942 Details : A performance regression has been discovered in the MPT Fusion
943 driver between versions 3.02.73rh and 3.02.99.00rh. As a
944 consequence, we have downgraded the MPT Fusion driver in the RHEL4
945 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
947 Severity : enhancement
949 Description: SNMP support enhancement
950 Details : Adding total number of sampled request for an MDS node in snmp
953 Severity : enhancement
955 Description: Optimize ldlm waiting list processing for PR extent locks
956 Details : When processing waiting list for read extent lock and meeting read
957 lock that is same or wider to it that is not contended, skip
958 processing rest of the list and immediatelly return current
959 status of conflictness, since we are guaranteed there are no
960 conflicting locks in the rest of the list.
964 Description: Time out and refuse to reconnect
965 Details : When the failover node is the primary node, it is possible
966 to have two identical connections in imp_conn_list. We must
967 compare not conn's pointers but NIDs, otherwise we can defeat
968 connection throttling.
972 Description: port llog fixes from b1_6 into HEAD
973 Details : Port llog reference couting and some llog cleanups from b1_6
974 (bug 10800) into HEAD, for protect from panic and access to already
975 free llog structures.
979 Description: Detect stride IO mode in read-ahead
980 Details : When a client does stride read, read-ahead should detect that and
981 read-ahead pages according to the detected stride pattern.
985 Description: data checksumming impacts single node performance
986 Details : add support for several checksum algorithm. Currently, only CRC32
987 and Adler-32 are supported. The checksum type can be changed on
988 the fly via /proc/fs/lustre/osc/*/checksum_type.
992 Description: use adler32 for page checksums
993 Details : when available, use the Adler-32 algorithm instead of CRC32 for
998 Description: build for x2 fails
999 Details : fix compile issue on Cray systems.
1003 Description: Properly match for duplicate locks
1004 Details : Due to different lock order from skiplists code, we need to
1005 traverse entire list for now
1008 Frequency : only on PPC/SLES10
1010 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1011 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1012 be defined. Add a hack in mkfs_lustre.c to work around this compile
1017 Description: LASSERT on MDS when client holding flock lock dies
1018 Details : ldlm pool logic depends on number of granted locks equal to
1019 number of released locks which is not true for flock locks, so
1020 just exclude such locks from consideration.
1024 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1025 Details : Use fsfilt_sync() for both the whole filesystem sync and
1026 individual file sync to eliminate dangerous inode locking
1027 with I_LOCK that can lead to a deadlock.
1031 Description: performance in 1.6.3
1032 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1033 sure that 1MB requests are not fragmented by the block layer.
1035 Severity : enhancement
1037 Description: organize the server-side client stats on per-nid basis
1038 Details : Change the structure of stats under obdfilter and mds to
1046 The "uuid"s file would list the uuids of _active_ exports.
1047 And the clear entry is to clear all stats and stale nids.
1049 Severity : enhancement
1051 Description: eliminate client locks in face of contention
1052 Details : file contention detection and lockless i/o implementation
1053 for contended files.
1057 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1060 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1062 Description: MDS may lose file striping (and hence file data) in some cases
1063 Details : If there are additional extended attributes stored on the MDS,
1064 in particular ACLs, SELinux, or user attributes (if user_xattr
1065 is specified for the client mount options) then there is a risk
1066 of attribute loss. Additionally, the Lustre file striping
1067 needs to be larger than default (e.g. striped over all OSTs),
1068 and an additional attribute must be stored initially in the
1069 inode and then increase in size enough to be moved to the
1070 external attribute block (e.g. ACL growing in size) for file
1075 Description: skiplist implementation simplification
1076 Details : skiplists are used to group compatible locks on granted list
1077 that was implemented as tracking first and last lock of each lock group
1078 the patch changes that to using doubly linked lists
1082 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1083 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1084 because LASSERT on both the data supplied by a client, and the data
1085 on disk is dangerous and incorrect.
1087 --------------------------------------------------------------------------------
1089 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1091 * Support for kernels:
1092 2.6.5-7.283 (SLES 9),
1093 2.6.9-55.EL (RHEL 4),
1094 2.6.16.46-0.14 (SLES 10),
1095 2.6.18.8 vanilla (kernel.org)
1096 * Client support for unpatched kernels:
1097 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1098 2.6.16 - 2.6.22 vanilla (kernel.org)
1099 * Due to recently discovered recovery problems, we do not recommend
1100 using patchless RHEL 4 clients with this or any earlier release.
1101 * Recommended e2fsprogs version: 1.39.cfs8
1102 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1103 * Starting with this release, the ldiskfs backing filesystem required
1104 by Lustre is now in its own package, lustre-ldiskfs. This package
1105 should be installed. It is versioned separately from Lustre and
1106 may be released separately in future.
1110 Description: block reactivating mgc import until all deactivates complete
1111 Details : Fix race when failing back MDT/MGS to itself (testing)
1114 Frequency : at statup only
1116 Description: mds_lov_synchronize race leads to various problems
1117 Details : simultaneous MDT->OST connections at startup can cause the
1118 sync to abort, leaving the OSC in a bad state.
1120 Severity : enhancement
1122 Description: add optional extra BUILD_VERSION info
1123 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1124 to override the lustre version.
1127 Frequency : 2.6.18 servers only
1129 Description: ll_kern_mount() doesn't release the module reference
1130 Details : The ldiskfs module reference count never drops down to 0
1131 because ll_kern_mount() doesn't release the module reference.
1136 Description: server LBUG when using old ost_num_threads parameter
1137 Details : Accept the old ost_num_threads parameter but warn that it
1138 is deprecated, and fix an off-by-one error that caused an LBUG.
1143 Description: Transient SCSI error results in persistent IO issue
1144 Details : iobuf->dr_error is not reinitialized to 0 between two
1148 Frequency : sometimes when underlying device returns I/O errors
1150 Description: OSTs not going read-only during write failures
1151 Details : OSTs are not remounted read-only when the journal commit threads
1152 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1153 instead of the ext3 wrappers.
1157 Description: poor connect scaling with increasing client count
1158 Details : Don't run filter_grant_sanity_check for more than 100 exports
1159 to improve scaling for large numbers of clients.
1162 Frequency : SLES10 only
1164 Description: sanity-quota.sh quotacheck failed: rc = -22
1165 Details : Quotas cannot be enabled on SLES10.
1168 Frequency : liblustre clients only
1170 Description: getdirentries does not give error when run on compute nodes
1171 Details : getdirentries does not fail when the size specified as an argument
1172 is too small to contain at least one entry
1174 Severity : enhancement
1176 Description: Add LNET router traceability for debug purposes
1177 Details : If a checksum failure occurs with a router as part of the
1178 IO path, the NID of the last router that forwarded the bulk data
1179 is printed so it can be identified.
1184 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1185 Details : Due to a race condition, liblustre clients were occasionally
1186 evicted incorrectly.
1188 Severity : enhancement
1190 Description: lfs setstripe use optional parameters instead of postional
1193 Severity : enhancement
1195 Description: Nanosecond timestamp support for ldiskfs
1196 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1197 resolution timestamps. There is not yet support for this at
1198 the Lustre filesystem level.
1201 Frequency : during server recovery
1203 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1204 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1208 Frequency : nfs export on patchless client
1210 Description: connectathon hang when test nfs export over patchless client
1211 Details : Disconnected dentry cannot be found with lookup, so we do not need
1212 to unhash it or make it invalid
1215 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1216 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1217 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1222 Description: open req refcounting wrong on reconnect
1223 Details : If reconnect happened between getting open reply from server and
1224 call to mdc_set_replay_data in ll_file_open, we will schedule
1225 replay for unreferenced request that we are about to free.
1226 Subsequent close will crash in variety of ways.
1227 Check that request is still eligible for replay in
1228 mdc_set_replay_data().
1233 Description: disable writes to filesystem when reading health_check file
1234 Details : the default for reading the health_check proc file has changed
1235 to NOT do a journal transaction and write to disk, because this
1236 can cause reads of the /proc file to hang and block HA state
1237 checking on a healthy but otherwise heavily loaded system. It
1238 is possible to return to the previous behaviour during configure
1239 with --enable-health-write.
1241 Severity : enhancement
1243 Description: 64-bit inode version
1244 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1245 to the inode. This will be required for version-based recovery.
1250 Description: MDS fails to start if a duplicate client export is detected
1251 Details : in some rare cases it was possible for a client to connect to
1252 an MDS multiple times. Upon recovery the MDS would detect this
1253 and fail during startup. Handle this more gracefully.
1255 Severity : enhancement
1257 Description: Add -o localflock option to simulate old noflock
1259 Details : This will achieve local-only flock/fcntl locks
1265 Description: log_commit_thread vs filter_destroy race leads to crash
1266 Details : Take import reference before releasing llog record semaphore
1271 Description: Wrong request locking in request set processing
1272 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1273 imp_delayed_list, in this place should be used imp_lock.
1276 Frequency : when reconnection
1278 Description: Grant Leak when osc reconnect to OST
1279 Details : When osc reconnect ost, OST(filter) should check whether it
1280 should grant more space to client by comparing fed_grant and
1281 cl_avail_grant, and return the granted space to client instead
1282 of "new granted" space, because client will call osc_init_grant
1283 to update the client grant space info.
1286 Frequency : when client reconnect to OST
1288 Description: Grant Leak when osc do resend and replay bulk write
1289 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1290 bulk write request, because the grant info while be sync between
1291 OSC and OST when reconnect, and we should ignore the grant info
1292 these of resend/replay write req.
1297 Description: Grant space more than avaiable left space sometimes.
1298 Details : When then OST is about to be full, if two bulk writing from
1299 different clients came to OST. Accord the avaliable space of the
1300 OST, the first req should be permitted, and the second one
1301 should be denied by ENOSPC. But if the seconde arrived before
1302 the first one is commited. The OST might wrongly permit second
1303 writing, which will cause grant space > avaiable space.
1306 Frequency : when client is evicted
1308 Description: Grant might be wrongly erased when osc is evicted by OST
1309 Details : when the import is evicted by server, it will fork another
1310 thread ptlrpc_invalidate_import_thread to invalidate the
1311 import, where the grant will be set to 0. While the original
1312 thread will update the grant it got when connecting. So if
1313 the former happened latter, the grant will be wrongly errased
1314 because of this race.
1319 Description: Checking Stale with correct fid
1320 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1321 to check whether it is stale, because sometimes, we want the
1322 enqueue happened anyway, and op_data.fid2 will not be initialized.
1324 Severity : enhancement
1326 Description: update patchless client
1327 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1330 Frequency : only with 2.4 kernel
1332 Description: random memory corruption
1333 Details : size of struct ll_inode_info is to big for union inode.u and this
1334 can be cause of random memory corruption.
1339 Description: Memory leak in recovery
1340 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1341 It should also check obd_fail before fsfilt_start, otherwise if
1342 fsfilt_start return -EROFS,(failover mds during mds recovery).
1343 then the req will return with repmsg->transno = 0 and rc = EROFS.
1344 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1345 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1346 be freed no matter whether fsfilt_commit success or not.
1349 Frequency : only with huge count clients
1351 Description: Prevents from taking the superblock lock in llap_from_page for
1353 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1354 for llap_from_page call in ll_removepage prevents from taking the
1355 superblock lock for a soon died page.
1360 Description: Not check open intent error before release open handle
1361 Details : in some rare cases, the open intent error is not checked before
1362 release open handle, which may cause
1363 ASSERTION(open_req->rq_transno != 0), because it tries to release
1364 the failed open handle.
1369 Description: Set cat log bitmap only after create log success.
1370 Details : in some rare cases, the cat log bitmap is set too early. and it
1371 should be set only after create log success.
1375 Description: Accessing a block bevice can re-enable I/O when Lustre is
1376 tearing down a device.
1377 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1382 Description: service threads may hog cpus when there are a lot of requests
1383 Details : Insert cond_resched to give other threads a chance to use some CPU
1388 Description: the cat log was not initialized in recovery
1389 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1390 unlink log on mds will not be initialized until mds post
1391 recovery. And also in mds post recovery, the unlink log will
1392 initialization will be done asynchronausly, so there will be race
1393 between add unlink log and unlink log initialization.
1397 Description: brw_stats were being printed incorrectly
1398 Details : brw_stats were being printed as log2 but all of them were not
1399 recorded as log2. Also remove some code duplication arising from
1400 filter_tally_{read,write}.
1404 Frequency : rare, only in recovery.
1405 Description: ASSERTION(req->rq_type != LI_POISON) failed
1406 Details : imp_lock should be held while iterating over imp_sending_list for
1407 prevent destroy request after get timeout in ptlrpc_queue_wait.
1411 Description: replay-single.sh test 52 fails
1412 Details : A lock's skiplist need to be cleanup when it being unlinked
1413 from its resource list.
1417 Description: Short directio read returns full requested size rather than
1419 Details : Direct I/O operations should return actual amount of bytes
1420 transferred rather than requested size.
1422 Severity : enhancement
1424 Description: metadata RPC reduction (e.g. for rm performance)
1425 Details : decrease the amount of synchronous RPC between clients and servers
1426 by canceling conflicing lock before the operation on the client
1427 and packing thier handles into the main operation RPC to server.
1429 Severity : enhancement
1431 Description: Async OSC create to avoid the blocking unnecessarily.
1432 Details : If a OST has no remain object, system will block on the creating
1433 when need to create a new object on this OST. Now, ways use
1434 pre-created objects when available, instead of blocking on an
1435 empty osc while others are not empty. If we must block, we block
1436 for the shortest possible period of time.
1438 Severity : enhancement
1440 Description: refine locking for avoid write wrong info into lov_objid file
1441 Details : fix possible races with add new target and write/update data in
1444 --------------------------------------------------------------------------------
1446 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1451 Frequency : on some architectures
1453 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1454 Details : Uninitialized flags sometimes cause configuration commands to
1458 Frequency : patchless clients only
1460 Description: missing __iget() symbol export
1461 Details : The __iget() symbol export is missing. To avoid the need for
1462 this on patchless clients the deathrow inode reaper is turned
1463 off, and we depend on the VM to clean up old inodes. This
1464 dependency was during via the fix for bug 12181.
1466 --------------------------------------------------------------------------------
1468 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1470 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1471 INTEROPERATE with older versions automatically. In many cases a
1472 special upgrade step is needed. Please read the
1473 user documentation before upgrading any part of a live system.
1474 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1475 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1476 * WARNING: Lustre configuration and startup changes are required with
1477 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1483 Severity : enhancement
1485 Description: MountConf
1486 Details : Lustre configuration is now managed via mkfs and mount
1487 commands instead of lmc and lconf. New obd types (MGS, MGC)
1488 are added for dynamic configuration management. See
1489 https://mail.clusterfs.com/wikis/lustre/MountConf for
1492 Severity : enhancement
1494 Description: dynamic OST addition
1495 Details : OSTs can now be added to a live filesystem
1497 Severity : enhancement
1499 Description: startup order invariance
1500 Details : MDTs and OSTs can be started in any order. Clients only
1501 require the MDT to complete startup.
1503 Severity : enhancement
1505 Description: parallel, asynchronous orphan cleanup
1506 Details : orphan cleanup is now performed in separate threads for each
1507 OST, allowing parallel non-blocking operation.
1509 Severity : enhancement
1511 Description: optimized stripe assignment
1512 Details : stripe assignments are now made based on ost space available,
1513 ost previous usage, and OSS previous usage, in order to try
1514 to optimize storage space and networking resources.
1516 Severity : enhancement
1518 Description: Permanently set tunables
1519 Details : All writable /proc/fs/lustre tunables can now be permanently
1520 set on a per-server basis, at mkfs time or on a live system.
1522 Severity : enhancement
1524 Description: Lustre message v2
1525 Details : Add lustre message format v2.
1527 Severity : enhancement
1529 Description: client OST exclusion list
1530 Details : Clients can be started with a list of OSTs that should be
1531 declared "inactive" for known non-responsive OSTs.
1535 Description: SPEC SFS validation failure on NFS v2 over lustre.
1536 Details : Changes the blocksize for regular files to be 2x RPC size,
1537 and not depend on stripe size.
1539 Severity : enhancement
1541 Description: Multiple MD RPCs in flight.
1542 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1543 To avoid overly-overloading MDS, introduce a limit on number of
1544 MDS RPCs in flight for a single client and add /proc controls
1545 to adjust this limit.
1547 Severity : enhancement
1549 Description: client read/write statistics
1550 Details : Add client read/write call usage stats for performance
1551 analysis of user processes.
1552 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1553 file access. extents_stats shows chunk size distribution.
1554 extents_stats_per_process show chunk size distribution per
1557 Severity : enhancement
1559 Description: mds statistics
1560 Details : Add detailed mds operations statistics in
1561 /proc/fs/lustre/mds/*/stats.
1565 Description: Failure of copying files with lustre special EAs.
1566 Details : Client side always return success for setxattr call for lustre
1567 special xattr (currently only "trusted.lov").
1572 Description: Refcount LNET uuids
1573 Details : The global LNET uuid list grew linearly with every startup;
1574 refcount repeated list entries instead of always adding to
1577 Severity : enhancement
1579 Description: Dynamic service threads
1580 Details : Within a small range, start extra service threads
1581 automatically when the request queue builds up.
1584 Frequency : mixed-endian client/server environments
1586 Description: mixed-endian crashes
1587 Details : The new msg_v2 system had some failures in mixed-endian
1590 Severity : enhancement
1592 Description: Easy OST removal
1593 Details : OSTs can be permanently deactivated with e.g. 'lctl
1594 conf_param lustre-OST0001.osc.active=0'
1596 Severity : enhancement
1598 Description: MGS proc entries
1599 Details : Added basic proc entries for the MGS showing what filesystems
1602 Severity : enhancement
1604 Description: provide MGS failover
1605 Details : Added config lock reacquisition after MGS server failover.
1607 Severity : enhancement
1609 Description: add Linux 2.4 support
1610 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1614 Description: a large application tries to do I/O to the same resource and dies
1615 in the middle of it.
1616 Details : Check the req->rq_arrival time after the call to
1617 ost_brw_lock_get(), but before we do anything about
1618 processing it & sending the BULK transfer request. This
1619 should help move old stale pending locks off the queue as
1620 quickly as obd_timeout.
1623 Frequency : when an incorrect nid is specified during startup
1625 Description: ptlrpc connect to non-existant node causes kernel crash
1626 Details : LNET can't be re-entered from an event callback, which
1627 happened when we expire a message after the export has been
1628 cleaned up. Instead, hand the zombie cleanup off to another
1631 Severity : enhancement
1633 Description: plain/inodebits lock performance improvement
1634 Details : Grouping plain/inodebits in granted list by their request modes
1635 and bits policy, thus improving the performance of search through
1639 Frequency : only if OST filesystem is corrupted
1641 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1642 Details : for a short time RPCs with bulk IO are in the replay list,
1643 but replay of bulk IOs is unimplemented. If the OST filesystem
1644 is corrupted due to disk cache incoherency and then replay is
1645 started it is possible to trip an assertion. Avoid putting
1646 committed RPCs into the replay list at all to avoid this issue.
1649 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1651 Description: System hang on startup
1652 Details : This bug allowed the liblustre (e.g. catamount) client to
1653 return to the app before handling all startup RPCs. This
1654 could leave the node unresponsive to lustre network traffic
1655 and manifested as a server ptllnd timeout.
1657 Severity : enhancement
1659 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1660 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1661 boolean to control whether to print peer debug info when a
1662 client's RPC times out.
1665 Frequency : only for kernels with patches from Lustre below 1.4.3
1667 Description: Remove old rdonly API
1668 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1671 Frequency : only for devices with external journals
1673 Description: Set external device read-only also
1674 Details : During a commanded failover stop, we set the disk device
1675 read-only while the server shuts down. We now also set any
1676 external journal device read-only at the same time.
1679 Frequency : when upgrading from 1.4 while trying to change parameters
1681 Description: The wrong (new) MDC name was used when setting parameters for
1682 upgraded MDT's. Also allows changing of OSC (and MDC)
1683 parameters if --writeconf is specified at tunefs upgrade time.
1686 Frequency : when setting specific ost indicies
1688 Description: QOS code breaks on skipped indicies
1689 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1690 created with --index need not be sequential.
1692 Severity : enhancement
1694 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1695 Details : The uninit_groups feature works in conjunction with the kernel
1696 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1697 up the pass1 processing of e2fsck. This is a read-only feature
1698 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1699 mount filesystems that have had this feature enabled.
1701 Severity : enhancement
1703 Description: Improve multi-block allocation algorithm to avoid fragmentation
1704 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1705 allocation locality and avoid filesystem fragmentation.
1707 ------------------------------------------------------------------------------
1709 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1711 * Support for kernels:
1712 2.4.21-47.0.1.EL (RHEL 3)
1713 2.6.5-7.283 (SLES 9)
1714 2.6.9-42.0.10.EL (RHEL 4)
1715 2.6.12.6 vanilla (kernel.org)
1716 2.6.16.27-0.9 (SLES 10)
1717 * Recommended e2fsprogs version: 1.39.cfs5
1719 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1723 Frequency : occasional, depends on client load and configuration
1724 Bugzilla : 12181, 12203
1725 Description: data loss for recently-modified files
1727 Details : In some cases it is possible that recently written or created
1728 files may not be written to disk in a timely manner (this should
1729 normally be within 30s unless client IO load is very high).
1730 The problem appears as zero-length files or files that are a
1731 multiple of 1MB in size after a client crash or client eviction
1732 that are missing data at the end of the file.
1734 This problem is more likely to be hit on clients where files are
1735 repeatedly created and unlinked in the same directory, clients
1736 have a large amount of RAM, have many CPUs, the filesystem has
1737 many OSTs, the clients are rebooted frequently, and/or the files
1738 are not accessed by other nodes after being written.
1740 The presence of the problem can be detected by looking at
1741 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1742 smaller than the second (nr_unused) then dirty files will not
1743 be flushed automatically to disk. "sync; sleep 10" should be
1744 run several times on the node before unmounting it to update
1745 Lustre (this is also safe to run on nodes without this problem).
1747 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1748 that can cause this same problem, so customers using that kernel
1749 also need to update the kernel in addition to Lustre. In order
1750 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1752 It is normal that files written just before a client crash (less
1753 than 30s) may not yet have been flushed to disk, even for local
1757 Frequency : frequent on thin XT3 nodes
1759 Description: UUID collision on thin XT3 Linux nodes
1760 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1761 randomly, since we relied on an insufficiently-seeded PRNG.
1766 Description: OSS hangs after "All ost request buffers busy"
1767 Details : A deadlock between quota and journal operations caused OSS
1768 hangs after printing "All ost request buffers busy."
1771 Frequency : always on liblustre builds
1773 Description: Cleanup compiler warnings on liblustre
1776 Frequency : always on liblustre builds on XT3
1778 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1782 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1783 > 32000 subdirectories
1784 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1785 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1786 subdirectory count crosses 32000. This will aid e2fsck to
1787 correctly handle more than 32000 subdirectories.
1790 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1792 Description: System hang on startup
1793 Details : This bug allowed the liblustre (e.g. catamount) client to
1794 return to the app before handling all startup RPCs. This
1795 could leave the node unresponsive to lustre network traffic
1796 and manifested as a server ptllnd timeout.
1798 Severity : enhancement
1800 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1801 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1802 boolean to control whether to print peer debug info when a
1803 client's RPC times out.
1808 Description: make O_SYNC working on 2.6 kernels
1809 Details : 2.6 kernels use different method for mark pages for write,
1810 so need add a code to lustre for O_SYNC work.
1815 Description: Failure to close file and release space on NFS
1816 Details : Put inode details into lock acquired in ll_intent_file_open.
1817 Use mdc_intent_lock in ll_intent_open to properly
1818 detect all kind of errors unhandled by mdc_enqueue.
1823 Description: proc file read during shutdown sometimes raced obd removal,
1825 Details : Add lock to prevent obd access after proc file removal.
1828 Frequency : Only for files larger than 4GB on 32-bit clients.
1830 Description: improperly doing page alignment of locks
1831 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1832 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1837 Description: RPCs being resent when they shouldn't be
1838 Details : Some RPCs that should not be resent are being resent. This
1839 can cause inconsistencies in the RPC state machine. Do not
1840 resend such requests.
1843 Frequency : rare, only with NFS export
1845 Description: Crash on NFS re-export node
1846 Details : under very unusual load conditions an assertion is hit in
1847 ll_intent_file_open()
1850 Frequency : only if OST filesystem is corrupted
1852 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1853 Details : for a short time RPCs with bulk IO are in the replay list,
1854 but replay of bulk IOs is unimplemented. If the OST filesystem
1855 is corrupted due to disk cache incoherency and then replay is
1856 started it is possible to trip an assertion. Avoid putting
1857 committed RPCs into the replay list at all to avoid this issue.
1862 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1863 Details : Large single O_DIRECT read and write calls can fail to allocate
1864 a sufficiently large buffer to process the request. In case of
1865 allocation failure the allocation is retried with a smaller
1866 buffer and broken into smaller requests.
1868 Severity : enhancement
1870 Description: Add -o localflock option to simulate old noflock behaviour.
1871 Details : This will achieve local-only flock/fcntl locks coherentness.
1876 Description: versioning check is incomplete
1877 Details : Checking the version difference of client vs. server, report
1878 error if the gap is too big.
1883 Description: add support PG_writeback bit
1884 Details : add support for PG_writeback bit for Lustre, for more carefull
1885 work with page cache in 2.6 kernel. This also fix some deadlocks
1886 and remove hack for work O_SYNC with 2.6 kernel.
1888 Severity : enhancement
1890 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1891 Details : The uninit_groups feature works in conjunction with the kernel
1892 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1893 up the pass1 processing of e2fsck. This is a read-only feature
1894 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1895 mount filesystems that have had this feature enabled.
1897 Severity : enhancement
1899 Description: Improve multi-block allocation algorithm to avoid fragmentation
1900 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1901 allocation locality and avoid filesystem fragmentation.
1903 ------------------------------------------------------------------------------
1905 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1907 * Support for kernels:
1908 2.6.9-42.0.3.EL (RHEL 4)
1909 2.6.5-7.276 (SLES 9)
1910 2.4.21-47.0.1.EL (RHEL 3)
1911 2.6.12.6 vanilla (kernel.org)
1912 2.6.16.21-0.8 (SLES10)
1913 * Recommended e2fsprogs version: 1.39.cfs2-0
1915 * The backwards-compatible /proc/sys/portals symlink has been removed
1916 in this release. Before upgrading, please ensure that you change
1917 any configuration scripts or /etc/sysctl.conf files that access
1918 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1919 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1920 in advance of the upgrade on any system running Lustre 1.4.6 or
1921 newer, since /proc/sys/lnet was added in that version.
1922 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1926 Frequency : only when quota is used
1928 Description: avoid scanning export list for quota master
1929 Details : Change the algorithms to avoid scanning export list in order
1930 to improve the efficiency.
1933 Frequency : MDS failover only, very rarely
1935 Description: "went back in time" messages on mds failover
1936 Details : The greatest transno may be lost when the current operation
1937 finishes with an error (transno==0) and the client's last_rcvd
1938 record is over-written. Save the greatest transno in the
1939 mds_last_transno for this case.
1942 Frequency : always for specific kernels and striping counts
1944 Description: client may get "Matching packet too big" without ACL support
1945 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1946 when trying to access files in certain configurations. The
1947 clients should in fact be denied when mounting because they do
1948 not understand ACLs.
1951 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1953 Description: many clients connecting with IO in progress causes connect timeouts
1954 Details : Avoid synchronous journal commits to avoid delays caused by many
1955 clients connecting/disconnecting when bulk IO is in progress.
1956 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1957 OST_IO_PORTAL to avoid delays behind potentially many pending
1961 Frequency : occasionally with multiple writers to a single file
1963 Description: shared writes to file may result in wrong size reported by stat()
1964 Details : Allow growing of kms when extent lock is cancelled
1967 Frequency : always with random mmap IO to multi-striped file
1969 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1970 Details : Only if the hole is at the end of OST object so that kms is too
1971 small. Fix is to increase kms accordingly in ll_nopage.
1974 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1976 Description: writes to a missing object would leak memory on the OST
1977 Details : If there is an inconsistency between the MDS and OST filesystems,
1978 such that the MDS references an object that doesn't exist, writes
1979 to that object will leak memory due to incorrect cleanup in the
1980 error handling path, eventually running out of memory on the OST.
1985 Description: Creating too long symlink causes lustre errors
1986 Details : Check symlink and name lengths before sending requests to MDS.
1989 Frequency : only if flock is enabled (not on by default)
1991 Description: posix locks not released on fd closure on 2.6.9+
1992 Details : We failed to add posix locks to list of inode locks on 2.6.9+
1993 kernels, this caused such locks not to be released on fd close and
1994 then assertions on fs unmount about still used locks.
1997 Frequency : MDS failover only, very rarely
1999 Description: clients may get ASSERTION(granted_lock != NULL)
2000 Details : When request was taking a long time, and a client was resending
2001 a getattr by name lock request. The were multiple lock requests
2002 with the same client lock handle and
2003 mds_getattr_name->fixup_handle_for_resent_request found one of the
2004 lock handles but later failed with ASSERTION(granted_lock != NULL).
2009 Description: handle->h_buffer_credits > 0, assertion failure
2010 Details : h_buffer_credits is zero after truncate, causing assertion
2011 failure. This patch extends the transaction or creates a new
2015 Frequency : NFS re-export or patchless client
2016 Bugzilla : 11179, 10796
2017 Description: Crash on NFS re-export node (__d_move)
2018 Details : We do not want to hash the dentry if we don't have a lock.
2019 But if this dentry is later used in d_move, we'd hit uninitialised
2020 list head d_hash, so we just do this to init d_hash field but
2021 leave dentry unhashed.
2024 Frequency : NFS re-export or patchless client
2026 Description: NFS exports has problem with symbolic link
2027 Details : lustre client didn't properly install dentry when re-exported
2028 to NFS or running patchless client.
2031 Frequency : NFS re-export or patchless client
2033 Description: Various nfs/patchless fixes.
2034 Details : fixes reuse disconected alias for lookup process - this fixes
2035 warning "find_exported_dentry: npd != pd",
2036 fix permission error with open files at nfs.
2037 fix apply umask when do revalidate.
2040 Frequency : occasional
2042 Description: Crash on NFS re-export node
2043 Details : calling clear_page() on the wrong pointer triggered oops in
2044 generic_mapping_read().
2047 Frequency : rarely, using O_DIRECT IO
2049 Description: unaligned directio crashes client with LASSERT
2050 Details : check for unaligned buffers before trying any requests.
2053 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2055 Description: stale data returned from RAID cache
2056 Details : If only a small amount of IO is done to the RAID device before
2057 reading it again it is possible to get stale data from the RAID
2058 cache instead of reading it from disk.
2061 Frequency : always for sles10 kernel
2063 Description: sles10 support
2064 Details : ll_follow_link: compile fixes and using of nd_set_link
2065 under newer kernels.
2068 Frequency : depends on arch, kernel and compiler version, always on sles10
2071 Description: recursive or deep enough symlinks cause stack overflow
2072 Details : getting rid of large stack-allocated variable in
2076 Frequency : depends on hardware
2078 Description: lustre write performance loss in the SLES10 kernel
2079 Details : the performance loss is caused by using of write barriers in the
2080 ext3 code. The SLES10 kernel turns barrier support on by
2081 default. The fix is to undo that change for ldiskfs.
2083 ------------------------------------------------------------------------------
2085 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2087 * Support for kernels:
2088 2.6.9-42.0.3EL (RHEL 4)
2089 2.6.5-7.276 (SLES 9)
2090 2.4.21-47.0.1.EL (RHEL 3)
2091 2.6.12.6 vanilla (kernel.org)
2095 Frequency : quota enabled and large files being deleted
2097 Description: releasing more than 4GB of quota at once hangs OST
2098 Details : If a user deletes more than 4GB of files on a single OST it
2099 will cause the OST to spin in an infinite loop. Release
2100 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2105 Description: statfs data retrieved from /proc may be stale or zero
2106 Details : When reading per-device statfs data from /proc, in the
2107 {kbytes,files}_{total,free,avail} files, it may appear
2108 as zero or be out of date.
2111 Frequency : systems with MD RAID1 external journal devices
2113 Description: lconf's call to blkid is confused by RAID1 journal devices
2114 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2115 of returning all block devices that match the journal UUID.
2118 Frequency : always, for aggregate stripe size over 4GB
2120 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2121 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2122 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2125 Frequency : always if "lfs find" used on a local file/directory
2127 Description: "lfs find" segfaults if used on a local file/directory
2128 Details : The case where a directory component was not specified wasn't
2129 handled correctly. Handle this properly.
2132 Frequency : always on ppc64
2134 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2135 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2136 perform endianness conversion.
2139 Frequency : rarely (truncate to non-zero file size after write under load)
2140 Bugzilla : 10730, 10687
2141 Description: Files padded with zeros to next 4K multiple
2142 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2143 it is possible that files that are truncated to a non-zero size
2144 immediately after being written are filled with zero bytes beyond
2145 the truncated size. No file data is lost.
2147 Severity : enhancement
2149 Description: Allow recovery/failover for liblustre clients.
2150 Details : liblustre clients were unaware of failover configurations until
2153 Severity : enhancement
2155 Description: user file locks should fail when not mounting with flock option
2156 Details : Set up an error-returning stub in ll_file_operations.lock field
2157 to prevent incorrect behaviour when client is mounted without
2158 flock option. Also, set up properly f_op->flock field for
2162 Frequency : always on ia64
2164 Description: "lfs df" loops on printing out MDS statfs information
2165 Details : The obd_ioctl_data was not initialized and in some systems
2166 this caused a failure during the ioctl that did not return
2167 an error. Initialize the struct and return an error on failure.
2170 Frequency : SLES 9 only
2172 Description: Error of copying files with lustre special EAs as root
2173 Details : Client side always return success for setxattr call for lustre
2174 special xattr (currently only "trusted.lov").
2177 Frequency : rarely on clusters with both ia64+i386 clients
2179 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2180 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2181 clients are doing concurrent writes to the same file it is
2182 possible that the ia64 clients may LASSERT because the OST
2183 extent locks are not PAGE_SIZE aligned. Ensure that grown
2184 locks are always aligned on the request boundary.
2187 Frequency : specific use, occasional
2189 Description: Overwriting in use executable truncates on-disk binary image
2190 Details : If one node attempts to overwrite an executable in use by
2191 another node, we now correctly return ETXTBSY instead of
2192 truncating the file.
2194 Severity : enhancement
2196 Description: Async OSC create to avoid the blocking unnecessarily.
2197 Details : If a OST has no remain object, system will block on the creating
2198 when need to create a new object on this OST. Now, ways use
2199 pre-created objects when available, instead of blocking on an
2200 empty osc while others are not empty. If we must block, we block
2201 for the shortest possible period of time.
2206 Description: chmod on Lustre root is propagated to other clients
2207 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2208 invalid by the follow_mount time.
2211 Frequency : liblustre clients only
2213 Description: Race in 'instant cancel' lock handling could lead to such locks
2214 never to be granted in case of SMP MDS
2215 Details : Do not destroy not yet granted but cbpending locks in
2219 Frequency : replay/resend of open
2221 Description: non null lock assetion failure in mds_intent_policy
2222 Details : Trying to replay/resend lockless open requests resulted in
2223 mds_open() returning 0 with no lock. Now it sets a flag if
2224 a lock is going to be returned.
2226 Severity : enhancement
2228 Description: Checksum enhancements
2229 Details : New checksum enhancements allow for resending RPCs that failed
2232 Severity : enhancement
2234 Description: Tunables on number of dirty pages in cacche
2235 Details : Allow to set limit on number of dirty pages cached.
2240 Description: client crash on unmount - lock still has references
2241 Details : In some error handling cases it was possible to leak a lock
2242 reference on a client while accessing a file. This was not
2243 harmful to the client during operation, but would cause the
2244 client to crash when the filesystem is unmounted.
2247 Frequency : specific case, rare
2249 Description: ETXTBSY on mds though file not in use by client
2250 Details : ETXTBSY is no longer incorrectly returned when attempting to
2251 chmod or chown a directory that the user previously tried to
2252 execute or a currently-executing binary.
2255 Frequency : extremely rare except on liblustre-based clients
2257 Description: Lustre space not freed when files are deleted
2258 Details : Clean up open-unlinked files after client eviction. Previously
2259 the unlink was skipped and the files remained as orphans.
2264 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2265 Details : In some cases it was possible to send a blocking callback to
2266 a client doing a glimpse, even though that client didn't get
2267 a lock granted. When the glimpse lock is cancelled on the OST
2268 the freed lock is left on the waiting list and corrupted the list.
2271 Frequency : all core dumps
2273 Description: Broke core dumps to lustre
2274 Details : Negative dentry may be unhashed if parent does not have UPDATE
2275 lock, but some callers, e.g. do_coredump, expect dentry to be
2276 hashed after successful create, hash it in ll_create_it.
2278 ------------------------------------------------------------------------------
2280 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2282 * Support for kernels:
2283 2.6.9-42.0.2.EL (RHEL 4)
2284 2.6.5-7.276 (SLES 9)
2285 2.4.21-40.EL (RHEL 3)
2286 2.6.12.6 vanilla (kernel.org)
2290 Frequency : always on RHEL 3
2292 Description: Number of open files grows over time
2293 Details : The number of open files grows over time, whether or not
2294 Lustre is started. This was due to a filp leak introduced
2295 by one of our kernel patches.
2297 ------------------------------------------------------------------------------
2299 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2301 * Support for kernels:
2302 2.6.9-42.EL (RHEL 4)
2303 2.6.5-7.276 (SLES 9)
2304 2.4.21-40.EL (RHEL 3)
2305 2.6.12.6 vanilla (kernel.org)
2310 Bugzilla : 5719, 9635, 9792, 9684
2311 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2312 Details : If a server is under heavy load and cannot reply to new
2313 connection requests before the client resends the (re)connect,
2314 the connection handling code can behave badly if two service
2315 threads are concurrently handing separate (re)connections from
2316 the same client. Add better locking to the connection handling
2317 code, and ensure that only a single connection will be processed
2318 for a given client UUID, even if the lock is dropped.
2320 Severity : enhancement
2322 Description: add TCP zero-copy support to kernel
2323 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2324 sends if the hardware supports scatter-gather and checksumming.
2325 This allows socklnd to do client-write and server-read more
2326 efficiently and reduce CPU utilization from skbuf copying.
2329 Frequency : only if NFS exporting from client
2331 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2332 Details : If setting attributes on a file created under NFS that had
2333 never been opened it would be possible to oops the client
2334 if the file had no objects.
2337 Frequency : always for liblustre
2339 Description: liblustre client does MDS+OSTs setattr RPC for each write
2340 Details : When doing a write from a liblustre client, the client
2341 incorrectly issued an RPC to the MDS and each OST the file was
2342 striped over in order to update the timestamps. When writing
2343 with small chunks and many clients this could overwhelm the MDS
2344 with RPCs. In all cases it would slow down the write because
2345 these RPCs are unnecessary.
2347 Severity : enhancement
2349 Description: allow number of MDS service threads to be changed at module load
2350 Details : It is now possible to change the number of MDS service threads
2351 running. Adding "options mds mds_num_threads={N}" to the MDS's
2352 /etc/modprobe.conf will set the number of threads for the next
2353 time Lustre is restarted (assuming the "mds" module is also
2354 reloaded at that time). The default number of threads will
2355 stay the same, 32 for most systems.
2360 Description: OST crash if filesystem is unformatted or corrupt
2361 Details : If an OST is started on a device that has never been formatted
2362 or if the filesystem is corrupt and cannot even mount then the
2363 error handling cleanup routines would dereference a NULL pointer.
2368 Description: NULL pointer deref in llap_from_page.
2369 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2370 incorrect) mapping if the page was truncated/reclaimed while it was
2371 searched for. Check for this condition and skip such pages when
2372 doing readahead. Introduce extra check to llap_from_page() to
2373 verify page->mapping->host is non-NULL (so page is not anonymous).
2376 Frequency : Sometimes when using sys_sendfile
2378 Description: "page not covered by a lock" warnings from ll_readpage
2379 Details : sendfile called ll_readpage without right page locks present.
2380 Now we introduced ll_file_sendfile that does necessary locking
2381 around call to generic_file_sendfile() much like we do in
2385 Frequency : with certain MDS communication failures at client mount time
2387 Description: NULL pointer deref after failed client mount
2388 Details : a client connection request may delayed by the network layer
2389 and not be sent until after the PTLRPC layer has timed out the
2390 request. If the client fails the mount immediately it will try
2391 to clean up before the network times out the request. Add a
2392 reference from the request import to the obd device and delay
2393 the cleanup until the network drops the request.
2396 Frequency : occasionally during client (re)connect
2398 Description: assertion failure during client (re)connect
2399 Details : processing a client connection request may be delayed by the
2400 client or server longer than the client connect timeout. This
2401 causes the client to resend the connection request. If the
2402 original connection request is replied in this interval, the
2403 client may trip an assertion failure in ptlrpc_connect_interpret()
2404 which thought it would be the only running connect process.
2407 Frequency : only with obd_echo servers and clients that are rebooted
2409 Description: kernel BUG accessing uninitialized data structure
2410 Details : When running an obd_echo server it did not start the ping_evictor
2411 thread, and when a client was evicted an uninitialized data
2412 structure was accessed. Start the ping_evictor in the RPC
2413 service startup instead of the OBD startup.
2415 Severity : enhancement
2416 Bugzilla : 10193 (patchless)
2417 Description: Remove dependency on various unexported kernel interfaces.
2418 Details : No longer need reparent_to_init, exit_mm, exit_files,
2419 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2422 Frequency : rare (only users of deprecated and unsupported LDAP config)
2424 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2425 Details : LDAP apparently contains 'lustreName' attributes instead of
2426 'name'. A simple remapping of the name is sufficient.
2429 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2431 Description: waiting_locks_callback trips kernel BUG if client is evicted
2432 Details : Running with the dump_on_timeout debug flag turned on makes
2433 it possible that the waiting_locks_callback() can try to dump
2434 the Lustre kernel debug logs from an interrupt handler. Defer
2435 this log dumping to the expired_lock_main() thread.
2437 Severity : enhancement
2439 Description: Support NFS exporting on 2.6 kernels.
2440 Details : Implement non-rawops metadata methods for NFS server to use without
2441 changing NFS server code.
2444 Frequency : very rare (synthetic metadata workload only)
2446 Description: two racing renames might cause an MDS thread to deadlock
2447 Details : Running the "racer" program may cause one MDS thread to rename
2448 a file from being the source of a rename to being the target of
2449 a rename at exactly the same time that another thread is doing
2450 so, and the second thread has already enqueued these locks after
2451 doing a lookup of the target and is trying to relock them in
2452 order. Ensure that we don't try to re-lock the same resource.
2455 Frequency : only very large systems with liblustre clients
2457 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2458 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2459 all OSTs in parallel. This allows the eviction of stale liblustre
2460 clients to proceed much faster than if they were done in series,
2461 and also offers similar improvements for other set_info RPCs.
2466 Description: excessive CPU usage during initial read phase on client
2467 Details : During the initial read phase on a client, it would agressively
2468 retry readahead on the file, consuming too much CPU and impacting
2469 performance (since 1.4.5.8). Improve the readahead algorithm
2470 to avoid this, and also improve some other common cases (read
2471 of small files in particular, where "small" is files smaller than
2472 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2477 Description: MDS crash when receiving packet with unknown intent.
2478 Details : Do not LBUG in unknown intent case, just return -EFAULT
2480 Severity : enhancement
2481 Bugzilla : 9293, 9385
2482 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2483 Details : Do not serialize getattr (non-intent version) and statfs.
2486 Frequency : occasional, when OST network is overloaded/intermittent
2488 Description: client evicted by OST after bulk IO timeout
2489 Details : If a client sends a bulk IO request (read or write) the OST
2490 may evict the client if it is unresposive to its data GET/PUT
2491 request. This is incorrect if the network is overloaded (takes
2492 too long to transfer the RPC data) or dropped the OST GET/PUT
2493 request. There is no need to evict the client at all, since
2494 the pinger and/or lock callbacks will handle this, and the
2495 client can restart the bulk request.
2498 Frequency : Always when mmapping file with no objects
2500 Description: client crashes when mmapping file with no objects
2501 Details : Check that we actually have objects in a file before doing any
2502 operations on objects in ll_vm_open, ll_vm_close and
2508 Description: Request leak when working with deleted CWD
2509 Details : Introduce advanced request refcount tracking for requests
2510 referenced from lustre intent.
2512 Severity : Enhancement
2514 Description: Cache open file handles on client.
2515 Details : MDS now will return special lock along with openhandle, if
2516 requested and client is allowed to hold openhandle, even if unused,
2517 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2518 closing files for every read/write openration.
2520 Severity : Enhancement
2522 Description: Cache open negative dentries on client when possible.
2523 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2524 negative dentries on lock revocation.
2529 Description: Remounting a client read-only wasn't possible with a zconf mount
2530 Details : It wasn't possible to remount a client read-only with llmount.
2532 Severity : enhancement
2533 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2534 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2535 MPICH is included to add Lustre-specific ADIO interfaces.
2536 This is based closely on the UFS ADIO layer and only differs
2537 in file creation, in order to allow the OST striping to be set.
2538 This is user-contributed code and not supported by CFS.
2543 Description: extended inode attributes (immutable, append-only) work improperly
2544 when 2.4 and 2.6 kernels are used on client/server or vice versa
2545 Details : Introduce kernel-independent values for these flags.
2547 Severity : enhancement
2550 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2551 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2552 granularity. Now, Fractional MB granularity is be supported,
2553 this is very useful for low memory system.
2555 Severity : enhancement
2557 Description: Getattr by fid
2558 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2559 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2560 have locks and avoids getattr rpc requests for it on every stat.
2563 Frequency : Always, for filesystems larger than 2TB
2565 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2566 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2567 filesystems larger than 2TB to be mounted with mballoc enabled.
2570 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2572 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2573 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2574 filesystems larger than 2TB is mounted then the kernel will
2575 silently corrupt the start of the filesystem. CONFIG_LBD is
2576 enabled for all CFS-supported kernels, but the possibility of
2577 this happening with a modified kernel config exists.
2579 Severity : enhancement
2581 Description: add client O_DIRECT support for 2.6 kernels
2582 Details : It is now possible to do O_DIRECT reads and writes to files
2583 in the Lustre client mountpoint on 2.6 kernel clients.
2585 Severity : enhancement
2587 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2588 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2589 parallel, not waiting for response from every OST before sending
2590 a rpc to the next OST.
2595 Description: setattr vs write race when updating file timestamps
2596 Details : Client processes that update a file timestamp into the past
2597 right after writing to the file (e.g. tar) it is possible that
2598 the updated file modification time can be reset to the current
2599 time due to a race between processing the setattr and write RPC.
2601 Severity : enhancement
2603 Description: Bring 'lfs find' closer in line with regular Linux find.
2604 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2605 -print0 options and obtains all the needed info through the lustre
2608 Severity : enhancement
2610 Description: support up to 1024 configured devices on one node
2611 Details : change obd_dev array from statically allocated to dynamically
2612 allocated structs as they are first used to reduce memory usage
2617 Description: Flush dirty partially truncated pages during truncate
2618 Details : Immediatelly flush partially truncated pages in filter_setattr,
2619 this way we completely avoid having any pages in page cache on OST
2620 and can retire ugly workarounds during writes to flush such pages.
2625 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2626 Details : protect inode from truncation within vfs_unlink() context
2627 just take a reference before calling vfs_unlink() and release it
2628 when parent's i_sem is free.
2633 Description: last_id value checked outside lock on OST caused LASSERT failure
2634 Details : If there were multiple MDS->OST object precreate requests in
2635 flight, it was possible that the OST's last object id was checked
2636 outside a lock and incorrectly tripped an assertion. Move checks
2637 inside locks, and discard old precreate requests.
2640 Frequency : always, if extents are used on OSTs
2642 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2643 Details : OSTs using the extents format would not zero the high 16 bits of
2644 the index physical block number. This is not a problem for any
2645 OST filesystems smaller than 16TB, and no kernels support ext3
2646 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2647 new/modified files) and can be fixed for existing filesystems
2648 with e2fsprogs-1.39-cfs1.
2653 Description: import connection selection may be incorrect if timer wraps
2654 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2655 import connections to be ignored if the 32-bit jiffies counter
2656 wraps. Use a 64-bit jiffies counter.
2659 Frequency : very large clusters immediately after boot
2661 Description: LNET request buffers exhausted under heavy short-term load
2662 Details : If a large number of client requests are generated on a service
2663 that has previously never seen so many requests it is possible
2664 that the request buffer growth cannot keep up with the spike in
2665 demand. Instead of dropping incoming requests, they are held in
2666 the LND until the RPC service can accept more requests.
2669 Frequency : Sometimes during replay
2671 Description: Assertion failure in ll_local_open after replay.
2672 Details : If replay happened on an open request reply before we were able
2673 to set replay handler, reply will become not swabbed tripping the
2674 assertion in ll_local_open. Now we set the handler right after
2675 recognising of open request
2678 Frequency : very rare
2680 Description: kernel reports "badness in vsnprintf"
2681 Details : Reading from the "recovery_status" /proc file in small chunks
2682 may cause a negative length in lprocfs_obd_rd_recovery_status()
2683 call to vsnprintf() (which is otherwise harmless). Exit early
2684 if there is no more space in the output buffer.
2686 Severity : enhancement
2688 Description: clear OBD RPC statistics by writing to them
2689 Details : It is now possible to clear the OBD RPC statistics by writing
2690 to the "stats" file.
2695 Description: Client mtime is not the same on different clients after utimes
2696 Details : In some cases, the client was using the utimes() syscall on
2697 a file cached on another node. The clients now validate the
2698 ctime from the MDS + OSTs to determine which one is right.
2703 Description: Inability to activate failout mode
2704 Details : lconf script incorrectly assumed that in pythong string's numeric
2705 value is used in comparisons.
2708 Frequency : always with multiple stripes per file
2710 Description: Inefficient object allocation for mutli-stripe files
2711 Details : When selecting which OSTs to stripe files over, for files with
2712 a stripe count that divides evenly into the number of OSTs,
2713 the MDS is always picking the same starting OST for each file.
2714 Return the OST selection heuristic to the original design.
2719 Description: mount failures may take full timeout to return an error
2720 Details : Under some heavy load conditions it is possible that a
2721 failed mount can wait for the full obd_timeout interval,
2722 possibly several minutes, before reporting an error.
2723 Instead return an error as soon as the status is known.
2725 Frequency : quota enabled and large files being deleted
2727 Description: releasing more than 4GB of quota at once hangs OST
2728 Details : If a user deletes more than 4GB of files on a single OST it
2729 will cause the OST to spin in an infinite loop. Release
2730 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2735 Description: statfs data retrieved from /proc may be stale or zero
2736 Details : When reading per-device statfs data from /proc, in the
2737 {kbytes,files}_{total,free,avail} files, it may appear
2738 as zero or be out of date.
2741 Frequency : systems with MD RAID1 external journal devices
2743 Description: lconf's call to blkid is confused by RAID1 journal devices
2744 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2745 of returning all block devices that match the journal UUID.
2748 Frequency : always, for aggregate stripe size over 4GB
2750 Description: assertion fails when trying to use 4GB stripe size
2751 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2752 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2755 Frequency : always on ppc64
2757 Description: the first write on an ext3 filesystem with mballoc got stuck
2758 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2759 perform endianness conversion.
2761 ------------------------------------------------------------------------------
2763 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2765 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2766 INTEROPERATE with older versions automatically. Please read the
2767 user documentation before upgrading any part of a live system.
2768 * WARNING: Lustre networking configuration changes are required with
2769 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2772 * Support for kernels:
2773 2.6.9-22.0.2.EL (RHEL 4)
2774 2.6.5-7.244 (SLES 9)
2775 2.6.12.6 vanilla (kernel.org)
2778 Severity : enhancement
2779 Bugzilla : 7981/8208
2780 Description: Introduced Lustre Networking (LNET)
2781 Details : LNET is new networking infrastructure for Lustre, it includes
2782 a reorganized network configuration mode (see the user
2783 documentation for full details) as well as support for routing
2784 between different network fabrics. Lustre Networking Devices
2785 (LNDs) for the supported network fabrics have also been
2786 created for this new infrastructure.
2788 Severity : enhancement
2789 Description: Introduced Access control lists
2790 Details : clients can set ACLs on files and directories in order to have
2791 more fine-grained permissions than the standard Unix UGO+RWX.
2792 The MDS must be started with the "-o acl" mount option.
2794 Severity : enhancement
2795 Description: Introduced filesystem quotas
2796 Details : Administrators may now establish per-user quotas on the
2799 Severity : enhancement
2801 Description: Configuration change for the XT3
2802 The PTLLND is now used to run Lustre over Portals on the XT3
2803 The configure option(s) --with-cray-portals are no longer used.
2804 Rather --with-portals=<path-to-portals-includes> is used to
2805 enable building on the XT3. In addition to enable XT3 specific
2806 features the option --enable-cray-xt3 must be used.
2811 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2812 Details : A race between allocating a new llcd and re-getting the llcd_lock
2813 allowed another thread to grab newly-allocated llcd.
2815 Severity : enhancement
2817 Description: 2.6 OST async journal commit and locking fix to improve performance
2818 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2819 2.6 kernels are now async as they already were in 2.4 kernels so
2820 that they can commit concurrently with the network bulk transfer.
2821 For block-allocated files the filter allocation semaphore is held
2822 to avoid filesystem fragmentation during allocation. BKL lock
2823 removed for 2.6 xattr operations where it is no longer needed.
2828 Description: lconf incorrectly determined whether two IP networks could talk
2829 Details : In some more complicated routing and multiple-network
2830 configurations, lconf will avoid trying to make a network
2831 connection to a disjoint part of the IP space. It was doing the
2832 math incorrectly for one set of cases.
2837 Description: Fix for potential infinite loop processing records in an llog.
2838 Details : If an llog record is corrupted/zeroed, it is possible to loop
2839 forever in llog_process(). Validate the llog record length
2840 and skip the remainder of the block on error.
2843 Frequency : occasional (liblustre only)
2845 Description: liblustre could not open files whose last component is a symlink
2846 Details : sysio_path_walk() would incorrectly pass the open intent to
2847 intermediate path components.
2850 Frequency : rare (liblustre only with non-standard tuning)
2851 Bugzilla : 7201 (7350)
2852 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2853 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2854 after it has been released, only do this for referenced locks
2856 Severity : enhancement
2858 Description: specifying an (invalid) directory default stripe_size of -1
2859 would reset the directory default striping
2860 Details : stripe_size -1 was used internally to signal directory stripe
2861 removal, now use "all default" to signal dir stripe removal
2862 as a directory striping of "all default" is not useful
2865 Frequency : common for large clusters running liblustre clients
2867 Description: doing an ls when liblustre clients are running is slow
2868 Details : sending a glimpse AST to a liblustre client waits for every AST
2869 to time out, as liblustre clients will not respond. Since they
2870 cannot cache data we refresh the OST lock LVB from disk instead.
2872 Severity : enhancement
2874 Description: doing an ls at the same time as file IO can be slow
2875 Details : enqueue and other "small" requests can be blocked behind many
2876 large IO requests. Create a new OST IO portal for non-IO
2877 requests so they can be processed faster.
2880 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2882 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2883 created file with mode 000
2884 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2885 allow an ftruncate() as a non-root user to a file with mode 000.
2886 Lustre now respects this flag to disable mode checks when
2887 truncating a file owned by the user
2890 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2893 Description: Revoking locks from clients that went dead or catatonic might take
2895 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2896 cancellation of such locks instant on servers without waiting for
2897 any reply from clients. Clients drops these locks when cancel
2898 notification from server is received without replying.
2901 Frequency : liblustre-only, when liblustre client dies or becomes busy
2903 Description: Doing ls on Linux clients can take a long time with active
2905 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2906 granting such locks to it in the first place if possible. Locks
2907 are taken by proxy on the OST during the read or write and
2908 dropped immediately afterward. Add connect flags handling, do
2909 not grant locks to liblustre clients for glimpse ASTs.
2911 Severity : enhancement
2913 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2914 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2915 sequential reads and behaves badly with applications doing random
2916 reads. Improve it by reading ahead at least read region, and
2917 avoiding excessive large RPC for small reads.
2919 Severity : enhancement
2921 Description: Creating more than 1000 files for a single job may cause a load
2922 imbalance on the OSTs if there are also a large number of OSTs.
2923 Details : qos_prep_create() uses an OST index reseed value that is an
2924 even multiple of the number of available OSTs so that if the
2925 reseed happens in the middle of the object allocation it will
2926 still utilize the OSTs as uniformly as possible.
2931 Description: OST or MDS may oops in ping_evictor_main()
2932 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2933 but doesn't restart at beginning of obd_exports_timed list
2936 Severity : enhancement
2938 Description: improve by-nid export eviction on the MDS and OST
2939 Details : allow multiple exports with the same NID to be evicted at one
2940 time without re-searching the exports list.
2943 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2945 Description: MDS may oops in groups_free()
2946 Details : in rare race conditions a newly allocated group_info struct is
2947 freed again, and this can be NULL. The 2.4 compatibility code
2948 for groups_free() checked for a NULL pointer, but 2.6 did not.
2951 Frequency : common for liblustre clients doing little filesystem IO
2952 Bugzilla : 9352, 7313
2953 Description: server may evict liblustre clients accessing contended locks
2954 Details : if a client is granted a lock or receives a completion AST
2955 with a blocking AST already set it would not reply to the AST
2956 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2959 Frequency : lfs setstripe, only systems with more than 160 OSTs
2961 Description: unable to set striping with a starting offset beyond OST 160
2962 Details : llapi_create_file() incorrectly limited the starting stripe
2963 index to the maximum single-file stripe count.
2966 Frequency : LDAP users only
2968 Description: lconf did not handle in-kernel recovery with LDAP properly
2969 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2971 Severity : enhancement
2973 Description: bind OST threads to NUMA nodes to improve performance
2974 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2975 node and do their allocations there to localize memory access
2977 Severity : enhancement
2979 Description: llmount can determine client NID directly from Myrinet (GM)
2980 Details : the client NID code from gmnalnid was moved directly into
2981 llmount, removing the need to use this or specifying the
2982 client NID explicitly when mounting GM clients with zeroconf
2985 Frequency : if client is started with down MDS
2987 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
2988 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
2989 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
2990 ptlrpc_import_delayed_req() because we didn't check if the
2991 request was interrupted, and we also didn't break out of the
2992 event loop if there was no timeout
2997 Description: data loss during non-page-aligned writes to a single file from
2998 both multiple nodes and multiple threads on one node at same time
2999 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3000 inconsistency led to false short-reads, that were cached and later
3001 used by ->prepare_write() to fill in partially written page,
3002 leading to data loss.
3005 Frequency : always, if lconf --abort_recovery used
3007 Description: lconf --abort_recovery fails with 'Operation not supported'
3008 Details : lconf was attempting to abort recovery on the MDT device and not
3011 Severity : enhancement
3013 Description: remove cleanup logs
3014 Details : replace lconf-generated cleanup logs with lustre internal
3015 cleanup routines. Eliminates the need for client-cleanup and
3018 Severity : enhancement
3020 Description: add support for EAs (user and system) on lustre filesystems
3021 Details : it is now possible to store extended attributes in the Lustre
3022 client filesystem, and with the user_xattr mount option it
3023 is possible to allow users to store EAs on their files also
3025 Severity : enhancement
3027 Description: Add possibility (config option) to show minimal available OST free
3029 Details : When compiled with --enable-mindf configure option, statfs(2)
3030 (and so, df) will return least minimal free space available from
3031 all OSTs as amount of free space on FS, instead of summary of
3032 free spaces of all OSTs.
3034 Severity : enhancement
3036 Description: do not expand extent locks acquired on OST-side
3037 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3038 by server: they are not cached anyway.
3041 Frequency : when mmap is used/binaries executed from Lustre
3043 Description: Unmmap pages before throwing them away from read cache.
3044 Details : llap_shrink cache now attempts to unmap pages before discarding
3045 them (if unmapping failed - do not discard). SLES9 kernel has
3046 extra checks that trigger if this unmapping is not done first.
3051 Description: lconf didn't resolve symlinks before checking to see whether a
3052 given mountpoint was already in use
3055 Frequency : when migrating failover services
3056 Bugzilla : 6395, 9514
3057 Description: When migrating a subset of services from a node (e.g. failback
3058 from a failover service node) the remaining services would
3059 time out and evict clients.
3060 Details : lconf --force (implied by --failover) sets the global obd_timeout
3061 to 5 seconds in order to quickly disconnect, but this caused
3062 other RPCs to time out too quickly. Do not change the global
3063 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3065 Severity : enhancement
3066 Frequency : if MDS is started with down OST
3067 Bugzilla : 9439,5706
3068 Description: Allow startup/shutdown of an MDS without depending on the
3069 availability of the OSTs.
3070 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3071 Add appropriate locking and lov-osc refcounts for safe
3072 cleaning. Add osc abort_inflight calls in case the
3073 synchronize never started.
3076 Frequency : occasional (Cray XT3 only)
3078 Description: root not authorized to access files in CRAY_PORTALS environment
3079 Details : The client process capabilities were not honoured on the MDS in
3080 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3081 been accessed by an authorized user then root was able to access
3082 the file on the local client also. The root user capabilities
3083 are now allowed on the MDS, as this environment has secure UID.
3086 Frequency : occasional
3088 Description: ldiskfs "too long searching" message happens too often
3089 Details : A debugging message (otherwise harmless) prints too often on
3090 the OST console. This has been reduced to only happen when
3091 there are fragmentation problems on the filesystem.
3096 Description: Division by zero in statfs when all OSCs are inactive
3097 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3098 lov_statfs divides by value returned by lov_get_stripecnt().
3102 Bugzilla : 9489, 3273
3103 Description: First write from each client to each OST was only 4kB in size,
3104 to initialize client writeback cache, which caused sub-optimal
3105 RPCs and poor layout on disk for the first writen file.
3106 Details : Clients now request an initial cache grant at (re)connect time
3107 and so that they can start streaming writes to the cache right
3108 away and always do full-sized RPCs if there is enough data.
3109 If the OST is rebooted the client also re-establishes its grant
3110 so that client cached writes will be honoured under the grant.
3115 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3116 Details : Now I/O RPCs go to different portal number and (presumably) fast
3117 lock requests (and glimses) and other RPCs get their own service
3118 threads pool that should be able to service those RPCs
3121 Severity : enhancement
3123 Description: Ability to exchange lustre version between client and servers and
3124 issue warnings at client side if client is too old. Also for
3125 liblustre clients there is ability to refuse connection of too old
3127 Details : New 'version' field is added to connect data structure that is
3128 filled with version info. That info is later checked by server and
3132 Frequency : rare, liblustre only.
3133 Bugzilla : 9296, 9581
3134 Description: Two simultaneous writes from liblustre at offset within same page
3135 might proceed at the same time overwriting eachother with stale
3137 Details : I/O lock withing llu_file_prwv was released too early, before data
3138 actually was hitting the wire. Extended lock-holding time until
3139 server acknowledges receiving data.
3142 Frequency : extremely rare. Never observed in practice.
3144 Description: avoid generating lustre_handle cookie of 0.
3145 Details : class_handle_hash() generates handle cookies by incrementing
3146 global counter, and can hit 0 occasionaly (this is unlikely, but
3147 not impossible, because initial value of cookie counter is
3148 selected randonly). Value of 0 is used as a sentinel meaning
3149 "unassigned handle" --- avoid it. Also coalesce two critical
3150 sections in this function into one.
3152 Severity : enhancement
3154 Description: allow liblustre clients to delegate truncate locking to OST
3155 Details : To avoid overhead of locking, liblustre client instructs OST to
3156 take extent lock in ost_punch() on client's behalf. New connection
3157 flag is added to handle backward compatibility.
3159 Severity : enhancement
3160 Bugzilla : 4928, 7341, 9758
3161 Description: allow number of OST service threads to be specified
3162 Details : a module parameter allows the number of OST service threads
3163 to be specified via "options ost ost_num_threads={N}" in the
3164 OSS's /etc/modules.conf or /etc/modprobe.conf.
3168 Bugzilla : 6146, 9635, 9895
3169 Description: servers crash with bad pointer in target_handle_connect()
3170 Details : In rare cases when a client is reconnecting it was possible that
3171 the connection request was the last reference for that export.
3172 We would temporarily drop the export reference and get a new
3173 one, but this may have been the last reference and the export
3174 was just destroyed. Get new reference before dropping old one.
3176 Severity : enhancement
3177 Frequency : if client is started with failover MDS
3179 Description: Allow multiple MDS hostnames in the mount command
3180 Details : Try to read the configuration from all specified MDS
3181 hostnames during a client mount in case the "primary"
3184 Severity : enhancement
3186 Description: Stop sending data to evicted clients as soon as possible.
3187 Details : Check if the client we are about to send or are sending data to
3188 was evicted already. (Check is done every second of waiting,
3189 for which l_wait_event interface was extended to allow checking
3190 of exit condition at specified intervals).
3193 Frequency : rare, normally only when NFS exporting is done from client
3195 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3197 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3200 Frequency : always, if srand() is called before liblustre initialization
3202 Description: Liblustre uses system PRNG disturbing its usage by user application
3203 Details : Introduce internal to lustre fast and high-quality PRNG for
3204 lustre usage and make liblustre and some other places in generic
3205 lustre code to use it.
3207 Severity : enhancement
3208 Bugzilla : 9477, 9557, 9870
3209 Description: Verify that the MDS configuration logs are updated when xml is
3210 Details : Check if the .xml configuration logs are newer than the config
3211 logs stored on the MDS and report an error if this is the case.
3212 Request --write-conf, or allow starting with --old_conf.
3214 Severity : enhancement
3216 Description: Handle symlinks in the path when checking if Lustre is mounted.
3217 Details : Resolve intermediate symlinks when checking if a client has
3218 mounted a filesystem to avoid duplicate client mounts.
3223 Description: lconf can hit an error exception but still return success.
3224 Details : The lconf command catches the Command error exception at the top
3225 level script context and will exit with the associated exit
3226 status, but doesn't ensure that this exit status is non-zero.
3231 Description: failure of ptlrpc thread startup can cause oops
3232 Details : Starting a ptlrpc service thread can fail if there are a large
3233 number of threads or the server memory is very fragmented.
3234 Handle this without oopsing.
3237 Frequency : always, only if liblustre and non-default acceptor port was used
3239 Description: liblustre cannot connect to servers with non-default acceptor port
3240 Details : tcpnal_set_default_params() was not called and was therefore
3241 ignoring the environment varaible TCPNAL_PORT, as well as other
3242 TCPNAL_ environment variables
3247 Description: two objects could be created on the same OST for a single file
3248 Details : If an OST is down, in some cases it was possible to create two
3249 objects on a single OST for a single file. No problems other
3250 than potential performance impact and spurious error messages.
3254 Bugzilla : 5681, 9562
3255 Description: Client may oops in ll_unhash_aliases
3256 Details : Client dcache may become inconsistent in race condition.
3257 In some cases "getcwd" can fail if the current directory is
3263 Description: Inode refcounting problems in NFS export code
3264 Details : link_raw functions used to call d_instantiate without obtaining
3265 extra inode reference first.
3269 Bugzilla : 9942, 9903
3270 Description: Referencing freed requests leading to crash, memleaks with NFS.
3271 Details : We used to require that call to ll_revalidate_it was always
3272 followed by ll_lookup_it. Also with revalidate_special() it is
3273 possible to call ll_revalidate_it() twice for the same dentry
3274 even if first occurence returned success. This fix changes semantic
3275 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3276 reference on a request referred from the intent.
3277 ll_intent_release() then releases such a request.
3280 Frequency : rare, normally benchmark loads only
3282 Description: unlinked inodes were kept in memory on the client
3283 Details : If a client is repeatedly creating and unlinking files it
3284 can accumulate a lot of stale inodes in the inode slab cache.
3285 If there is no other client load running this can cause the
3286 client node to run out of memory. Instead flush old inodes
3287 from client cache that have the same inode number as a new inode.
3290 Frequency : SLES9 2.6.5 kernel and long filenames only
3291 Bugzilla : 9969, 10379
3292 Description: utime reports stale NFS file handle
3293 Details : SLES9 uses out-of-dentry names in some cases, which confused
3294 the lustre dentry revalidation. Change it to always use the
3298 Frequency : rare, unless heavy write-truncate concurrency is continuous
3299 Bugzilla : 4180, 6984, 7171, 9963, 9331
3300 Description: OST becomes very slow and/or deadlocked during object unlink
3301 Details : filter_destroy() was holding onto the parent directory lock
3302 while truncating+unlinking objects. For very large objects this
3303 may block other threads for a long time and slow overall OST
3304 responsiveness. It may also be possible to get a lock ordering
3305 deadlock in this case, or run out of journal credits because of
3306 the combined truncate+unlink. Solution is to do object truncate
3307 first in one transaction without parent lock, and then do the
3308 final unlink in a new transaction with the parent lock. This
3309 reduces the lock hold time dramatically.
3312 Frequency : rare, 2.4 kernels only
3314 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3315 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3316 before unmounting their filesystem in order to avoid deadlock.
3317 The kernel_locked() function in 2.4 kernels only checks whether
3318 the kernel lock is held, not whether it is this process that is
3319 holding it as 2.6 kernels do.
3324 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3325 Details : The client ptlrpc code may be trying to reconnect to a down
3326 server before a previous connection attempt has timed out.
3327 Increase the reconnect interval to be longer than the connection
3328 timeout interval to avoid sending duplicate connections to
3332 Frequency : echo_client brw_test command
3334 Description: fix echo_client to work with OST preallocated code
3335 Details : OST preallocation code (5137) didn't take echo_client IO path
3336 into account: echo_client calls filter methods outside of any
3337 OST thread and, hence, there is no per-thread preallocated
3338 pages and buffers to use. Solution: hijack pga pages for IO. As
3339 a byproduct, this avoids unnecessary data copying.
3343 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3344 Description: Client can oops in mdc_commit_close() after open replay
3345 Details : It was possible for the MDS to return an open request with no
3346 transaction number in mds_finish_transno() if the client was
3347 evicted, but without actually returning an error. Clients
3348 would later try to replay that open and may trip an assertion
3349 Simplify the client close codepath, and always return an error
3350 from the MDS in case the open is not successful.
3353 Frequency : rare, 2.6 OSTs only
3355 Description: OST may deadlock under high load on fragmented files
3356 Details : If there was a heavy load and highly-fragmented OST filesystems
3357 it was possible to have all the OST threads deadlock waiting on
3358 allocation of biovecs, because the biovecs were not released
3359 until the entire RPC IO was completed. Instead, release biovecs
3360 as soon as they are complete to ensure forward IO progress.
3362 Severity : enhancement
3364 Description: Support for specifying external journal device at mount
3365 Details : If an OST or MDS device is formatted with an external journal
3366 device, this device major/minor is stored in the ext3 superblock
3367 and may not be valid for failover. Allow detecting and
3368 specifying the external journal at mount time.
3373 Description: Mounting an MDS with pending unlinked files may cause oops
3374 Details : target_finish_recovery() calls mds_postrecov() which returned
3375 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3376 considers this an error and immediately begins cleaning up the
3377 lov, just after starting the mds_lov process
3379 Severity : enhancement
3381 Description: Implement 'lfs df' to report actual free space on per-OST basis
3382 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3383 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3384 usage of objects; '-h' to report in human readable format.
3386 ------------------------------------------------------------------------------
3388 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3395 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3396 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3397 at startup during orphan unlinking. Instead, skip buddy update
3398 before setup, as it will be regenerated after recovery is complete.
3403 Description: If an OST is inactive, its locks might reference stale inodes.
3404 Details : lov_change_cbdata() must iterate over all namespaces, even if
3405 they are inactive to clear inode references from the lock.
3407 Severity : enhancement
3408 Frequency : occasional, if non-standard max_dirty_mb used
3410 Description: Client will block write RPCs if not enough grant
3411 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3412 then the client will block writes while waiting for another RPC
3413 to complete instead of consuming its dirty limit. With change
3414 we get improved performance when max_dirty_mb is small.
3416 Severity : enhancement
3417 Bugzilla : 3389, 6253
3418 Description: Add support for supplementary groups on the MDS.
3419 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3420 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3421 lookups for user supplementary groups into a cache.
3425 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3426 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3427 for liblustre O_EXCL is enforced by MDS.
3431 Description: readdir never returns NULL in liblustre.
3432 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3433 dentry in struct dirent.
3437 Frequency : liblustre only, depends on application IO pattern
3438 Description: liblustre clients evicted if not contacting servers
3439 Details : Don't put liblustre clients into the ping_evictor list, so
3440 they will not be evicted by the pinger ever.
3442 Severity : enhancement
3444 Description: Add ability to evict clients by NID from MDS.
3445 Details : By echoing "nid:$NID" string into
3446 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3447 $NID would be instantly evicted from this MDS and from all active
3448 OSTs connected to it.
3452 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3453 Details : lookup_it_finish() used to query file size from OSTs that was not
3458 Description: service threads change working directory to that of init
3459 Details : Starting lustre service threads may pin the working directory
3460 of the parent thread, making that filesystem busy. Threads
3461 now change to the working directory of init to avoid this.
3465 Frequency : during shutdown only
3466 Description: shutdown with a failed MDS or OST can cause unmount to hang
3467 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3472 Frequency : 2.6 only
3473 Description: chmod/chown may include an extra supplementary group
3474 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3475 supplementary group and if none is specified this is used.
3478 Bugzilla : 5479 (6816)
3480 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3481 Details : If lookup is in progress on a file that is unlinked we might try
3482 to revalidate the inode and fail in revalidate after lookup is
3483 complete and ll_file_open() enqueues the open again but
3484 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3487 Frequency : always, if lconf --abort_recovery used
3489 Description: lconf --abort_recovery fails with 'Operation not supported'
3490 Details : lconf was attempting to abort recovery on the MDT device and not
3493 ------------------------------------------------------------------------------
3495 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3500 Frequency : rare (only unsupported configurations with a node running as an
3502 Bugzilla : 6514, 5137
3503 Description: Mounting a Lustre file system on a node running as an OST could
3505 Details : OSTs now preallocates memory needed to write out data at
3506 startup, instead of when needed, to avoid having to
3507 allocate memory in possibly low memory situations.
3508 Specifically, if the file system is mounted on on OST,
3509 memory pressure could force it to try to write out data,
3510 which it needed to allocate memory to do. Due to the low
3511 memory, it would be unable to do so and the node would
3512 become unresponsive.
3514 Severity : enhancement
3516 Description: Addition of lconf --service command line option
3517 Details : lconf now accepts a '--service <arg>' option, which is
3518 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3520 Severity : enhancement
3522 Description: Failover mode is now the default for OSTs.
3523 Details : By default, OSTs will now run in failover mode. To return to
3524 the old behaviour, add '--failout' to the lmc line for OSTs.
3526 Severity : enhancement
3528 Description: Health checks are now provided for MDS and OSTs
3529 Details : Additional detailed health check information on MSD and OSTs
3530 is now provided through the procfs health_check value.
3533 Frequency : occasional, depends on IO load
3535 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3536 after numerous create/delete cycles
3537 Details : The ext3 inode allocation policy would not allocate new inodes
3538 very well on the OSTs because there are no new directories
3539 being created. Instead we look for groups with free space if
3540 the parent directories are nearly full.
3545 Description: Network or server problems during mount may cause partially
3546 mounted clients instead of returning an error.
3547 Details : The config llog parsing code may overwrite the error return
3548 code during mount error handling, returning success instead
3554 Description: MDS can fail to allocate large reply buffers
3555 Details : After long uptimes the MDS can fail to allocate large reply
3556 buffers (e.g. zconf client mount config records) due to memory
3557 fragmentation or consumption by the buffer cache. Preallocate
3558 some large reply buffers so that these replies can be sent even
3559 under memory pressure.
3563 Frequency : rare (liblustre)
3564 Description: fsx running with liblustre complained that using truncate() to
3565 extend the file doesn't work. This patch corrects that issue.
3566 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3567 ATTR_SIZE and lsm use in llu_setattr_raw.
3571 Frequency : rare, only 2.6 kernels
3572 Description: Unusual file access patterns on the MDS may result in inode
3573 data being lost in very rare circumstances.
3574 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3575 code in the 2.6 kernel caused the inode and/or EA data not to
3576 be read from disk, causing single-file corruption.
3580 Frequency : rare, only 2.6 filesystems using extents
3581 Description: Heavy concurrent write and delete load may cause data corruption.
3582 Details : It was possible under high-load situations to have an extent
3583 metadata block in the block device cache from a just-unlinked
3584 file overwrite a newly-allocated data block. We now unmap any
3585 metadata buffers that alias just-allocated data blocks.
3589 Frequency : filesystems with default stripe_count larger than 77
3590 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3591 Details : lconf specifies an inode size of 4096 bytes when the default
3592 stripe_count is larger than 77. This conflicts with the default
3593 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3594 this case to avoid pinning too much memory for large EAs.
3596 ------------------------------------------------------------------------------
3598 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3603 Frequency : rare (extremely heavy IO load with hundreds of clients)
3605 Description: Client is evicted, gets IO error writing to file
3606 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3607 caused clients to be evicted for AST timeouts. The fixes in
3608 bug 5192 mean we no longer need to have such short AST timeouts
3609 so ldlm_timeout has been increased.
3612 Frequency : occasional during --force or --failover shutdown under load
3613 Bugzilla : 5949, 4834
3614 Description: Server oops/LBUG if stopped with --force or --failover under load
3615 Details : a collection of import/export refcount and cleanup ordering
3616 issues fixed for safer force cleanup
3619 Frequency : only filesystems larger than 120 OSTs
3620 Bugzilla : 5990, 6223
3621 Description: lfs getstripe would oops on a very large filesystem
3622 Details : lov_getconfig used kfree on vmalloc'd memory
3625 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3626 Bugzilla : 6242, 6243
3627 Description: reading from files that had been truncated to a non-zero size
3628 but never opened returned no data
3629 Details : ll_file_read() reads zeros from no-object files to EOF
3634 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3635 Details : There's little harm in aborting MDS/OSS recovery and letting it
3636 try again, so I removed the LASSERT and return an error instead.
3638 Severity : enhancement
3640 Description: New debugging infrastructure for tracking down data corruption
3641 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3642 (b) cover more of the client-side code path, and (c) try to narrow
3643 down where problems occurred
3647 Bugzilla : 3819, 4364, 4397, 6313
3648 Description: Racing close and eviction MDS could cause assertion in mds_close
3649 Details : It was possible to get multiple mfd references during close and
3650 client eviction, leading to one thread referencing a freed mfd.
3652 Severity: : enhancement
3653 Bugzilla : 3262, 6359
3654 Description: Attempts to reconnect to servers are now more aggressive.
3655 Details : This builds on the enhanced upcall-less recovery that was added
3656 in 1.4.2. When trying to reconnect to servers, clients will
3657 now try each server in the failover group every 10 seconds. By
3658 default, clients would previously try one server every 25 seconds.
3663 Description: After recovery, certain operations trigger a failed
3664 assertion on a client.
3665 Details : Failing over an mds, using lconf -d --failover, while a
3666 client was doing a readdir() call would cause the client to
3667 LBUG after recovery completed and the readdir() was resent.
3669 Severity : enhancement
3671 Description: Default groups are now added by lconf
3672 Details : You can now run lconf --group <servicename> without having to
3673 manually add groups with lmc.
3676 Frequency : occasional
3678 Description: Nodes with an elan id of 0 trigger a failed assertion
3681 Frequency : always when accessing e.g. tty/console device nodes
3683 Description: tty and some other devices nodes cannot be used on lustre
3684 Details : file's private_data field is used by device data and lustre
3685 values in there got lost. New field was added to struct file to
3686 store fs-specific private data.
3689 Frequency : when exporting Lustre via NFS
3691 Description: NFSD failed occasionally when looking up a path component
3692 Details : NFSD is looking up ".." which was broken in ext3 directories
3693 that had grown large enough to become hashed.
3696 Frequency : Clusters with multiple interfaces not on the same subnet
3698 Description: Nodes will repeatedly try to reconnect to an interface which it
3699 cannot reach and report an error to the log.
3700 Details : Extra peer list entries will be created by lconf with some peers
3701 unreachable. lconf now validates the peer before adding it.
3704 Frequency : Only if a default stripe is set on the filesystem root.
3706 Description: Setting a default stripe on the filesystem root prevented the
3707 filesystem from being remounted.
3708 Details : The client was sending extra request flags in the root getattr
3709 request and did not allocate a reply buffer for the dir EA.
3712 Frequency : occasional, higher if lots of files are accessed by one client
3713 Bugzilla : 6159, 6097
3714 Description: Client trips assertion regarding lsm mismatch/magic
3715 Details : While revalidating inodes the VFS looks up inodes with ifind()
3716 and in rare cases can find an inode that is being freed.
3717 The ll_test_inode() code will free the lsm during ifind()
3718 when it finds an existing inode and then the VFS later attaches
3719 this free lsm to a new inode.
3723 Bugzilla : 6422, 7030
3724 Description: MDS deadlock between mkdir and client eviction
3725 Details : Creating a new file via mkdir or mknod (starting a transaction
3726 and getting the ns lock) can deadlock with client eviction
3727 (gets ns lock and trying to finish a synchronous transaction).
3730 Frequency : occasional
3731 Description: While starting a server, the fsfilt_ext3 module could not be
3733 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3734 kernels. Previously, lconf would still use the ext3 name
3735 when trying to load modules. Now, it will correctly use
3736 ext3 on 2.4 and ldiskfs on 2.6.
3738 Severity : enhancement
3739 Description: The default stripe count has been changed to 1
3740 Details : The interpretation of the default stripe count (0, to lfs
3741 or lmc) has been changed to mean striping across a single
3742 OST, rather than all available. For general usage we have
3743 found a stripe count of 1 or 2 works best.
3745 Severity : enhancement
3746 Description: Add support for compiling against Cray portals.
3747 Details : Conditional compiling for some areas that are different
3751 Frequency : occasional
3752 Bugzilla : 6409, 6834
3753 Description: Creating files with an explicit stripe count may lead to
3754 a failed assertion on the MDS
3755 Details : If some OSTs are full or unavailable, creating files may
3756 trigger a failed assertion on the MDS. Now, Lustre will
3757 try to use other servers or return an error to the
3761 Frequency : occasional
3763 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3764 caused lock timeout during readahead (since 1.4.2).
3765 Details : Processes doing readahead might match a lock that hasn't been
3766 granted yet if there are overlapping and conflicting lock
3767 requests. The readahead process waits on ungranted lock
3768 (original lock is CBPENDING), while OST waits for that process
3769 to cancel CBPENDING read lock and eventually evicts client.
3771 Severity : enhancement
3773 Description: Initial enabling of flock support for clients
3774 Details : Implements fcntl advisory locking and file status functions.
3775 This feature is provided as an optional mount flag (default
3776 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3777 locking are implemented yet, and those that are are not guaranteed
3778 to be completely correct in production environments.
3779 mount -t lustre -o [flock|noflock] ...
3782 Frequency : occasional
3784 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3785 assertion in the ext3 JBD (journaling) layer.
3786 Details : The b_committed_data struct is protected by the big kernel lock
3787 in 2.4 kernels, serializing journal_commit_transaction() and
3788 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3789 access to this struct. In 2.6 kernels there is finer grained
3790 locking to improve SMP performance of the JBD layer.
3794 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3799 Description: merge and cleanup kernel patches.
3802 -----------------------------------------------------------------------------
3804 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3806 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3807 versions of Lustre. Please update all servers and clients to
3808 version 1.4.2 or later at the same time. You must also run
3809 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3810 to update the configuration logs.
3812 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3813 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3814 - swab llog records of type '0' so we get proper header size/idx (5861)
3815 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3816 - fix rename of one directory over another leaking an inode (5953)
3817 - avoid SetPageDirty on 2.6 (5981)
3818 - don't re-add just-being-destroyed locks to the waiting list (5653)
3819 - when creating new directories, inherit the parent's custom
3820 striping settings if present parent (3048)
3821 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3822 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3823 - don't hold client locks on temporary worklist from l_lru (5666)
3824 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3825 - automatically evict dead clients (5921)
3826 - Update file size properly in create+truncate+fstat case (6196)
3827 - Do not unhash mountpoint dentries, do not allow removal of
3829 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3830 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3831 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3832 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3833 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3834 - Correctly handle reads of files with no objects (6243)
3835 - lctl recover will also mark a device active if deactivate used (5933)
3837 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3838 - allow --write-conf on an MDS with different nettype than client (5619)
3839 - don't write config llogs to MDS for mounts not from that MDS (5617)
3840 - lconf should create multiple TCP connections from a client (5201)
3841 - init scripts are now turned off by default; run chkconfig --on
3842 lustre and chkconfig --on lustrefs to use them
3843 - upcalls are no longer needed for clients to recover to failover
3845 - add --abort-recovery option to lconf to abort recovery on device
3847 - add support for an arbitrary number of OSTs (3026)
3848 - Quota support protocol changes.
3849 - forward compatibility changes to wire structs (6007)
3850 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3851 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3852 - improved "lustre status" script
3853 - initialize blocksize for non-regular files (6062)
3854 - added --disable-server and --disable-client configure options (5782)
3855 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3856 - Vanilla 2.4.29 support
3857 - increase maximum number of obd devices to 520 (6242)
3858 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3859 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3860 SLES 9 2.6.5 kernels
3861 - limit stripes per file to 160 (the maximum EA size) (6093)
3863 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3866 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3867 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3868 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3869 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3870 - small ext3 extents cleanups and fixes (5733)
3871 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3872 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3873 - avoid needless client->OST connect, fix handle mismatch (5317)
3874 - fix DLM error path that led to out-of-sync client, long delays (5779)
3875 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3876 - fix several locking issues related to i_size (5492,5624,5654,5672)
3877 - don't move pending lock onto export if it is already evicted (5683)
3878 - fix kernel oops when creating .foo in unlinked directory (5548)
3879 - fix deadlock in obdfilter statistics vs. object create (5811)
3880 - use time_{before,after} to avoid timer jiffies wrap (5882)
3881 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3882 - Do not leak request if server was not able to process it (5154)
3883 - If mds_open unable to find parent dir, make that negative lookup(5154)
3884 - don't create new directories with extent-mapping (5909, 5936)
3886 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3887 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3888 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3889 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3890 - support for RHEL4 kernel on i686 (5773)
3891 - provide error messages when incompatible logs are encountered (5898)
3893 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3894 * version 1.4.0.10 (1.4.1 release candidate 1)
3896 - don't keep a lock reference when lock is not granted (4238)
3897 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3898 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3899 - drop import inflight refcount on signal_completed_replay error (5255)
3900 - unlock page after async write error during send (3677)
3901 - handle missing objects in filter_preprw_read properly (5265)
3902 - no transno return for symlink open, don't save no-trasno open (3440)
3903 - don't try to complete elan receive that already failed (4012)
3904 - free RPC server reply state on error (5406)
3905 - clean up thread from ptlrpc_start_thread() on error (5160)
3906 - readahead could read extra page into cache that wasn't ejected (5388)
3907 - prevent races in class_attach/setup/cleanup/detach (5260)
3908 - don't dereference de->d_inode after l_dput of de (5458)
3909 - use "int" for stripe value returned from lock_to_stripe (5544)
3910 - mballoc allocation and error-checking fixes in 2.6 (5504)
3911 - block device patches to fix I/O request sizes in 2.6 (5482)
3912 - look up hostnames for IB nals (5602)
3913 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3914 - don't start multiple acceptors for the same port (5277)
3915 - fix incorrect LASSERT in mds_getattr_name (5635)
3916 - export a proc file for general "ping" checking (5628)
3917 - fix "lfs check" to not block when the MDS is down (5628)
3919 - service request history (4965)
3920 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3921 - create an "evict_client" /proc entry on OSTs, like the MDS has
3922 - fix mount usage message, return errors per mount(8) (5168)
3923 - change grep [] to grep "[]" in tests so they work in more UMLs
3924 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3925 - remove ancient LOV_MAGIC_V0 EA support (5047)
3926 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3927 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3928 - allow readahead allocations to fail when low on memory (5383)
3929 - mmap locking landed again, after considerable improvement (2828)
3930 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3932 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3935 - send OST transaction number in read/write reply to free req (4966)
3936 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3937 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3938 - the watchdog thread now runs as interruptible (5246)
3939 - flock/lockf fixes (but it's still disabled, pending 5135)
3940 - don't use EXT3 constants in llite code (5094)
3941 - memory shortage at startup could cause assertion (5176)
3943 - reorganization of lov code
3944 - single portals codebase
3946 - add extents/mballoc support (5025)
3947 - direct I/O reads in the obdfilter (4048)
3948 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3950 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3953 - send OST transaction number in read/write reply to free req (4966)
3954 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3955 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3956 - the watchdog thread now runs as interruptible (5246)
3957 - handle missing objects in filter_preprw_read properly (5265)
3958 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3959 - drop import inflight refcount on signal_completed_replay error (5255)
3960 - unlock page after async write error during send (3677)
3961 - return original error code on reconstructed replies (3761)
3962 - no transno return for symlink open, don't save no-trasno open (3440)
3964 - add pid to ldlm debugging output (4922)
3965 - bump the watchdog timeouts -- we can't handle 30sec yet
3966 - extra debugging for orphan dentry/inode bug (5259)
3968 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3971 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3972 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3973 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3974 - fix NULL dereference in /proc/sys/portals/routes (4827)
3975 - allow failed mdc_close() operations to be interrupted (4561)
3976 - stop precreate on OST before MDS would time out on it (4778)
3977 - don't send partial-page writes before EOF from client (4410)
3978 - discard client grant for sub-page writes on large-page clients (4520)
3979 - don't free dentries not owned by NFS code, check generation (4806)
3980 - fix lsm leak if mds_create_objects() fails (4801)
3981 - limit debug_daemon file size, always print CERROR messages (4789)
3982 - use transno after validating reply (3892)
3983 - process timed out requests if import state changes (3754)
3984 - update mtime on OST during writes, return in glimpse (4829)
3985 - add mkfsoptions to LDAP (4679)
3986 - use ->max_readahead method instead of zapping global ra (5039)
3987 - don't interrupt __l_wait_event() during strace
3989 - add software watchdogs to catch hung threads quickly (4941)
3990 - make lustrefs init script start after nfs is mounted
3991 - fix CWARN/ERROR duplication (4930)
3992 - return async write errors to application if possible (2248)
3993 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
3994 - print NAL number in %x format (4645)
3995 - update barely-supported suse-2.4.21-171 series (4842)
3996 - support for sles 9 %post scripts
3997 - support for building 2.6 kernel-source packages
3998 - support for sles km_* packages
4000 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4003 - ignore -ENOENT errors in osc_destroy (3639)
4004 - notify osc create thread that OSC is being cleaned up (4600)
4005 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4006 - reconstruct ost_handle() like mds_handle() (4657)
4007 - create a new thread to do import eviction to avoid deadlock (3969)
4008 - let lconf resolve symlinked-to devices (4629)
4009 - don't unlink "objects" from directory with default EA (4554)
4010 - hold socknal file ref over connect in case target is down (4394)
4011 - allow more than 32000 subdirectories in a single directory (3244)
4012 - fix blocks count for O_DIRECT writes (3751)
4013 - OST returns ENOSPC from object create when no space left (4539)
4014 - don't send truncate RPC if file size isn't changing (4410)
4015 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4016 - bind to privileged port in socknal and tcpnal (3689)
4018 - rate limit CERROR/CWARN console message to avoid overload (4519)
4019 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4020 - basic mmap support (3918)
4021 - kernel patch series update from b1_4 (4711)
4023 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4026 - avoid crash during MDS cleanup with OST shut down (2775)
4027 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4028 - don't use bad inodes on the MDS (3744)
4029 - dynamic object preallocation to improve recovery speed (4236)
4030 - don't hold spinlock over lock dumping or change debug flags (4401)
4031 - don't zero obd_dev when it is force cleaned (3651)
4032 - print grants to console if they go negative (4431)
4033 - "lctl deactivate" will stop automatic recovery attempts (3406)
4034 - look for existing locks in ldlm_handle_enqueue() (3764)
4035 - don't resolve lock handle twice in recovery avoiding race (4401)
4036 - revalidate should check working dir is a directory (4134)
4038 - don't always mark "slow" obdfilter messages as errors (4418)
4040 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4043 - don't close LustreDB during write_conf until it is done (3860)
4044 - fix typo in lconf for_each_profile (3821)
4045 - allow dumping logs from multiple threads at one time (3820)
4046 - don't allow multiple threads in OSC recovery (3812)
4047 - fix debug_size parameters (3864)
4048 - fix mds_postrecov to initialize import for llog ctxt (3121)
4049 - replace config semaphore with spinlock (3306)
4050 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4051 - don't allow enqueue to complete on a destroyed export (3822)
4052 - down write_lock before checking llog header bitmap (3825)
4053 - recover from lock replay timeout (3764)
4054 - up llog sem before sending rpc (3652)
4055 - reduce ns lock hold times when setting kms (3267)
4056 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4057 - fix NULL deref and obd_dev leak on setup error (3312)
4058 - replace some LBUG about llog ops with error handling (3841)
4059 - don't match INVALID dentries from d_lookup and spin (3784)
4060 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4061 - fix invalid assertion in ptlrpc_set_wait (3880)
4063 - add libwrap support for the TCP acceptor (3996)
4064 - add /proc/sys/portals/routes for non-root route listing (3994)
4065 - allow setting MDS UUID in .xml (2580)
4066 - print the stack of a process that LBUGs (4228)
4068 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4071 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4072 - make sure to unset replay flag from failed open requests (3440)
4073 - if default stripe count is 0, use OST count for inode size (3636)
4074 - update parent mtime/ctime on client for create/unlink (2611)
4075 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4076 - free recovery state on server during a forced cleanup (3571)
4077 - unregister_reply for resent reqs (3063)
4078 - loop back devices mounting and status check on 2.6 (3563)
4079 - fix resource-creation race that can provoke i_size == 0 (3513)
4080 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4081 - more debugging for page-accounting assertion (3746)
4082 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4083 - avoid deadlock after precreation failure (3758)
4084 - fix race and lock order deadlock in orphan handling (3450, 3750)
4085 - add validity checks when grabbing inodes from l_ast_data (3599)
4087 - add /proc/.../recovery_status to obdfilter (3428)
4088 - lightweight CDEBUG infrastructure, debug daemon (3668)
4089 - change default OSC RPC parameters to be better on small clusters
4090 - turn off OST read cache for files smaller than 32MB
4091 - install man pages and include them in rpms (3100)
4092 - add new init script for (un)mounting lustre filesystems (2593)
4093 - run chkconfig in %post for init scripts (3701)
4094 - drop scimac NAL (unmaintained)
4096 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4099 - clean kiobufs before and after use (3485)
4100 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4101 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4102 - change init script to fail more clearly if not run as root (1528)
4103 - allow clients to reconnect during replay (1742)
4104 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4105 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4106 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4107 - fix cond_resched (3554)
4108 - only evict unfinished clients after recovery (3515)
4109 - allow bulk resend, prevent data loss (3570)
4110 - dynamic ptlrpc request buffer allocation (2102)
4111 - don't allow unlinking open directory if it isn't empty (2904)
4112 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4113 - remove extraneous obd dereference causing LASSERT failure (3334)
4114 - don't use get_cycles() when creating temp. files on the mds (3156)
4115 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4116 - handle EEXIST for set-stripe, set proper directory name (3336)
4118 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4119 - fix ksocknal_fmb_callback() error messages (2918)
4121 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4124 - don't copy lvb into (possibly NULL) reply on error (2983)
4125 - don't deref dentry after dput, don't free lvb on error (2922)
4126 - use the kms to determine writeback rpc length (2947)
4127 - increment oti_logcookies when osc is inactive (2948)
4128 - update client's i_blocks count via lvb messages (2543)
4129 - handle intent open/close of special files properly (1557)
4130 - mount MDS with errors=remount-ro, like obdfilter (2009)
4131 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4132 - don't use cancelling-locks' kms values (2947)
4133 - use highest lock extent for kms, not last one (2925)
4134 - don't dereference ERR_PTR() dentry in error handling path (3107)
4135 - fix thread race in portals_debug_dumplog() (3122)
4136 - create lprocfs device entries at setup instead of at attach (1519)
4137 - common AST error handler, don't evict client on completion race (3145)
4138 - zero nameidata in detach_mnt in 2.6 (3118)
4139 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4140 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4141 - initialize RPC timeout timer earlier for 2.6 (3219)
4142 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4143 - print nal/nid for unknown nid (3258)
4144 - additional checks for oscc recovery before doing precreate (3284)
4145 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4146 - don't crash in mdc_close for bad permissions on open (3285)
4147 - zero i_rdev for non-device files (3147)
4148 - clear page->private before handing to FS, better assertion (3119)
4149 - tune the read pipeline (3236)
4150 - fix incorrect decref of invalidated dentry (2350)
4151 - provide read-ahead stats and refine rpc in flight stats (3328)
4152 - don't hold journal transaction open across create RPC (3313)
4153 - update atime on MDS at close time (3265)
4154 - close LDAP connection when recovering to avoid server load (3315)
4155 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4156 - don't leak open file on MDS after open resend (3325)
4157 - serialize filter_precreate and filter_destroy_precreated (3329)
4158 - loop device shouldn't call sync_dev() for nul device (3092)
4159 - clear page cache after eviction (2766)
4160 - resynchronize MDS->OST in background (2824)
4161 - refuse to mount the same filesystem twice on same mountpoint (3394)
4162 - allow llmount to create routes for mounting behind routers (3320)
4163 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4164 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4165 - fix rare problem with rename on htree directories (3417)
4167 - allow default OST striping configuration per directory (1414)
4168 - fix compilation for qswnal for 2.6 kernels (3125)
4169 - increase maximum number of MDS request buffers for large systems
4170 - change liblustreapi to be useful for external progs like lfsck (3098)
4171 - increase local configuration timeout for slow disks (3353)
4172 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4174 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4177 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4178 - don't overwrite extent policy data in reply if lock was blocked (2901)
4179 - drop filter export grants atomically with removal from device (2663)
4180 - del obd_self_export from work_list in class_disconnect_exports (2908)
4181 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4182 - swab reply message in mdc_close, other PPC fixes (2464)
4183 - fix destroying of named logs (2325)
4184 - overwrite old logs when running lconf --write_conf (2264)
4185 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4186 - fix race in target_handle_connect (2898)
4187 - mds_reint_create() should take same inode create lock (2926)
4188 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4189 - don't close files for self_export to avoid uninitialized obd (2936)
4190 - allow MDS with the same name as client node (2939)
4191 - hold dentry reference for closed log files for unlink (2325)
4192 - reserve space for all logs during transactions (2059)
4193 - don't evict page beyond end of stripe extent (2925)
4194 - don't oops on a deleted current working directory (2399)
4195 - handle hard links to targets without a parent properly (2517)
4196 - don't dereference NULL lock when racing during eviction (2867)
4197 - don't grow lock extents when lots of conflicting locks (2919)
4199 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4202 - account for cache space usage on clients to avoid data loss (974)
4203 - lfsck support in lustre kernel code (2349)
4204 - reduce journal credits needed for BRW writes (2370)
4205 - orphan handling to avoid losing space on client/server crashes
4206 - ptlrpcd can be blocked, stopping ALL progress (2477)
4207 - use lock value blocks to assist in proper KMS, faster stat (1021)
4208 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4209 - recovery for initial connections (2355)
4210 - fixes for mds_cleanup_orphans (1934)
4211 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4212 - block all file creations until orphan recovery completes (1901)
4213 - client remove rq_connection from request struct (2423)
4214 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4215 - recovery timer race (2670)
4216 - mdc_close recovey bug (2532)
4217 - ptlrpc cleanup bug (2710)
4218 - mds timeout on local locks (2588)
4219 - namespace lock held during RPCs (2431)
4220 - handle interrupted sync write properly (2503)
4221 - don't try to handle a message that hasn't been replied to (2699)
4222 - client assert failure during cleanup after abort recovery (2701)
4223 - leak mdc device after failed mount (2712)
4224 - ptlrpc_check_set allows timedout requests to complete (2714)
4225 - wait for inflight reqs when ptlrpcd finishes (2710)
4226 - make sure unregistered services are removed from the srv_list
4227 - reset bulk XID's when resending them (caught by 1138 test)
4228 - unregister_bulk after timeout
4229 - fix lconf error (2694)
4230 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4231 - readahead locks pages, leaves pending causing memory pressure (2673)
4232 - increase OST request buffers to 4096 on large machines (2729)
4233 - fix up permission of existing directories in simple_mkdir (2661)
4234 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4235 - don't assign transno to errored transactions (2742)
4236 - don't delete objects on OST if given a bogus objid from MDS (2751)
4237 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4238 - if rq_no_resend, then timeout request after recovery (2432)
4239 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4240 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4241 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4242 - MDS recovery shouldn't delete objects that it has given out (2730)
4243 - if enqueue arrives after completion, don't clobber LVB (2819)
4244 - don't unlock pages twice when trigger_group_io returns error (2814)
4245 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4246 - don't write pages to disk if there was an error (1450)
4247 - don't ping imports that have recovery disabled (2676)
4248 - take buffered bytes into account when balancing socknal conn (2817)
4249 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4250 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4251 - remove little-endian swabbing of llog records (1987)
4252 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4253 - retry reposting request buffers if they fail (1191)
4254 - grow extent at grant time to avoid granting a revoked lock (2809)
4255 - lock revoke doesn't evict page if covered by a second lock (2765)
4256 - disable VM readahead to avoid reading outside lock extents (2805)
4258 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4259 - updated kernel patches for hp-2.4.20 kernel (2681)
4261 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4264 - fix truncated write corruption (2366)
4265 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4267 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4268 - ensure len(uuid) < 37 in lmc (1171)
4269 - fix ia64 OOPS in llog_test (2255)
4270 - zero end of page at obdfilter for partial page writes (2648)
4271 - don't leave stale dentries around after renames (bug 2428)
4272 - fix timeouts when evicting a client with a single lock held (2642)
4273 - set deadline for the initial HELLO message to drain (2634)
4274 - print out dotted-quad IP addresses in the socknal (2302)
4276 - additional debugging for MDS client eviction problem (2443)
4277 - fix mkfsoptions support for osts (2603, 2604)
4279 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4282 - add series for the vanilla 2.6.0 kernel
4283 - add series for the vanilla 2.4.24 kernel
4284 - add series for a cray x86/64 UL kernel drop
4285 - fix xattr patches for the vanilla 2.4.19 series
4287 - generate true UUIDs in lmc (1171)
4288 - have portals stack dumping break in UML (2466)
4289 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4290 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4291 - wake callback waiting threads on client eviction (2460)
4292 - Add --ptldebug and --subsystem to lmc (1719)
4293 - update assertion to allow safe interrupt allocation
4294 - set rq_no_resend for cancel requests (2432)
4295 - recalculate ptlrpcd timeout after resend (2494)
4296 - call vfs_rmdir when removing pending directories (2368)
4297 - fix renaming a file to itself (2429)
4298 - lmc creates a default one-stripe lov (2454)
4299 - expand procfs space to handle large clusters (2326)
4300 - increase UML stack to avoid overflow
4301 - update lconf's list of debug and subsystem masks
4302 - fix lfs find --obd (2510)
4303 - /proc tunable for disabling filter read caching (2591)
4304 - stop rpm packages from altering slapd.conf (2301)
4305 - disable nagle in the socknal under 0conf (2578)
4306 - choose mds inode size based on stripe count (2572)
4307 - fix kernel-source rpm problems (2516)
4309 - add --disable-doc to avoid pdf generation (2421)
4310 - update documentation, tests, type-os, comments
4311 - avoid format warnings on ia64
4312 - remove the TOE NAL
4313 - tiny code cleanups by removing unused fields
4315 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4318 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4319 - remove the most blatant lies from BUILDING file (2371)
4320 - change default debug level to reasonable production setting
4321 - reduce client side cache size to reduce cache flush time
4322 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4323 - make TCP zerocopy and pinger support enabled by default (2476)
4324 - sync writes completed after process exits caused crashes (2319)
4325 - maintain correct mount count on the MDS (2356)
4326 - backout 1557, because 2316 wasn't really fixed
4327 - better file I/O statistics gathering in /proc
4328 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4329 - another kernel patch to fix zero-copy TCP function export
4330 - don't take duplicate lock when processing re-sent getattr (2420)
4331 - lctl uses obd_self_export instead of creating new conn (2353)
4332 - MDS/OST recovery case which requires object creation asserted (2425)
4333 - move lfs from /usr/sbin to /usr/bin in packages
4334 - fix race between mds_client_add and mds_client_free (2417)
4335 - use kmalloc instead of slabs in portals (2430)
4336 - don't create duplicate records when a failover MDS is present (2442)
4337 - remove unnecessary mount age check (2332)
4338 - don't remove directory inodes from locks prematurely (2451)
4339 - don't break if MDS service name is the same as hostname (2103)
4340 - fix races in client write RPC generation when cache full (2482)
4342 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4345 - remove now-unused request->rq_obd (278)
4346 - if an allocation fails, print out how much memory we've used (1933)
4347 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4348 - add the "configurable stack size" patch to most series files (1256)
4349 - ability to write large log records, for 100+ OST configs (2306)
4350 - fix NULL deref when filter_prep fails (2314)
4351 - fix operator precedence error in filter_sync
4352 - dynamic allocation of socknal TX descriptors (2315)
4353 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4354 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4355 - ensure that CWARN messages in Portals always get to the syslog
4356 - __init/__exit are not for prototype decls (ldlm_init/exit)
4357 - x86-64 compile warning fixes
4358 - fix gateway LMC keyword conflict (2318)
4359 - fix MDS lock inversions in getattr/reint paths (1844)
4360 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4361 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4362 - DLM race condition prevented some lock evictions (2328)
4363 - ENOMEM detection and retry on socknal sends (2230)
4364 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4365 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4366 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4367 - better error messages when a client is rejected during recovery (1505)
4368 - avoid cancelling locks which were never granted, after failure (2330)
4369 - fix i_sem/journal inversion in mds_client_add (2333)
4370 - fix truncate/getattr lock cycle deadlock (2334)
4371 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4372 - fix two rare exit paths which could leak an l_lock() ref (2321)
4373 - fencepost error in MDS/OST orphan recovery (2226)
4374 - make log record alignment 8 bytes (1988)
4375 - lstripe now fails when requested offset > ost_count (2237)
4376 - ensure that all kernel series have a complete list.h (1607)
4377 - fix crashes in special-file operations (2316)
4378 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4380 - allow configurable automake binary, for testing new versions
4381 - small update to the lfs documentation
4383 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4385 * fix negative export reference count in fsfilt_sync (2312)
4387 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4388 * release candidate 0.9.1
4390 - orphans are moved into the PENDING directory for possible recovery
4391 - replayed opens now open by fid for orphan/rename safety (1042)
4392 - last close of an orphan inode generates a transno (683)
4393 - chdir() and mount() now pin the directory entry (1020)
4394 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4395 - discard very old requests without processing them (1502)
4396 - remove some common, well-understood CERRORs (1505)
4397 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4398 - clear "grant" flags in OST replies until OST grant code lands (1644)
4399 - fix read performance by not clobbering i_blksize on client (1598)
4400 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4401 - make LRU size a /proc tunable, clears locks when reduced (707)
4402 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4403 - server locks take references on exports now (1558)
4404 - build fixes for 2.4.20-rh trees (1663)
4405 - return an error from lov_create if all OSCs are inactive (1751)
4406 - fix import levels when a reconnect happens without a timeout (1597)
4407 - exit early from mds_open if we get a lookup error (1749)
4408 - partial page read at EOF wouldn't wait for disk before sending (1642)
4409 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4410 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4411 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4412 - write/truncate lock inversion (1639)
4413 - Don't auto-load obdclass, portals modules during cleanup (1495)
4414 - fix timestamps from jumping to "now" (1763)
4415 - extra journal assertions (1648)
4416 - add an extra multiunlink test (1771)
4417 - fix read_record/write_record API (1776)
4418 - fix leak of offset_extent, possible incorrect i_size later (1772)
4419 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4420 - Debugging for the kqswnal_get_idle_tx problems (1820)
4421 - Allow recovery to be attempted multiple times (1536)
4422 - Write out MDS last_rcvd file after it is first created (1600)
4423 - Fix tx_descriptor leak in failed transmit situations (1827)
4424 - ext3 journaling fixes for assertion failure after IO error (1871)
4425 - class_export_put() on freed export after completion AST error (1896)
4426 - Fix revalidate looping in VFS (1322)
4427 - Don't access a freed export during MDS_REINT timeout (1521)
4428 - Add open-unlink recovery support on the MDS (1673,1764)
4429 - Return an error if no MDS data was read from last_rcvd (1946)
4430 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4431 - Don't setup a disk device that doesn't match exported UUID (317)
4432 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4433 - avoid committing NULL handle in force close
4434 - local.sh is now a one-stripe LOV configuration
4435 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4436 - fix ext3 htree duplicate directory entry corruption (1516)
4437 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4438 - update to new LOV EA format (2097)
4439 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4441 2003-06-15 Phil Schwan <phil@clusterfs.com>
4444 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4445 - per-import/export recovery handling (958, 931, 959)
4446 - multiple last-rcvd slots, for serving multiple FSes (949)
4447 - connections are again shared between multiple imp/exports (963, 964)
4448 - "umount -f" would hang if any requests needed to be sent (393, 978)
4449 - avoid pinning large req buffer by copying for queued messages (989)
4450 - add "uuid" to "lctl device" command to help upcalls (991)
4451 - "open" RPCs with transnos would confuse recovery counters (1037)
4452 - do proper endian conversion of all wire messages (288, 340, 891)
4453 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4454 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4455 - fix ptlrpc_connection leak in target_handle_connect (1174)
4456 - fix import refcounting bug in OST and MDS cleanup (1134)
4457 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4458 - fix very unlikely obd_types race condition (501)
4459 - remove osc_open hack for echo_client (1187)
4460 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4461 - a failure in read_inode2 leads to deadlock (1139)
4462 - cancel ack-locks as soon as transaction is committed (1072)
4463 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4464 - make sure to commitrw after any preprw to avoid deadlock (1162)
4465 - failing to execute a file in a lustre FS would lock inode (1203)
4466 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4467 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4468 - obd_uuid2tgt would walk off the end of the list (1255)
4469 - on IA64 the osc would give portals incorrect bulk size (1258)
4470 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4471 - fix lock inversion caused by new llite matching code (1282)
4472 - limit the number of dirty pages on a client to 10MB (1286)
4473 - timed out locks were not being corrected cancelled (1289)
4474 - fix O_DIRECT above 4GB on IA-32 (1292)
4475 * major user-visible changes
4476 - fail out/fail over policy now controlled by the upcall (993)
4478 - add OBD_PING to check server availability and failure (954)
4479 - lustre messages are now sent in sending host order (288, 340, 891)
4480 - add eadatalen to MDS getattr reply (340)
4481 - OST read replies may contain second buffer, with per-page status (593)
4483 2003-03-11 Phil Schwan <phil@clusterfs.com>
4486 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4487 - failed open()s could cause deadlock; fixed (867, 869)
4488 - stop cancelling OST locks when files are closed (481)
4489 - overlapping XID spaces caused network corruption (851, 853)
4490 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4491 - setattr_raw added, to do single-RPC, server-side setattrs
4492 - lmc/lconf syntax change for OST UUIDs
4493 - fix crashy race condition between ptlrpc_free_req and osc_close
4494 - don't use request in mdc_enqueue if we hit a timeout (889)
4495 - don't set the inode i_size for regular files from the MDS (896)
4496 - handle out of order completion AST (842)
4497 - don't LBUG if a lock request times out after receiving AST (913)
4498 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4499 - if a bad lock AST arrives, send an error instead of dropping entirely
4500 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4501 - fix leak in bulk IO when only partially completed (899, 900, 926)
4502 - fix O_DIRECT for ia64 (55)
4503 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4504 - C-z after timeout could hang a process forever; fixed (977)
4506 - client-side I/O cache (678, 924, 929, 941, 970)
4508 - READPAGE and SETATTRs which don't take server-side locks get
4511 2003-02-11 Phil Schwan <phil@clusterfs.com>
4514 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4515 - Fix fsfilt_extN_readpage() to read a full page of directory
4516 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4517 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4518 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4519 - Most intent operations were removed, in favour of a new RPC mode
4520 that does a single RPC to the server and bypasses most of the VFS
4521 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4522 - Aggressively cancel local locks on DLM servers
4523 - mds_reint_unlink sends EA to the client if it's the last nlink.
4524 client uses that EA to unlink OST objects.
4525 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4526 - recursive symlinks were fixed (439)
4527 - fixed NULL deref in DEBUG_REQ
4528 - filter_update_lastobjid no longer calls sync, which annoyed extN
4529 - fixed multi-client small-writes to a single file problem (445)
4530 - fixed mtime updates during file writes (607)
4531 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4532 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4533 - fixed Linux OST size reporting problem (444, 656)
4534 - OST now updates object mtime with writes or setattr (607, 619)
4535 - client verifies file size before zeroing page past EOF (445)
4536 - OST now writes last allocated objid to disk with allocation (108)
4537 - LOV on echo now works (409)
4539 - mds_reint_unlink sends a new buffer, with the EA included. this
4540 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4541 set if a regular file was being unlinked, and it was the last link
4542 - use PtlGet from the target for bulk writes (315)
4543 - OST now updates object mtime with writes or setattr (607, 619)
4544 - LDLM now has a grant-time callback to revalidate locked items, if
4546 - Many MDS operations were reorganized to combat race conditions
4548 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4549 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4551 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4554 - Fully reactivate OST imports after reconnection (512, others)
4555 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4556 - More graceful error handling for truncating on dead OST (515)
4557 - Don't error out unless we're actually accessing dead stripes (474)
4558 - Fix garbage sizes when stripes are missing (410)
4559 - LRU counters were broken, causing constant lock purge (433, 432)
4560 - garbage on read from stripes with failed OSTs (441)
4561 - mark OSCs as active before reconnecting during recovery (438)
4562 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4563 - lfind did not preserve OST order in output (443)
4564 - symlinks cause hung clients, incorrect data (439)
4565 - stop dereferencing request after dropping refcount (457)
4566 - don't LASSERT(spin_is_locked) on non-SMP (455)
4567 - fixes for many rename() bugs
4568 - fstat didn't correctly synchronize attributes (399)
4569 - server must handle lock cancellation during blocking AST prep (487)
4570 - bulk descriptors were free()d too soon (511)
4571 - fix paths in lconf, which would load incorrect modules (451, 507)
4572 - fix confusing lconf 'host not found' error message (386)
4573 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4574 - fix race condition in mdc_blocking_ast() for inode access (526)
4575 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4576 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4577 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4578 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4579 - fix lstripe in several ways (kernel side) (527)
4580 - fix request leak in ldlm_cli_enqueue (262)
4581 - incorrect OSC was marked inactive after OST failure
4582 - call mds_fs_cleanup before unmounting filesystem (524)
4583 - fix races between taking ns_lock and ldlm_lock_change_resource
4584 - fix races updating LOV export open file list
4585 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4586 - fix recovery NULL deref in ldlm_cli_cancel_unused
4587 - fix some DLM races by using new hash table for lock handles (419)
4588 - permit the client to specify desired inodes, at replay
4589 - duplicate requests when we queue them for replay reintegration
4590 - fix last_rcvd offset calculation
4591 - sync after each recovered transaction, so we always make progress
4592 - never, not always, ERESTART requests without transnos
4593 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4595 - skip replay if the MDS says that the client is already connected
4596 - don't check for a recovery-enabled export to match lctl's UUID
4597 - don't INC_USE_COUNT for phantom exports
4598 - don't crash when cleaning up phantom exports (567)
4599 - don't double-finish or set replay data for errored mdc_open requests
4600 - abort requests when they time out, so we don't get old replies
4601 - send/receive replies for AST messages again
4602 - if the client says that it doesn't have the lock, cancel it on the
4604 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4605 mark it as destroyed, it will all work out when decref is called
4606 - fix module use counts (22, 581)
4608 - ASTs now expect a reply (server cancels lock on error reply)
4610 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4613 - fix many simultaneous client startup (392)
4614 - fix dentry->d_it clobbering
4615 - credentials weren't being shipped for readdir/getattr operations
4616 - remove invalid assertions triggered during some concurrent MD
4618 - proper Lustre versions added (336, 389)
4619 - fix memory leak for create error case (398)
4620 - fix LOV locking bug that would get cli/srv out of sync
4621 - fix echo client over LOV (409)
4622 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4623 - fix double-O_EXCL intent crash (424)
4624 - avoid sending multiple lock CANCELs (352)
4626 - MDS can do multi-client recovery (modulo bugs in new code)
4628 - many updates, edits, cleanups
4630 2002-11-18 Phil Schwan <phil@clusterfs.com>
4633 - fix null d_it dereference (346)
4634 - fix full OST/dbench hang (333)
4635 - fix permission problem with file removal (286)
4636 - fix removal of OSCs from LOV when they fail
4637 - fix NULL deref during bulk timeout (214)
4638 - fix problems related to multiple filesystems on one MDS (241)
4639 - fixed serious subtle metadata locking bugs
4640 - free locks on clients when inodes are removed due to memory
4642 - fix inode pointer in lock data (285)
4643 - partial support for multiple MDS on a single host (241)
4644 - data locks weren't cancelled at clear_inode time (290, 311)
4645 - intent locks could lead to unbounded lock growth (205)
4646 - added a maximum lock count, an LRU list, and a flusher
4647 - fix multiple rename (365)
4648 - properly abstracted the echo client
4649 - OSC locked 1 byte too many; fixed
4650 - rewrote brw callback code:
4651 - fixed recovery bugs related to LOVs (306)
4652 - fixed too-many-pages-in-one-write crash (191)
4653 - fixed (again) crash in sync_io_timeout (214)
4654 - probably fixed callback-related race (385)
4656 - Add capability to MDS protocol
4657 - LDLM cancellations and callbacks on different portals
4659 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4662 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4663 - defer open object destruction to close time (601981, 138)
4664 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4665 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4666 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4667 - don't call obd_setattr in ll_file_release for destroyed objects
4668 * protocol change to lustre_msg: move |version| and add |flags|
4669 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4670 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4671 * added replay of create, unlink, link and rename operations during
4672 MDS failover; recovery should be much more robust now
4673 * remove failed OSCs from LOVs (only lov_create uses this so far)
4674 * the lustre-HOWTO was brought (more) up to date (582544)
4676 2002-10-23 Phil Schwan <phil@clusterfs.com>
4679 - in-use dentries weren't being reused properly (617851)
4680 - prevent multiple LDLM setup (599178)
4681 - fix LOV size calculations for truncate (617853)
4682 - fix client handling of MDS intent errors (POSIX)
4683 - fix permission bug in lovstripe.c test (624321)
4684 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4685 - truncate past end of file could corrupt data
4686 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4687 - a race in recovery could return ETIMEDOUT to apps (623947)
4688 - building outside the source directory was fixed
4689 * the lustre-HOWTO was brought (more) up to date (582544)
4690 * major progress was made on recovery functionality
4692 2002-10-10 Phil Schwan <phil@clusterfs.com>
4695 - recovery deadlock fix
4696 - rm -rf causes LBUG fix (617817)
4697 - file open by multiple tasks fix (618962)
4698 - directory permissions bugs (602707 and 620007)
4699 - journal_stop fixed with locking (611313)
4700 - O_APPEND failures resolved (618273, perhaps 614459)
4701 - lconf PATH fix (619770)
4702 - IA64 build fix (621450)
4703 - RPC buffer sizes scale with amount of memory
4705 2002-10-01 Phil Schwan <phil@clusterfs.com>
4708 - locks would be cancelled without throwing away data pages,
4709 resulting in inconsistent data (605627)
4710 - inode attributes were not always being refreshed (605627, 612449)
4711 - lconf now continues to cleanup after lctl reports an error
4712 - MDS now enforces user permissions (602707)
4713 - lprocfs cleanup fixed, but not yet enabled (614157)
4714 - fixed infinite server hang, should a client not respond to an AST
4715 - avoid going into recovery if user calls readlink() with a buffer
4716 that's too small (613941)
4717 - AST RPCs no longer require replies (614867) -- this may be changed
4718 - don't crash server if client sends an IOV that's too big (611336)
4719 - fixed lock conversion deadlock (611892)
4720 - fixed the following of symlinks (614622)
4721 * recovery: the server can remove locks from a client that dies, other
4722 clients can make progress
4723 * more extN patch fixes
4724 * compile-time configurable ptlrpc buffer allocations
4726 - collaborative read cache document
4727 - Lustre Lite Performance CDR document-in-progress
4729 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4732 - fix typo in patch-2.4.18
4734 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4737 - clear ptlrpc request each time in handle_incoming_request()
4738 - unlink of files now destroys the object on the OST
4740 2002-09-19 Peter Braam <braam@clusterfs.com>
4742 * add hard link support
4743 * change obdfile creation method
4744 * kernel patch changed
4746 2002-09-19 Peter Braam <braam@clusterfs.com>
4749 - stack overflow bug in extN fixed
4751 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4753 * documentation updates
4754 - add man pages for config tools
4755 - update tests/README to describe testing with new config tools
4756 - finish metadata API descriptions
4757 * bug fixes and cleanups
4758 - statfs workaround for 16TB limit
4759 - LOV stripe allocation improved, can stripe on subset of OSTs
4760 - LOV file size/IO offset was wrong for files > 4GB in size
4761 - object EA data was being dropped, caused files to be unreadable
4762 - memory overflow with non-LOV OST caused memory corruption
4763 - fixed regression tests to work with new config tools, obdfilter
4764 - fixed bug when directory size became larger than 1 block
4765 - fixed bug (for single client case) when PWD was deleted
4766 - invalidate local directory pages when doing intent-based ops
4767 - avoid LDLM oops when lock callback contained bad data
4769 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4771 * documentation updates
4772 * bug fixes and cleanups
4773 - configuration tools
4776 - 64-bit compile warnings
4777 - 64-bit internal statfs data
4779 * test_brw on persistent OST devices
4781 * lprocfs (disabled)
4783 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4785 * documentation updates
4786 * bug fixes and cleanups
4787 * configuration tools
4789 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4792 * many small fixes to 0_5_4
4793 * io/network handling
4794 * thinkos in MDS operations
4796 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4799 * crucial basic fixes to 0.5.3
4800 * IOR, Iozone work over Elan
4803 2002-08-07 Phil Schwan <phil@clusterfs.com>
4804 * version 0_5_3, our first alpha
4805 * we use the new Portals iovs
4806 * documentation updates
4807 * bug fixes and cleanups
4808 * small changes in the DLM wire protocol
4810 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4811 * version 0_5_1 with some initial stability,
4812 * locking on MD and file I/O.
4813 * documentation updates
4814 * several bug fixes since 0.5.0
4815 * small changes in wire protocol
4817 2002-07-18 Phil Schwan <phil@clusterfs.com>
4819 * delivered as Lustre Light Alpha
4820 * fixed a crash after handling invalid MDS requests
4821 * fixed directory pages for architectures with non-4k pages sizes
4823 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4824 * release version v0_4_4
4825 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4827 * Fixes a number of interruption problems with OST operations.
4828 * Update documentation for portals header changes
4829 * Move all wire protocol structs/defines to lustre_idl.h
4830 * Fixes symlink length bug.
4831 * Add tcpdump to repository.
4833 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4834 * release version v0_4_3
4835 * Fixes statfs for inodes on extN.
4836 * Fixes bug in runtests which would delete /etc/hosts.
4837 * Use 64-bit object IDs wherever possible (not into VFS though)
4838 Remove ost_get_info, which is unused by lustre, and out of date.
4840 2002-07-03 Peter Braam <braam@clusterfs.com>
4841 * release version v0_4_2 Fixes a lookup error (type not passed)
4842 * move forward to head of Portals
4843 * move forward to latest Lustre kernel
4845 2002-06-25 Peter Braam <braam@clusterfs.com>
4846 * release version v0_4_1. Hopefully stable on single node use.