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).
17 Description: LBUG during stress test
18 Details : Need properly lock accesses the flock deadlock detection list.
22 Description: oops in page fault handler
23 Details : kernel page fault handler can return two special 'pages' in error case, don't
24 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
28 Description: timeout with invalidate import.
29 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
30 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
31 ptlrpc_invalidate_import and as result LASSERT.
33 Severity : enhancement
35 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
39 Description: enable MGS and MDT services start separately
40 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
41 a co-located MGS without starting the MGS, which is a complement
42 to 'nosvc' mount option.
45 Frequency : always with o2ib 1.3 and sles10
47 Description: fix build with SLES10 and o2ib v3.
48 Details : sles10 uses diffrent name for Module.symver file but configure
49 assume this file has same name on RHEL/SLES/vanila kernels.
53 Frequency : after recovery
54 Description: precreate to many object's after del orphan.
55 Details : del orphan st in oscc last_id == next_id and this triger growing
56 count of precreated objects. Set flag LOW to skip increase count
57 of precreated objects.
61 Frequency : rare, on clear nid stats
62 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
63 Details : when clean nid stats sometimes try destroy live entry,
64 and this produce panic in free.
68 Description: Stack overflow during MDS log replay
69 ease stack pressure by using a thread dealing llog_process.
73 Description: wait until IO finished before start new when do lock cancel.
74 Details : VM protocol want old IO finished before start new, in this case
75 need wait until PG_writeback is cleared until check dirty flag and
76 call writepages in lock cancel callback.
78 Severity : enhancement
80 Description: using special macro for print time and cleanup in includes.
84 Description: mds_mfd_close() ASSERTION(rc == 0)
85 Details : In mds_mfd_close(), we need protect inode's writecount change
86 within its orphan write semaphore to prevent possible races.
90 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
94 Frequency : rare, on shutdown ost
95 Description: don't hit live lock with umount ost.
96 Details : shrink_dcache_parent can be in long loop with destroy dentries,
97 use shrink_dcache_sb instead.
101 Description: don't panic with use echo client
102 Details : echo client pass NULL as client nid pointer and this produce null
107 Description: fix build on ppc32
108 Details : compile code with -m64 flag produce wrong object file for ppc32.
112 Description: add message levels for liblustreapi
116 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
117 Details : Move the 'good_osts' check before the 'total_bavail' check. This
118 will result in an -EAGAIN and in the exit call path we call
119 alloc_rr() which will with increasing aggressiveness attempt to
120 aquire precreated objects on the minimum number of required OSCs.
124 Description: Use old size assignment to avoid deadlock
125 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
126 the scheduling while holding a spinlock. We do not need locking
127 for size in ll_update_inode() because size is only updated from
128 the MDS for directories or files without objects, so there is no
129 other place to do the update, and concurrent access to such inodes
130 are protected by the inode lock.
134 Description: resolve "_IOWR redefined" build error on SLES10
138 Description: dump the memory debugging after all modules are unloaded to
139 suppress false negative in conf_sanity test 39
141 Severity : enhancement
143 Description: build kernel-ib packages for OFED 1.3 in our release cycle
148 Description: fix SLES kernel versioning
149 Details : the kernel version for our SLES 10 kernel did not include a "-"
150 before the "smp" at the end. while this was not a problem in
151 general, it did mean that software trying to use the kernel
152 version to try to detect a vendor specific kernel would fail.
153 this was most evident by the OFED build scripts.
157 Description: Don't update lov_desc members until making sure they are valid
158 Details : When updating lov_desc members via proc fs, need fix their
159 validities before doing the real update.
163 Description: don't put request into delay list while invalidate in flight.
164 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
165 import in flight. this produce timeout for invalidate and sometimes
166 can cause stale data.
170 Frequency : on ppc only
171 Description: not convert ost objects for directory because it's not exist.
172 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
176 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
178 Severity : enhancement
180 Description: Update to support 2.6.22.14 vanilla kernel.
184 Frequency : rare, on recovery
185 Description: read procfs can produce deadlock in some situation
186 Details : Holding lprocfs lock which send rpc can produce block for destroy
187 obd objects and this also block reconnect with -EALREADY. This isn't
188 fix all lprocfs bugs - but make it rare.
190 Severity : enhancement
192 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
195 Frequency : frequent on X2 node
197 Description: mdc_set_open_replay_data LBUG
198 Details : Set replay data for requests that are eligible for replay.
202 Description: lustre_mgs: operation 101 on unconnected MGS
203 Details : When MGC is disconnected from MGS long enough, MGS will evict the
204 MGC, and late on MGC cannot successfully connect to MGS and a lot
205 of the error messages complaining that MGS is not connected.
209 Frequency : on network error
210 Description: panic with double free request if network error
211 Details : mdc_finish_enqueue is finish request if any network error ocuring,
212 but it's true only for synchronus enqueue, for async enqueue
213 (via ptlrpcd) this incorrect and ptlrpcd want finish request
216 Severity : enhancement
218 Description: client-side metadata stat-ahead during readdir(directory readahead)
219 Details : perform client-side metadata stat-ahead when the client detects
220 readdir and sequential stat of dir entries therein
223 Frequency : on start mds
225 Description: Implement get_info(last_id) in obdfilter.
228 Frequency : occasional
230 Description: Correctly check stale fid, not start epoch if ost not support SOM
231 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
232 overwrited with new generated fid, but mds can anwer with one of these
233 two fids and both is not stale. setattr incorectly start epoch and
234 assume will be called done_writeting, but without SOM done_writing
238 Frequency : rare, depends on device drivers and load
240 Description: MDS or OSS nodes crash due to stack overflow
241 Details : Code changes in 1.8.0 increased the stack usage of some functions.
242 In some cases, in conjunction with device drivers that use a lot
243 of stack the MDS (or possibly OSS) service threads could overflow
244 the stack. One change which was identified to consume additional
245 stack has been reworked to avoid the extra stack usage.
248 Frequency : occasional
250 Description: Do not fail import if osc_interpret_create gets -EAGAIN
251 Details : If osc_interpret_create got -EAGAIN it immediately exits and
252 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
253 oscc_has_objects and see OSC has no objests and call
254 oscc_internal_create to resend create request.
256 Severity : enhancement
258 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
260 Severity : enhancement
262 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
265 Frequency : very rare
267 Description: Fix replay if there is an un-replied request and open
268 Details : In some cases, older replay request will revert the
269 mcd->mcd_last_xid on MDS which is used to record the client's
272 Severity : enhancement
274 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
276 Severity : enhancement
278 Description: Add rhel5 support to HEAD.
280 Serverity : enhancement
282 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
287 Description: Don't allow skipping OSTs if index has been specified.
288 Details : Don't allow skipping OSTs if index has been specified, make locking
289 in internal create lots better.
293 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
294 Details : access to bitfield in structure is always rounded to long
295 and this produce problem with not atomic change any bit.
299 Description: Lustre make rpms failed.
300 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
301 builds Lustre rpms from tarball.
304 Frequency : rare on shutdown ost
306 Description: If llog cancel was not send before clean_exports phase, this can
307 produce deadlock in llog code.
308 Details : If llog thread has last reference to obd and call class_import_put
309 this produce deadlock because llog_cleanup_commit_master wait when
310 last llog_commit_thread exited, but this never success because was
311 called from llog_commit_thread.
315 Description: allow userland application know is lost one of stripes.
316 Details : fill lvb_blocks with error code on ost and return it to
317 application if error flag found.
321 Description: NULL lov_tgts causing MDS oops
322 Details : more safe checks for NULL lov_tgts for avoid oops.
324 Severity : enhancement
326 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
330 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
331 Details : If an OST is down the MDS will hang indefinitely in
332 obd_statfs_rqset() waiting for the statfs data. While for
333 MDS QOS usage of statfs, it should not stuck in waiting.
335 Severity : enhancement
337 Description: remote_acl support
338 Details : Support ACL-based permission check for remote user.
339 Support setfacl/getfacl for remote user with the utils
340 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
341 the system "{s,g}etfacl" utils.
343 Severity : enhancement
345 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
347 Severity : enhancement
349 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
353 Description: make mgs_setparam() handle fsname containing dash
354 Details : fsname containing a dash does not work with lctl conf_param
356 Severity : enhancement
358 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
360 Severity : enhancement
362 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
366 Frequency : rare, at shutdown
367 Description: access already free / zero obd_namespace.
368 Details : if client_disconnect_export was called without force flag set,
369 and exist connect request in flight, this can produce access to
370 NULL pointer (or already free pointer) when connect_interpret
371 store ocd flags in obd_namespace.
375 Frequency : only at startup
376 Description: not alloc memory with spinlock held.
377 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
378 if any spinlock held.
380 Severity : enhancement
382 Description: make lustre randomly fail allocating memory
383 Details : Make lustre randomly failed allocating memory for testing purpose.
385 Severity : enhancement
387 Description: lost problems with lov objid file
388 Details : Fixes some scability and access to not inited memory problems
389 in work with lov objdid file.
394 Description: lfs find does not continue on file error
395 Details : Continue other files processing when a file/dir is absent.
399 Description: Inconsistent usage of lustre_pack_reply()
400 Details : Standardize the usage of lustre_pack_reply() such that it
401 always generate a CERROR on failure.
404 Frequency : occasional
406 Description: MDS hang or stay in waiting lock
407 Details : If client receive lock with CBPENDING flag ldlm need send lock
408 cancel as separate rpc, to avoid situation when cancel request
409 can't processed due all i/o threads stay in wait lock.
413 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
417 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
421 Description: Update to SLES9 kernel-2.6.5-7.287.3.
425 Description: Update to RHEL4 latest kernel.
427 Severity : enhancement
429 Description: Build SLES10 patchless client fails
430 Details : The configure was broken by run ./configure with
431 --with-linux-obj=.... argument for patchless client. When the
432 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
433 can't search header adequately. Use absolute path such as
434 -I($LINUX)/include instead.
438 Description: interrupt oig_wait produce painc on resend.
439 Details : brw_redo_request can be used for resend requests from ptlrpcd and
440 private set, and this produce situation when rq_ptlrpcd_data not
441 copyed to new allocated request and triggered LBUG on assert
442 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
443 ptlrpcd set if request is changed and can be safety changed to use
448 Description: LASSERT_{REQ,REP}SWAB macros are buggy
449 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
450 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
451 these macros and replace them with their difinitions instead.
455 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
456 Details : Update which_patch & target file for SLES10 latest kernel.
459 Description: Remove client patches from SLES 10 kernel.
460 Details : This causes SLES 10 clients to behave as patchless clients
461 even on a Lustre-patched (server) kernel.
463 Severity : enhancement
465 Description: self-adjustable client's lru lists
466 Details : use adaptive algorithm for managing client cached locks lru
467 lists according to current server load, other client's work
468 pattern, memory activities, etc. Both, server and client
469 side namespaces provide number of proc tunables for controlling
472 Severity : enhancement
474 Description: light-weight GSS support
475 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
476 and reduce performance overhead.
478 Severity : enhancement
480 Description: Linux keyring support
481 Details : Support using service of Linux keyring for Lustre GSS internal
482 context refresh/cache mechanism.
486 Description: Fix errors in lfs documentation
487 Details : Fixes man pages
491 Description: don't use GFP_* in generic Lustre code.
492 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
496 Description: obdclass is limited by single OBD_ALLOC(idarray)
497 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
501 Description: warnings with build patchless client with vanila 2.6.19 and up
502 Details : change old ctl_table style and replace ctl_table/ctl_table_header
503 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
507 Frequency : Only for SLES
508 Description: sanity_quota fail test_1
509 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
510 file, which makes TSTID[2] inunique.
514 Frequency : Only for SLES9
515 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
516 Details : Update target/ChangeLog/which_patch .
520 Frequency : Only for test_10 in sanity-quota.sh
521 Description: a bug in quota test code
522 Details : checking a test flag in a wrong place
526 Frequency : Only for test_13 in sanity-quota.sh
527 Description: a bug in quota test script
528 Details : shouldn't check overfull allocated quota
532 Description: jbd statistics
533 Details : Port older jbd statistics patch for sles10
535 Severity : enhancement
537 Description: Add printing inode info into message about error in writepage.
542 Description: reply_lock_interpret crash due to race with it and lock cancel.
543 Details : Do not replay locks that are being cancelled. Do not reference
544 locks by their address during replay, just by their handle.
548 Frequency : When flocks are used.
549 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
550 Details : Flock locks might destroy just granted lock if it could be merged
551 with another existing flock, this is done in completion handler,
552 so teach ldlm_cli_enquque_fini that this is a valid case for
558 Description: Oops in read and write path when failing to allocate lock.
559 Details : Check if lock allocation failed and return error back.
563 Description: lstripe command fails for valid OST index
564 Details : The stripe offset is compared to lov->desc.ld_tgt_count
565 instead of lov->desc.ld_active_tgt_count.
569 Description: sanity.sh failed test 103
570 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
575 Description: df doesn't work properly if diskfs blocksize != 4K
576 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
580 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
581 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
585 Description: obd_health_check_timeout too short
586 Details : set obd_health_check_timeout as 1.5x of obd_timeout
590 Description: llapi_file_create() does not allow some changes
591 Details : add llapi_file_open() that allows specifying the mode and
592 open flags, and also returns an open file handle.
596 Description: merge and cleanup kernel patches.
597 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
601 Description: Add journal checksum support.(Kernel part)
602 Details : The journal checksum feature adds two new flags i.e
603 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
604 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
605 indicates that the commit block contains the checksum for
606 the blocks described by the descriptor blocks. Now commit
607 record can be sent to disk without waiting for descriptor
608 blocks to be written to disk. This behavior is controlled
609 using JBD2_FEATURE_ASYNC_COMMIT flag.
613 Description: OSS needs mutliple precreate threads
614 Details : Add ability to start more than one create thread per OSS.
618 Description: Sanity.sh test_65k failed: setstripe should have succeeded
619 Details : execute lfs setstripe on client
623 Description: mds_obd_create error creating tmp object
624 Details : When the user sets quota on root, llog will be affected and can't
625 create files and write files.
628 Frequency : Always on ia64 patchless client, and possibly others.
630 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
631 Details : This allows the patchless client to be loaded on architectures
636 Description: RedHat Update kernel for RHEL5
637 Details : Add the kernel config file for RHEL5.
641 Description: RedHat Update kernel for RHEL5
642 Details : Modify the kernel config file more closer RHEL5.
646 Description: Build failure against Centos5 (RHEL5)
647 Details : Define PAGE_SIZE when it isn't present.
651 Description: test_7 on sanity-quota.sh can't be used on separate machines
652 Details : add facet to handle it
656 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
657 Details : with NFS, the anon dentry's parent was set to itself in
658 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
659 corresponding dentry other than use rec->ur_name.
661 Severity : enhancement
663 Description: lfs setstripe enhancement
664 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
668 Description: enable data checksumming by default
669 Details : enable checksum by default, allow --disable-checksum
670 configure option and "-o nochecksum" mount option. Checksums
671 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
675 Description: lustre support for RHEL5
676 Details : Add support for RHEL5.
680 Description: Client eviction due to failover config
681 Details : after a connection loss, the lustre client should attempt to
682 reconnect to the last active server first before trying the
683 other potential connections.
687 Description: when mds and osts use different quota unit(32bit and 64bit),
688 quota will be released repeatly.
689 Details : void sending multiple quota reqs to mds, which will keep the
690 status between the reqs.
694 Description: osts not allocated evenly to files
695 Details : change the condition to increase offset_idx
698 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
700 Description: Data corruption for OSTs that are formatted larger than 2TB
702 Details : When generating the bio request for lustre file writes the
703 sector number would overflow a temporary variable before being
704 used for the IO. The data reads correctly from Lustre (which
705 will overflow in a similar manner) but other file data or
706 filesystem metadata may be corrupted in some cases.
710 Description: Tune the kernel for good SCSI performance.
711 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
712 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
717 Description: rewrite ext2-derived code in obdclass/uuid.c
718 Details : rewrite inherited code (uuid parsing code from ext2 utils)
719 from scratch preserving functionality.
723 Description: "lfs find -obd UUID" prints directories
724 Details : "lfs find -obd UUID" will return all directory names instead
725 of just file names. It is incorrect because the directories
726 do not reside on the OSTs.
730 Description: use i_size_read and i_size_write in 2.6 port
731 Details : replace inode->i_size access with i_size_read/write()
734 Frequency : when removing large files
736 Description: scheduling issue during removal of large Lustre files
737 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
738 It causes scheduling issues when removing large files (17TB in the
742 Frequency : only with liblustre clients on XT3
744 Description: evictions taking too long
745 Details : allow llrd to evict clients directly on OSTs
748 Frequency : only on ppc
750 Description: /proc/fs/lustre/devices broken on ppc
751 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
752 We should make sure the type of 'index' is loff_t and then cast
753 explicitly as needed below. Do not assign an explicitly cast
757 Frequency : only for rhel5
759 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
760 Details : Modify the target file & which_kernel.
762 Severity : enhancement
764 Description: omit set fsid for export NFS
765 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
766 as FS_REQUIRES_DEV for avoid problems with generate fsid.
770 Frequency : Always, for kernels after 2.6.16
771 Description: Fix warning idr_remove called for id=.. which is not allocated.
772 Details : Last kernels save old s_dev before kill super and not allow
773 to restore from callback - restore it before call kill_anon_super.
777 Description: buffer overruns could theoretically occur
778 Details : llapi_semantic_traverse() modifies the "path" argument by
779 appending values to the end of the origin string, and a
780 overrun may occur. Adding buffer overrun check in liblustreapi.
784 Description: Fix error on 'ls .' at the top of the Lustre mount.
785 Details : Don't revalidate dentry if it is a root dentry.
789 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
790 Details : Modify vm-tunables-rhel4.patch.
794 Description: Kernel config for 2.6.18-vanilla.
795 Details : Modify targets/2.6-vanilla.target.in.
796 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
797 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
798 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
799 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
803 Description: improve handling recoverable errors
804 Details : if request processig with error which can be recoverable on server
805 request should be resend, otherwise page released from cache and
810 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
811 Details : Modify target file & which_patch.
812 A flaw was found in the IA32 system call emulation provided
813 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
814 value could be stored in the %RAX register, which could trigger an
815 out-of-bounds system call table access. An untrusted local user
816 could exploit this flaw to run code in the kernel
817 (ie a root privilege escalation). (CVE-2007-4573).
821 Description: O_DIRECT bypasses client statistics.
822 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
823 read_ahead_stats were updated. copied the stats section from
824 osc_send_oap_rpc() into async_internal().
828 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
832 Description: change order of libsysio includes
833 Details : '#include sysio.h' should always come before '#include xtio.h'
837 Description: handle "serious error: objid * already exists" more gracefully
838 Details : If LAST_ID value on disk is smaller than the objects existing in
839 the O/0/d* directories, it indicates disk corruption and causes an
840 LBUG(). If the object is 0-length, then we should use the existing
841 object. This will help to avoid a full fsck in most cases.
843 Severity : enhancement
845 Description: adapt the lustre_config script to support the upgrade case
846 Details : Add "-u" option for lustre_config script to support upgrading 1.4
847 server targets to 1.6 in parallel.
851 Description: To avoid grant space > avaible space when the disk is almost
852 full. Without this patch you might see the error "grant XXXX >
853 available" or some LBUG about grant, when the disk is almost
855 Details : In filter_check_grant, for non_grant cache write, we should
856 check the left space by if (*left > ungranted + bytes), instead
857 of (*left > ungranted), because only we are sure the left space
858 is enough for another "bytes", then the ungrant space should be
859 increase. In client, we should update cl_avail_grant only there
860 is OBD_MD_FLGRANT in the reply.
864 Description: Update RHEL 4 kernel to fix local root privilege escalation.
865 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
866 described in CVE-2007-4573. This problem could allow untrusted
867 local users to gain root access.
870 Frequency : when using O_DIRECT and quotas
872 Description: Incorrect file ownership on O_DIRECT output files
873 Details : block usage reported by 'lfs quota' does not take into account
874 files that have been written with O_DIRECT.
879 Description: touch file failed when fs is not full
880 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
881 otherwise we can get ENOSP while fs is not full.
885 Description: parallel lock callbacks
886 Details : Instead of sending blocking and completion callbacks as separated
887 requests, adding them to a set and sending in parallel.
890 Frequency : only for Cray XT3
891 Bugzilla : 12829/13455
892 Description: Changing primary group doesn't change the group lustre assigns to
894 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
895 overridden with the primary group provided by the group upcall,
896 whereas the supplied fsgid can be trusted if it is in the list of
897 supplementary groups returned by the group upcall.
899 Severity : enhancement
901 Description: Allow masking D_WARNING, D_ERROR messages from console
902 Details : Console messages can now be disabled via lnet.printk.
906 Description: User code with malformed file open parameter crashes client node
907 Details : Before packing join_file req, all the related reference should be
908 checked carefully in case some malformed flags cause fake
909 join_file req on client.
913 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
914 Details : ldlm_completion_ast() assumes that a lock is granted when the req
915 mode is equal to the granted mode. However, it should also check
916 that LDLM_FL_CP_REQD is not set.
920 Description: Heavy nfs access might result in deadlocks
921 Details : After ELC code landed, it is now improper to enqueue any mds
922 locks under och_sem, because enqueue might want to decide to
923 cancel open locks for same inode we are holding och_sem for.
927 Description: Client eviction while running blogbench
928 Details : A lot of unlink operations with concurrent I/O can lead to a
929 deadlock causing evictions. To address the problem, the number of
930 oustanding OST_DESTROY requests is now throttled to
931 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
932 callbacks are processed in priority.
936 Description: enable ACLs on MDS by default
937 Details : ACLs must be enabled on MDS by default.
940 Frequency : PPC/PPC64 only
942 Description: conflicts between asm-ppc64/types.h and lustre_types.h
943 Details : fix duplicated definitions between asm-ppc64/types.h and
944 lustre_types.h on PPC.
947 Frequency : PPC/PPC64 only
949 Description: asm-ppc/segment.h does not exist
950 Details : fix compile issue on PPC.
954 Description: better handle error messages in extents code
957 Frequency : RHEL4 only
959 Description: mkfs is very slow on IA64/RHEL4
960 Details : A performance regression has been discovered in the MPT Fusion
961 driver between versions 3.02.73rh and 3.02.99.00rh. As a
962 consequence, we have downgraded the MPT Fusion driver in the RHEL4
963 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
965 Severity : enhancement
967 Description: SNMP support enhancement
968 Details : Adding total number of sampled request for an MDS node in snmp
971 Severity : enhancement
973 Description: Optimize ldlm waiting list processing for PR extent locks
974 Details : When processing waiting list for read extent lock and meeting read
975 lock that is same or wider to it that is not contended, skip
976 processing rest of the list and immediatelly return current
977 status of conflictness, since we are guaranteed there are no
978 conflicting locks in the rest of the list.
982 Description: Time out and refuse to reconnect
983 Details : When the failover node is the primary node, it is possible
984 to have two identical connections in imp_conn_list. We must
985 compare not conn's pointers but NIDs, otherwise we can defeat
986 connection throttling.
990 Description: port llog fixes from b1_6 into HEAD
991 Details : Port llog reference couting and some llog cleanups from b1_6
992 (bug 10800) into HEAD, for protect from panic and access to already
993 free llog structures.
997 Description: Detect stride IO mode in read-ahead
998 Details : When a client does stride read, read-ahead should detect that and
999 read-ahead pages according to the detected stride pattern.
1003 Description: data checksumming impacts single node performance
1004 Details : add support for several checksum algorithm. Currently, only CRC32
1005 and Adler-32 are supported. The checksum type can be changed on
1006 the fly via /proc/fs/lustre/osc/*/checksum_type.
1010 Description: use adler32 for page checksums
1011 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1016 Description: build for x2 fails
1017 Details : fix compile issue on Cray systems.
1021 Description: Properly match for duplicate locks
1022 Details : Due to different lock order from skiplists code, we need to
1023 traverse entire list for now
1026 Frequency : only on PPC/SLES10
1028 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1029 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1030 be defined. Add a hack in mkfs_lustre.c to work around this compile
1035 Description: LASSERT on MDS when client holding flock lock dies
1036 Details : ldlm pool logic depends on number of granted locks equal to
1037 number of released locks which is not true for flock locks, so
1038 just exclude such locks from consideration.
1042 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1043 Details : Use fsfilt_sync() for both the whole filesystem sync and
1044 individual file sync to eliminate dangerous inode locking
1045 with I_LOCK that can lead to a deadlock.
1049 Description: performance in 1.6.3
1050 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1051 sure that 1MB requests are not fragmented by the block layer.
1053 Severity : enhancement
1055 Description: organize the server-side client stats on per-nid basis
1056 Details : Change the structure of stats under obdfilter and mds to
1064 The "uuid"s file would list the uuids of _active_ exports.
1065 And the clear entry is to clear all stats and stale nids.
1067 Severity : enhancement
1069 Description: eliminate client locks in face of contention
1070 Details : file contention detection and lockless i/o implementation
1071 for contended files.
1075 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1078 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1080 Description: MDS may lose file striping (and hence file data) in some cases
1081 Details : If there are additional extended attributes stored on the MDS,
1082 in particular ACLs, SELinux, or user attributes (if user_xattr
1083 is specified for the client mount options) then there is a risk
1084 of attribute loss. Additionally, the Lustre file striping
1085 needs to be larger than default (e.g. striped over all OSTs),
1086 and an additional attribute must be stored initially in the
1087 inode and then increase in size enough to be moved to the
1088 external attribute block (e.g. ACL growing in size) for file
1093 Description: skiplist implementation simplification
1094 Details : skiplists are used to group compatible locks on granted list
1095 that was implemented as tracking first and last lock of each lock group
1096 the patch changes that to using doubly linked lists
1100 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1101 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1102 because LASSERT on both the data supplied by a client, and the data
1103 on disk is dangerous and incorrect.
1105 Severity : enhancement
1107 Description: Slow trucate/writes to huge files at high offsets.
1108 Details : Directly associate cached pages to lock that protect those pages,
1109 this allows us to quickly find what pages to write and remove
1110 once lock callback is received.
1112 --------------------------------------------------------------------------------
1114 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1116 * Support for kernels:
1117 2.6.5-7.283 (SLES 9),
1118 2.6.9-55.EL (RHEL 4),
1119 2.6.16.46-0.14 (SLES 10),
1120 2.6.18.8 vanilla (kernel.org)
1121 * Client support for unpatched kernels:
1122 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1123 2.6.16 - 2.6.22 vanilla (kernel.org)
1124 * Due to recently discovered recovery problems, we do not recommend
1125 using patchless RHEL 4 clients with this or any earlier release.
1126 * Recommended e2fsprogs version: 1.39.cfs8
1127 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1128 * Starting with this release, the ldiskfs backing filesystem required
1129 by Lustre is now in its own package, lustre-ldiskfs. This package
1130 should be installed. It is versioned separately from Lustre and
1131 may be released separately in future.
1135 Description: block reactivating mgc import until all deactivates complete
1136 Details : Fix race when failing back MDT/MGS to itself (testing)
1139 Frequency : at statup only
1141 Description: mds_lov_synchronize race leads to various problems
1142 Details : simultaneous MDT->OST connections at startup can cause the
1143 sync to abort, leaving the OSC in a bad state.
1145 Severity : enhancement
1147 Description: add optional extra BUILD_VERSION info
1148 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1149 to override the lustre version.
1152 Frequency : 2.6.18 servers only
1154 Description: ll_kern_mount() doesn't release the module reference
1155 Details : The ldiskfs module reference count never drops down to 0
1156 because ll_kern_mount() doesn't release the module reference.
1161 Description: server LBUG when using old ost_num_threads parameter
1162 Details : Accept the old ost_num_threads parameter but warn that it
1163 is deprecated, and fix an off-by-one error that caused an LBUG.
1168 Description: Transient SCSI error results in persistent IO issue
1169 Details : iobuf->dr_error is not reinitialized to 0 between two
1173 Frequency : sometimes when underlying device returns I/O errors
1175 Description: OSTs not going read-only during write failures
1176 Details : OSTs are not remounted read-only when the journal commit threads
1177 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1178 instead of the ext3 wrappers.
1182 Description: poor connect scaling with increasing client count
1183 Details : Don't run filter_grant_sanity_check for more than 100 exports
1184 to improve scaling for large numbers of clients.
1187 Frequency : SLES10 only
1189 Description: sanity-quota.sh quotacheck failed: rc = -22
1190 Details : Quotas cannot be enabled on SLES10.
1193 Frequency : liblustre clients only
1195 Description: getdirentries does not give error when run on compute nodes
1196 Details : getdirentries does not fail when the size specified as an argument
1197 is too small to contain at least one entry
1199 Severity : enhancement
1201 Description: Add LNET router traceability for debug purposes
1202 Details : If a checksum failure occurs with a router as part of the
1203 IO path, the NID of the last router that forwarded the bulk data
1204 is printed so it can be identified.
1209 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1210 Details : Due to a race condition, liblustre clients were occasionally
1211 evicted incorrectly.
1213 Severity : enhancement
1215 Description: lfs setstripe use optional parameters instead of postional
1218 Severity : enhancement
1220 Description: Nanosecond timestamp support for ldiskfs
1221 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1222 resolution timestamps. There is not yet support for this at
1223 the Lustre filesystem level.
1226 Frequency : during server recovery
1228 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1229 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1233 Frequency : nfs export on patchless client
1235 Description: connectathon hang when test nfs export over patchless client
1236 Details : Disconnected dentry cannot be found with lookup, so we do not need
1237 to unhash it or make it invalid
1240 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1241 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1242 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1247 Description: open req refcounting wrong on reconnect
1248 Details : If reconnect happened between getting open reply from server and
1249 call to mdc_set_replay_data in ll_file_open, we will schedule
1250 replay for unreferenced request that we are about to free.
1251 Subsequent close will crash in variety of ways.
1252 Check that request is still eligible for replay in
1253 mdc_set_replay_data().
1258 Description: disable writes to filesystem when reading health_check file
1259 Details : the default for reading the health_check proc file has changed
1260 to NOT do a journal transaction and write to disk, because this
1261 can cause reads of the /proc file to hang and block HA state
1262 checking on a healthy but otherwise heavily loaded system. It
1263 is possible to return to the previous behaviour during configure
1264 with --enable-health-write.
1266 Severity : enhancement
1268 Description: 64-bit inode version
1269 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1270 to the inode. This will be required for version-based recovery.
1275 Description: MDS fails to start if a duplicate client export is detected
1276 Details : in some rare cases it was possible for a client to connect to
1277 an MDS multiple times. Upon recovery the MDS would detect this
1278 and fail during startup. Handle this more gracefully.
1280 Severity : enhancement
1282 Description: Add -o localflock option to simulate old noflock
1284 Details : This will achieve local-only flock/fcntl locks
1290 Description: log_commit_thread vs filter_destroy race leads to crash
1291 Details : Take import reference before releasing llog record semaphore
1296 Description: Wrong request locking in request set processing
1297 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1298 imp_delayed_list, in this place should be used imp_lock.
1301 Frequency : when reconnection
1303 Description: Grant Leak when osc reconnect to OST
1304 Details : When osc reconnect ost, OST(filter) should check whether it
1305 should grant more space to client by comparing fed_grant and
1306 cl_avail_grant, and return the granted space to client instead
1307 of "new granted" space, because client will call osc_init_grant
1308 to update the client grant space info.
1311 Frequency : when client reconnect to OST
1313 Description: Grant Leak when osc do resend and replay bulk write
1314 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1315 bulk write request, because the grant info while be sync between
1316 OSC and OST when reconnect, and we should ignore the grant info
1317 these of resend/replay write req.
1322 Description: Grant space more than avaiable left space sometimes.
1323 Details : When then OST is about to be full, if two bulk writing from
1324 different clients came to OST. Accord the avaliable space of the
1325 OST, the first req should be permitted, and the second one
1326 should be denied by ENOSPC. But if the seconde arrived before
1327 the first one is commited. The OST might wrongly permit second
1328 writing, which will cause grant space > avaiable space.
1331 Frequency : when client is evicted
1333 Description: Grant might be wrongly erased when osc is evicted by OST
1334 Details : when the import is evicted by server, it will fork another
1335 thread ptlrpc_invalidate_import_thread to invalidate the
1336 import, where the grant will be set to 0. While the original
1337 thread will update the grant it got when connecting. So if
1338 the former happened latter, the grant will be wrongly errased
1339 because of this race.
1344 Description: Checking Stale with correct fid
1345 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1346 to check whether it is stale, because sometimes, we want the
1347 enqueue happened anyway, and op_data.fid2 will not be initialized.
1349 Severity : enhancement
1351 Description: update patchless client
1352 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1355 Frequency : only with 2.4 kernel
1357 Description: random memory corruption
1358 Details : size of struct ll_inode_info is to big for union inode.u and this
1359 can be cause of random memory corruption.
1364 Description: Memory leak in recovery
1365 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1366 It should also check obd_fail before fsfilt_start, otherwise if
1367 fsfilt_start return -EROFS,(failover mds during mds recovery).
1368 then the req will return with repmsg->transno = 0 and rc = EROFS.
1369 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1370 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1371 be freed no matter whether fsfilt_commit success or not.
1374 Frequency : only with huge count clients
1376 Description: Prevents from taking the superblock lock in llap_from_page for
1378 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1379 for llap_from_page call in ll_removepage prevents from taking the
1380 superblock lock for a soon died page.
1385 Description: Not check open intent error before release open handle
1386 Details : in some rare cases, the open intent error is not checked before
1387 release open handle, which may cause
1388 ASSERTION(open_req->rq_transno != 0), because it tries to release
1389 the failed open handle.
1394 Description: Set cat log bitmap only after create log success.
1395 Details : in some rare cases, the cat log bitmap is set too early. and it
1396 should be set only after create log success.
1400 Description: Accessing a block bevice can re-enable I/O when Lustre is
1401 tearing down a device.
1402 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1407 Description: service threads may hog cpus when there are a lot of requests
1408 Details : Insert cond_resched to give other threads a chance to use some CPU
1413 Description: the cat log was not initialized in recovery
1414 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1415 unlink log on mds will not be initialized until mds post
1416 recovery. And also in mds post recovery, the unlink log will
1417 initialization will be done asynchronausly, so there will be race
1418 between add unlink log and unlink log initialization.
1422 Description: brw_stats were being printed incorrectly
1423 Details : brw_stats were being printed as log2 but all of them were not
1424 recorded as log2. Also remove some code duplication arising from
1425 filter_tally_{read,write}.
1429 Frequency : rare, only in recovery.
1430 Description: ASSERTION(req->rq_type != LI_POISON) failed
1431 Details : imp_lock should be held while iterating over imp_sending_list for
1432 prevent destroy request after get timeout in ptlrpc_queue_wait.
1436 Description: replay-single.sh test 52 fails
1437 Details : A lock's skiplist need to be cleanup when it being unlinked
1438 from its resource list.
1442 Description: Short directio read returns full requested size rather than
1444 Details : Direct I/O operations should return actual amount of bytes
1445 transferred rather than requested size.
1447 Severity : enhancement
1449 Description: metadata RPC reduction (e.g. for rm performance)
1450 Details : decrease the amount of synchronous RPC between clients and servers
1451 by canceling conflicing lock before the operation on the client
1452 and packing thier handles into the main operation RPC to server.
1454 Severity : enhancement
1456 Description: Async OSC create to avoid the blocking unnecessarily.
1457 Details : If a OST has no remain object, system will block on the creating
1458 when need to create a new object on this OST. Now, ways use
1459 pre-created objects when available, instead of blocking on an
1460 empty osc while others are not empty. If we must block, we block
1461 for the shortest possible period of time.
1463 Severity : enhancement
1465 Description: refine locking for avoid write wrong info into lov_objid file
1466 Details : fix possible races with add new target and write/update data in
1469 --------------------------------------------------------------------------------
1471 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1476 Frequency : on some architectures
1478 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1479 Details : Uninitialized flags sometimes cause configuration commands to
1483 Frequency : patchless clients only
1485 Description: missing __iget() symbol export
1486 Details : The __iget() symbol export is missing. To avoid the need for
1487 this on patchless clients the deathrow inode reaper is turned
1488 off, and we depend on the VM to clean up old inodes. This
1489 dependency was during via the fix for bug 12181.
1491 --------------------------------------------------------------------------------
1493 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1495 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1496 INTEROPERATE with older versions automatically. In many cases a
1497 special upgrade step is needed. Please read the
1498 user documentation before upgrading any part of a live system.
1499 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1500 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1501 * WARNING: Lustre configuration and startup changes are required with
1502 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1508 Severity : enhancement
1510 Description: MountConf
1511 Details : Lustre configuration is now managed via mkfs and mount
1512 commands instead of lmc and lconf. New obd types (MGS, MGC)
1513 are added for dynamic configuration management. See
1514 https://mail.clusterfs.com/wikis/lustre/MountConf for
1517 Severity : enhancement
1519 Description: dynamic OST addition
1520 Details : OSTs can now be added to a live filesystem
1522 Severity : enhancement
1524 Description: startup order invariance
1525 Details : MDTs and OSTs can be started in any order. Clients only
1526 require the MDT to complete startup.
1528 Severity : enhancement
1530 Description: parallel, asynchronous orphan cleanup
1531 Details : orphan cleanup is now performed in separate threads for each
1532 OST, allowing parallel non-blocking operation.
1534 Severity : enhancement
1536 Description: optimized stripe assignment
1537 Details : stripe assignments are now made based on ost space available,
1538 ost previous usage, and OSS previous usage, in order to try
1539 to optimize storage space and networking resources.
1541 Severity : enhancement
1543 Description: Permanently set tunables
1544 Details : All writable /proc/fs/lustre tunables can now be permanently
1545 set on a per-server basis, at mkfs time or on a live system.
1547 Severity : enhancement
1549 Description: Lustre message v2
1550 Details : Add lustre message format v2.
1552 Severity : enhancement
1554 Description: client OST exclusion list
1555 Details : Clients can be started with a list of OSTs that should be
1556 declared "inactive" for known non-responsive OSTs.
1560 Description: SPEC SFS validation failure on NFS v2 over lustre.
1561 Details : Changes the blocksize for regular files to be 2x RPC size,
1562 and not depend on stripe size.
1564 Severity : enhancement
1566 Description: Multiple MD RPCs in flight.
1567 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1568 To avoid overly-overloading MDS, introduce a limit on number of
1569 MDS RPCs in flight for a single client and add /proc controls
1570 to adjust this limit.
1572 Severity : enhancement
1574 Description: client read/write statistics
1575 Details : Add client read/write call usage stats for performance
1576 analysis of user processes.
1577 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1578 file access. extents_stats shows chunk size distribution.
1579 extents_stats_per_process show chunk size distribution per
1582 Severity : enhancement
1584 Description: mds statistics
1585 Details : Add detailed mds operations statistics in
1586 /proc/fs/lustre/mds/*/stats.
1590 Description: Failure of copying files with lustre special EAs.
1591 Details : Client side always return success for setxattr call for lustre
1592 special xattr (currently only "trusted.lov").
1597 Description: Refcount LNET uuids
1598 Details : The global LNET uuid list grew linearly with every startup;
1599 refcount repeated list entries instead of always adding to
1602 Severity : enhancement
1604 Description: Dynamic service threads
1605 Details : Within a small range, start extra service threads
1606 automatically when the request queue builds up.
1609 Frequency : mixed-endian client/server environments
1611 Description: mixed-endian crashes
1612 Details : The new msg_v2 system had some failures in mixed-endian
1615 Severity : enhancement
1617 Description: Easy OST removal
1618 Details : OSTs can be permanently deactivated with e.g. 'lctl
1619 conf_param lustre-OST0001.osc.active=0'
1621 Severity : enhancement
1623 Description: MGS proc entries
1624 Details : Added basic proc entries for the MGS showing what filesystems
1627 Severity : enhancement
1629 Description: provide MGS failover
1630 Details : Added config lock reacquisition after MGS server failover.
1632 Severity : enhancement
1634 Description: add Linux 2.4 support
1635 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1639 Description: a large application tries to do I/O to the same resource and dies
1640 in the middle of it.
1641 Details : Check the req->rq_arrival time after the call to
1642 ost_brw_lock_get(), but before we do anything about
1643 processing it & sending the BULK transfer request. This
1644 should help move old stale pending locks off the queue as
1645 quickly as obd_timeout.
1648 Frequency : when an incorrect nid is specified during startup
1650 Description: ptlrpc connect to non-existant node causes kernel crash
1651 Details : LNET can't be re-entered from an event callback, which
1652 happened when we expire a message after the export has been
1653 cleaned up. Instead, hand the zombie cleanup off to another
1656 Severity : enhancement
1658 Description: plain/inodebits lock performance improvement
1659 Details : Grouping plain/inodebits in granted list by their request modes
1660 and bits policy, thus improving the performance of search through
1664 Frequency : only if OST filesystem is corrupted
1666 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1667 Details : for a short time RPCs with bulk IO are in the replay list,
1668 but replay of bulk IOs is unimplemented. If the OST filesystem
1669 is corrupted due to disk cache incoherency and then replay is
1670 started it is possible to trip an assertion. Avoid putting
1671 committed RPCs into the replay list at all to avoid this issue.
1674 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1676 Description: System hang on startup
1677 Details : This bug allowed the liblustre (e.g. catamount) client to
1678 return to the app before handling all startup RPCs. This
1679 could leave the node unresponsive to lustre network traffic
1680 and manifested as a server ptllnd timeout.
1682 Severity : enhancement
1684 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1685 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1686 boolean to control whether to print peer debug info when a
1687 client's RPC times out.
1690 Frequency : only for kernels with patches from Lustre below 1.4.3
1692 Description: Remove old rdonly API
1693 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1696 Frequency : only for devices with external journals
1698 Description: Set external device read-only also
1699 Details : During a commanded failover stop, we set the disk device
1700 read-only while the server shuts down. We now also set any
1701 external journal device read-only at the same time.
1704 Frequency : when upgrading from 1.4 while trying to change parameters
1706 Description: The wrong (new) MDC name was used when setting parameters for
1707 upgraded MDT's. Also allows changing of OSC (and MDC)
1708 parameters if --writeconf is specified at tunefs upgrade time.
1711 Frequency : when setting specific ost indicies
1713 Description: QOS code breaks on skipped indicies
1714 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1715 created with --index need not be sequential.
1717 Severity : enhancement
1719 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1720 Details : The uninit_groups feature works in conjunction with the kernel
1721 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1722 up the pass1 processing of e2fsck. This is a read-only feature
1723 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1724 mount filesystems that have had this feature enabled.
1726 Severity : enhancement
1728 Description: Improve multi-block allocation algorithm to avoid fragmentation
1729 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1730 allocation locality and avoid filesystem fragmentation.
1732 ------------------------------------------------------------------------------
1734 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1736 * Support for kernels:
1737 2.4.21-47.0.1.EL (RHEL 3)
1738 2.6.5-7.283 (SLES 9)
1739 2.6.9-42.0.10.EL (RHEL 4)
1740 2.6.12.6 vanilla (kernel.org)
1741 2.6.16.27-0.9 (SLES 10)
1742 * Recommended e2fsprogs version: 1.39.cfs5
1744 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1748 Frequency : occasional, depends on client load and configuration
1749 Bugzilla : 12181, 12203
1750 Description: data loss for recently-modified files
1752 Details : In some cases it is possible that recently written or created
1753 files may not be written to disk in a timely manner (this should
1754 normally be within 30s unless client IO load is very high).
1755 The problem appears as zero-length files or files that are a
1756 multiple of 1MB in size after a client crash or client eviction
1757 that are missing data at the end of the file.
1759 This problem is more likely to be hit on clients where files are
1760 repeatedly created and unlinked in the same directory, clients
1761 have a large amount of RAM, have many CPUs, the filesystem has
1762 many OSTs, the clients are rebooted frequently, and/or the files
1763 are not accessed by other nodes after being written.
1765 The presence of the problem can be detected by looking at
1766 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1767 smaller than the second (nr_unused) then dirty files will not
1768 be flushed automatically to disk. "sync; sleep 10" should be
1769 run several times on the node before unmounting it to update
1770 Lustre (this is also safe to run on nodes without this problem).
1772 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1773 that can cause this same problem, so customers using that kernel
1774 also need to update the kernel in addition to Lustre. In order
1775 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1777 It is normal that files written just before a client crash (less
1778 than 30s) may not yet have been flushed to disk, even for local
1782 Frequency : frequent on thin XT3 nodes
1784 Description: UUID collision on thin XT3 Linux nodes
1785 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1786 randomly, since we relied on an insufficiently-seeded PRNG.
1791 Description: OSS hangs after "All ost request buffers busy"
1792 Details : A deadlock between quota and journal operations caused OSS
1793 hangs after printing "All ost request buffers busy."
1796 Frequency : always on liblustre builds
1798 Description: Cleanup compiler warnings on liblustre
1801 Frequency : always on liblustre builds on XT3
1803 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1807 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1808 > 32000 subdirectories
1809 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1810 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1811 subdirectory count crosses 32000. This will aid e2fsck to
1812 correctly handle more than 32000 subdirectories.
1815 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1817 Description: System hang on startup
1818 Details : This bug allowed the liblustre (e.g. catamount) client to
1819 return to the app before handling all startup RPCs. This
1820 could leave the node unresponsive to lustre network traffic
1821 and manifested as a server ptllnd timeout.
1823 Severity : enhancement
1825 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1826 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1827 boolean to control whether to print peer debug info when a
1828 client's RPC times out.
1833 Description: make O_SYNC working on 2.6 kernels
1834 Details : 2.6 kernels use different method for mark pages for write,
1835 so need add a code to lustre for O_SYNC work.
1840 Description: Failure to close file and release space on NFS
1841 Details : Put inode details into lock acquired in ll_intent_file_open.
1842 Use mdc_intent_lock in ll_intent_open to properly
1843 detect all kind of errors unhandled by mdc_enqueue.
1848 Description: proc file read during shutdown sometimes raced obd removal,
1850 Details : Add lock to prevent obd access after proc file removal.
1853 Frequency : Only for files larger than 4GB on 32-bit clients.
1855 Description: improperly doing page alignment of locks
1856 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1857 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1862 Description: RPCs being resent when they shouldn't be
1863 Details : Some RPCs that should not be resent are being resent. This
1864 can cause inconsistencies in the RPC state machine. Do not
1865 resend such requests.
1868 Frequency : rare, only with NFS export
1870 Description: Crash on NFS re-export node
1871 Details : under very unusual load conditions an assertion is hit in
1872 ll_intent_file_open()
1875 Frequency : only if OST filesystem is corrupted
1877 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1878 Details : for a short time RPCs with bulk IO are in the replay list,
1879 but replay of bulk IOs is unimplemented. If the OST filesystem
1880 is corrupted due to disk cache incoherency and then replay is
1881 started it is possible to trip an assertion. Avoid putting
1882 committed RPCs into the replay list at all to avoid this issue.
1887 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1888 Details : Large single O_DIRECT read and write calls can fail to allocate
1889 a sufficiently large buffer to process the request. In case of
1890 allocation failure the allocation is retried with a smaller
1891 buffer and broken into smaller requests.
1893 Severity : enhancement
1895 Description: Add -o localflock option to simulate old noflock behaviour.
1896 Details : This will achieve local-only flock/fcntl locks coherentness.
1901 Description: versioning check is incomplete
1902 Details : Checking the version difference of client vs. server, report
1903 error if the gap is too big.
1908 Description: add support PG_writeback bit
1909 Details : add support for PG_writeback bit for Lustre, for more carefull
1910 work with page cache in 2.6 kernel. This also fix some deadlocks
1911 and remove hack for work O_SYNC with 2.6 kernel.
1913 Severity : enhancement
1915 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1916 Details : The uninit_groups feature works in conjunction with the kernel
1917 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1918 up the pass1 processing of e2fsck. This is a read-only feature
1919 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1920 mount filesystems that have had this feature enabled.
1922 Severity : enhancement
1924 Description: Improve multi-block allocation algorithm to avoid fragmentation
1925 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1926 allocation locality and avoid filesystem fragmentation.
1928 ------------------------------------------------------------------------------
1930 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1932 * Support for kernels:
1933 2.6.9-42.0.3.EL (RHEL 4)
1934 2.6.5-7.276 (SLES 9)
1935 2.4.21-47.0.1.EL (RHEL 3)
1936 2.6.12.6 vanilla (kernel.org)
1937 2.6.16.21-0.8 (SLES10)
1938 * Recommended e2fsprogs version: 1.39.cfs2-0
1940 * The backwards-compatible /proc/sys/portals symlink has been removed
1941 in this release. Before upgrading, please ensure that you change
1942 any configuration scripts or /etc/sysctl.conf files that access
1943 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1944 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1945 in advance of the upgrade on any system running Lustre 1.4.6 or
1946 newer, since /proc/sys/lnet was added in that version.
1947 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1951 Frequency : only when quota is used
1953 Description: avoid scanning export list for quota master
1954 Details : Change the algorithms to avoid scanning export list in order
1955 to improve the efficiency.
1958 Frequency : MDS failover only, very rarely
1960 Description: "went back in time" messages on mds failover
1961 Details : The greatest transno may be lost when the current operation
1962 finishes with an error (transno==0) and the client's last_rcvd
1963 record is over-written. Save the greatest transno in the
1964 mds_last_transno for this case.
1967 Frequency : always for specific kernels and striping counts
1969 Description: client may get "Matching packet too big" without ACL support
1970 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1971 when trying to access files in certain configurations. The
1972 clients should in fact be denied when mounting because they do
1973 not understand ACLs.
1976 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1978 Description: many clients connecting with IO in progress causes connect timeouts
1979 Details : Avoid synchronous journal commits to avoid delays caused by many
1980 clients connecting/disconnecting when bulk IO is in progress.
1981 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1982 OST_IO_PORTAL to avoid delays behind potentially many pending
1986 Frequency : occasionally with multiple writers to a single file
1988 Description: shared writes to file may result in wrong size reported by stat()
1989 Details : Allow growing of kms when extent lock is cancelled
1992 Frequency : always with random mmap IO to multi-striped file
1994 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1995 Details : Only if the hole is at the end of OST object so that kms is too
1996 small. Fix is to increase kms accordingly in ll_nopage.
1999 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2001 Description: writes to a missing object would leak memory on the OST
2002 Details : If there is an inconsistency between the MDS and OST filesystems,
2003 such that the MDS references an object that doesn't exist, writes
2004 to that object will leak memory due to incorrect cleanup in the
2005 error handling path, eventually running out of memory on the OST.
2010 Description: Creating too long symlink causes lustre errors
2011 Details : Check symlink and name lengths before sending requests to MDS.
2014 Frequency : only if flock is enabled (not on by default)
2016 Description: posix locks not released on fd closure on 2.6.9+
2017 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2018 kernels, this caused such locks not to be released on fd close and
2019 then assertions on fs unmount about still used locks.
2022 Frequency : MDS failover only, very rarely
2024 Description: clients may get ASSERTION(granted_lock != NULL)
2025 Details : When request was taking a long time, and a client was resending
2026 a getattr by name lock request. The were multiple lock requests
2027 with the same client lock handle and
2028 mds_getattr_name->fixup_handle_for_resent_request found one of the
2029 lock handles but later failed with ASSERTION(granted_lock != NULL).
2034 Description: handle->h_buffer_credits > 0, assertion failure
2035 Details : h_buffer_credits is zero after truncate, causing assertion
2036 failure. This patch extends the transaction or creates a new
2040 Frequency : NFS re-export or patchless client
2041 Bugzilla : 11179, 10796
2042 Description: Crash on NFS re-export node (__d_move)
2043 Details : We do not want to hash the dentry if we don't have a lock.
2044 But if this dentry is later used in d_move, we'd hit uninitialised
2045 list head d_hash, so we just do this to init d_hash field but
2046 leave dentry unhashed.
2049 Frequency : NFS re-export or patchless client
2051 Description: NFS exports has problem with symbolic link
2052 Details : lustre client didn't properly install dentry when re-exported
2053 to NFS or running patchless client.
2056 Frequency : NFS re-export or patchless client
2058 Description: Various nfs/patchless fixes.
2059 Details : fixes reuse disconected alias for lookup process - this fixes
2060 warning "find_exported_dentry: npd != pd",
2061 fix permission error with open files at nfs.
2062 fix apply umask when do revalidate.
2065 Frequency : occasional
2067 Description: Crash on NFS re-export node
2068 Details : calling clear_page() on the wrong pointer triggered oops in
2069 generic_mapping_read().
2072 Frequency : rarely, using O_DIRECT IO
2074 Description: unaligned directio crashes client with LASSERT
2075 Details : check for unaligned buffers before trying any requests.
2078 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2080 Description: stale data returned from RAID cache
2081 Details : If only a small amount of IO is done to the RAID device before
2082 reading it again it is possible to get stale data from the RAID
2083 cache instead of reading it from disk.
2086 Frequency : always for sles10 kernel
2088 Description: sles10 support
2089 Details : ll_follow_link: compile fixes and using of nd_set_link
2090 under newer kernels.
2093 Frequency : depends on arch, kernel and compiler version, always on sles10
2096 Description: recursive or deep enough symlinks cause stack overflow
2097 Details : getting rid of large stack-allocated variable in
2101 Frequency : depends on hardware
2103 Description: lustre write performance loss in the SLES10 kernel
2104 Details : the performance loss is caused by using of write barriers in the
2105 ext3 code. The SLES10 kernel turns barrier support on by
2106 default. The fix is to undo that change for ldiskfs.
2108 ------------------------------------------------------------------------------
2110 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2112 * Support for kernels:
2113 2.6.9-42.0.3EL (RHEL 4)
2114 2.6.5-7.276 (SLES 9)
2115 2.4.21-47.0.1.EL (RHEL 3)
2116 2.6.12.6 vanilla (kernel.org)
2120 Frequency : quota enabled and large files being deleted
2122 Description: releasing more than 4GB of quota at once hangs OST
2123 Details : If a user deletes more than 4GB of files on a single OST it
2124 will cause the OST to spin in an infinite loop. Release
2125 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2130 Description: statfs data retrieved from /proc may be stale or zero
2131 Details : When reading per-device statfs data from /proc, in the
2132 {kbytes,files}_{total,free,avail} files, it may appear
2133 as zero or be out of date.
2136 Frequency : systems with MD RAID1 external journal devices
2138 Description: lconf's call to blkid is confused by RAID1 journal devices
2139 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2140 of returning all block devices that match the journal UUID.
2143 Frequency : always, for aggregate stripe size over 4GB
2145 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2146 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2147 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2150 Frequency : always if "lfs find" used on a local file/directory
2152 Description: "lfs find" segfaults if used on a local file/directory
2153 Details : The case where a directory component was not specified wasn't
2154 handled correctly. Handle this properly.
2157 Frequency : always on ppc64
2159 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2160 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2161 perform endianness conversion.
2164 Frequency : rarely (truncate to non-zero file size after write under load)
2165 Bugzilla : 10730, 10687
2166 Description: Files padded with zeros to next 4K multiple
2167 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2168 it is possible that files that are truncated to a non-zero size
2169 immediately after being written are filled with zero bytes beyond
2170 the truncated size. No file data is lost.
2172 Severity : enhancement
2174 Description: Allow recovery/failover for liblustre clients.
2175 Details : liblustre clients were unaware of failover configurations until
2178 Severity : enhancement
2180 Description: user file locks should fail when not mounting with flock option
2181 Details : Set up an error-returning stub in ll_file_operations.lock field
2182 to prevent incorrect behaviour when client is mounted without
2183 flock option. Also, set up properly f_op->flock field for
2187 Frequency : always on ia64
2189 Description: "lfs df" loops on printing out MDS statfs information
2190 Details : The obd_ioctl_data was not initialized and in some systems
2191 this caused a failure during the ioctl that did not return
2192 an error. Initialize the struct and return an error on failure.
2195 Frequency : SLES 9 only
2197 Description: Error of copying files with lustre special EAs as root
2198 Details : Client side always return success for setxattr call for lustre
2199 special xattr (currently only "trusted.lov").
2202 Frequency : rarely on clusters with both ia64+i386 clients
2204 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2205 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2206 clients are doing concurrent writes to the same file it is
2207 possible that the ia64 clients may LASSERT because the OST
2208 extent locks are not PAGE_SIZE aligned. Ensure that grown
2209 locks are always aligned on the request boundary.
2212 Frequency : specific use, occasional
2214 Description: Overwriting in use executable truncates on-disk binary image
2215 Details : If one node attempts to overwrite an executable in use by
2216 another node, we now correctly return ETXTBSY instead of
2217 truncating the file.
2219 Severity : enhancement
2221 Description: Async OSC create to avoid the blocking unnecessarily.
2222 Details : If a OST has no remain object, system will block on the creating
2223 when need to create a new object on this OST. Now, ways use
2224 pre-created objects when available, instead of blocking on an
2225 empty osc while others are not empty. If we must block, we block
2226 for the shortest possible period of time.
2231 Description: chmod on Lustre root is propagated to other clients
2232 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2233 invalid by the follow_mount time.
2236 Frequency : liblustre clients only
2238 Description: Race in 'instant cancel' lock handling could lead to such locks
2239 never to be granted in case of SMP MDS
2240 Details : Do not destroy not yet granted but cbpending locks in
2244 Frequency : replay/resend of open
2246 Description: non null lock assetion failure in mds_intent_policy
2247 Details : Trying to replay/resend lockless open requests resulted in
2248 mds_open() returning 0 with no lock. Now it sets a flag if
2249 a lock is going to be returned.
2251 Severity : enhancement
2253 Description: Checksum enhancements
2254 Details : New checksum enhancements allow for resending RPCs that failed
2257 Severity : enhancement
2259 Description: Tunables on number of dirty pages in cacche
2260 Details : Allow to set limit on number of dirty pages cached.
2265 Description: client crash on unmount - lock still has references
2266 Details : In some error handling cases it was possible to leak a lock
2267 reference on a client while accessing a file. This was not
2268 harmful to the client during operation, but would cause the
2269 client to crash when the filesystem is unmounted.
2272 Frequency : specific case, rare
2274 Description: ETXTBSY on mds though file not in use by client
2275 Details : ETXTBSY is no longer incorrectly returned when attempting to
2276 chmod or chown a directory that the user previously tried to
2277 execute or a currently-executing binary.
2280 Frequency : extremely rare except on liblustre-based clients
2282 Description: Lustre space not freed when files are deleted
2283 Details : Clean up open-unlinked files after client eviction. Previously
2284 the unlink was skipped and the files remained as orphans.
2289 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2290 Details : In some cases it was possible to send a blocking callback to
2291 a client doing a glimpse, even though that client didn't get
2292 a lock granted. When the glimpse lock is cancelled on the OST
2293 the freed lock is left on the waiting list and corrupted the list.
2296 Frequency : all core dumps
2298 Description: Broke core dumps to lustre
2299 Details : Negative dentry may be unhashed if parent does not have UPDATE
2300 lock, but some callers, e.g. do_coredump, expect dentry to be
2301 hashed after successful create, hash it in ll_create_it.
2303 ------------------------------------------------------------------------------
2305 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2307 * Support for kernels:
2308 2.6.9-42.0.2.EL (RHEL 4)
2309 2.6.5-7.276 (SLES 9)
2310 2.4.21-40.EL (RHEL 3)
2311 2.6.12.6 vanilla (kernel.org)
2315 Frequency : always on RHEL 3
2317 Description: Number of open files grows over time
2318 Details : The number of open files grows over time, whether or not
2319 Lustre is started. This was due to a filp leak introduced
2320 by one of our kernel patches.
2322 ------------------------------------------------------------------------------
2324 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2326 * Support for kernels:
2327 2.6.9-42.EL (RHEL 4)
2328 2.6.5-7.276 (SLES 9)
2329 2.4.21-40.EL (RHEL 3)
2330 2.6.12.6 vanilla (kernel.org)
2335 Bugzilla : 5719, 9635, 9792, 9684
2336 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2337 Details : If a server is under heavy load and cannot reply to new
2338 connection requests before the client resends the (re)connect,
2339 the connection handling code can behave badly if two service
2340 threads are concurrently handing separate (re)connections from
2341 the same client. Add better locking to the connection handling
2342 code, and ensure that only a single connection will be processed
2343 for a given client UUID, even if the lock is dropped.
2345 Severity : enhancement
2347 Description: add TCP zero-copy support to kernel
2348 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2349 sends if the hardware supports scatter-gather and checksumming.
2350 This allows socklnd to do client-write and server-read more
2351 efficiently and reduce CPU utilization from skbuf copying.
2354 Frequency : only if NFS exporting from client
2356 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2357 Details : If setting attributes on a file created under NFS that had
2358 never been opened it would be possible to oops the client
2359 if the file had no objects.
2362 Frequency : always for liblustre
2364 Description: liblustre client does MDS+OSTs setattr RPC for each write
2365 Details : When doing a write from a liblustre client, the client
2366 incorrectly issued an RPC to the MDS and each OST the file was
2367 striped over in order to update the timestamps. When writing
2368 with small chunks and many clients this could overwhelm the MDS
2369 with RPCs. In all cases it would slow down the write because
2370 these RPCs are unnecessary.
2372 Severity : enhancement
2374 Description: allow number of MDS service threads to be changed at module load
2375 Details : It is now possible to change the number of MDS service threads
2376 running. Adding "options mds mds_num_threads={N}" to the MDS's
2377 /etc/modprobe.conf will set the number of threads for the next
2378 time Lustre is restarted (assuming the "mds" module is also
2379 reloaded at that time). The default number of threads will
2380 stay the same, 32 for most systems.
2385 Description: OST crash if filesystem is unformatted or corrupt
2386 Details : If an OST is started on a device that has never been formatted
2387 or if the filesystem is corrupt and cannot even mount then the
2388 error handling cleanup routines would dereference a NULL pointer.
2393 Description: NULL pointer deref in llap_from_page.
2394 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2395 incorrect) mapping if the page was truncated/reclaimed while it was
2396 searched for. Check for this condition and skip such pages when
2397 doing readahead. Introduce extra check to llap_from_page() to
2398 verify page->mapping->host is non-NULL (so page is not anonymous).
2401 Frequency : Sometimes when using sys_sendfile
2403 Description: "page not covered by a lock" warnings from ll_readpage
2404 Details : sendfile called ll_readpage without right page locks present.
2405 Now we introduced ll_file_sendfile that does necessary locking
2406 around call to generic_file_sendfile() much like we do in
2410 Frequency : with certain MDS communication failures at client mount time
2412 Description: NULL pointer deref after failed client mount
2413 Details : a client connection request may delayed by the network layer
2414 and not be sent until after the PTLRPC layer has timed out the
2415 request. If the client fails the mount immediately it will try
2416 to clean up before the network times out the request. Add a
2417 reference from the request import to the obd device and delay
2418 the cleanup until the network drops the request.
2421 Frequency : occasionally during client (re)connect
2423 Description: assertion failure during client (re)connect
2424 Details : processing a client connection request may be delayed by the
2425 client or server longer than the client connect timeout. This
2426 causes the client to resend the connection request. If the
2427 original connection request is replied in this interval, the
2428 client may trip an assertion failure in ptlrpc_connect_interpret()
2429 which thought it would be the only running connect process.
2432 Frequency : only with obd_echo servers and clients that are rebooted
2434 Description: kernel BUG accessing uninitialized data structure
2435 Details : When running an obd_echo server it did not start the ping_evictor
2436 thread, and when a client was evicted an uninitialized data
2437 structure was accessed. Start the ping_evictor in the RPC
2438 service startup instead of the OBD startup.
2440 Severity : enhancement
2441 Bugzilla : 10193 (patchless)
2442 Description: Remove dependency on various unexported kernel interfaces.
2443 Details : No longer need reparent_to_init, exit_mm, exit_files,
2444 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2447 Frequency : rare (only users of deprecated and unsupported LDAP config)
2449 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2450 Details : LDAP apparently contains 'lustreName' attributes instead of
2451 'name'. A simple remapping of the name is sufficient.
2454 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2456 Description: waiting_locks_callback trips kernel BUG if client is evicted
2457 Details : Running with the dump_on_timeout debug flag turned on makes
2458 it possible that the waiting_locks_callback() can try to dump
2459 the Lustre kernel debug logs from an interrupt handler. Defer
2460 this log dumping to the expired_lock_main() thread.
2462 Severity : enhancement
2464 Description: Support NFS exporting on 2.6 kernels.
2465 Details : Implement non-rawops metadata methods for NFS server to use without
2466 changing NFS server code.
2469 Frequency : very rare (synthetic metadata workload only)
2471 Description: two racing renames might cause an MDS thread to deadlock
2472 Details : Running the "racer" program may cause one MDS thread to rename
2473 a file from being the source of a rename to being the target of
2474 a rename at exactly the same time that another thread is doing
2475 so, and the second thread has already enqueued these locks after
2476 doing a lookup of the target and is trying to relock them in
2477 order. Ensure that we don't try to re-lock the same resource.
2480 Frequency : only very large systems with liblustre clients
2482 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2483 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2484 all OSTs in parallel. This allows the eviction of stale liblustre
2485 clients to proceed much faster than if they were done in series,
2486 and also offers similar improvements for other set_info RPCs.
2491 Description: excessive CPU usage during initial read phase on client
2492 Details : During the initial read phase on a client, it would agressively
2493 retry readahead on the file, consuming too much CPU and impacting
2494 performance (since 1.4.5.8). Improve the readahead algorithm
2495 to avoid this, and also improve some other common cases (read
2496 of small files in particular, where "small" is files smaller than
2497 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2502 Description: MDS crash when receiving packet with unknown intent.
2503 Details : Do not LBUG in unknown intent case, just return -EFAULT
2505 Severity : enhancement
2506 Bugzilla : 9293, 9385
2507 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2508 Details : Do not serialize getattr (non-intent version) and statfs.
2511 Frequency : occasional, when OST network is overloaded/intermittent
2513 Description: client evicted by OST after bulk IO timeout
2514 Details : If a client sends a bulk IO request (read or write) the OST
2515 may evict the client if it is unresposive to its data GET/PUT
2516 request. This is incorrect if the network is overloaded (takes
2517 too long to transfer the RPC data) or dropped the OST GET/PUT
2518 request. There is no need to evict the client at all, since
2519 the pinger and/or lock callbacks will handle this, and the
2520 client can restart the bulk request.
2523 Frequency : Always when mmapping file with no objects
2525 Description: client crashes when mmapping file with no objects
2526 Details : Check that we actually have objects in a file before doing any
2527 operations on objects in ll_vm_open, ll_vm_close and
2533 Description: Request leak when working with deleted CWD
2534 Details : Introduce advanced request refcount tracking for requests
2535 referenced from lustre intent.
2537 Severity : Enhancement
2539 Description: Cache open file handles on client.
2540 Details : MDS now will return special lock along with openhandle, if
2541 requested and client is allowed to hold openhandle, even if unused,
2542 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2543 closing files for every read/write openration.
2545 Severity : Enhancement
2547 Description: Cache open negative dentries on client when possible.
2548 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2549 negative dentries on lock revocation.
2554 Description: Remounting a client read-only wasn't possible with a zconf mount
2555 Details : It wasn't possible to remount a client read-only with llmount.
2557 Severity : enhancement
2558 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2559 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2560 MPICH is included to add Lustre-specific ADIO interfaces.
2561 This is based closely on the UFS ADIO layer and only differs
2562 in file creation, in order to allow the OST striping to be set.
2563 This is user-contributed code and not supported by CFS.
2568 Description: extended inode attributes (immutable, append-only) work improperly
2569 when 2.4 and 2.6 kernels are used on client/server or vice versa
2570 Details : Introduce kernel-independent values for these flags.
2572 Severity : enhancement
2575 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2576 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2577 granularity. Now, Fractional MB granularity is be supported,
2578 this is very useful for low memory system.
2580 Severity : enhancement
2582 Description: Getattr by fid
2583 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2584 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2585 have locks and avoids getattr rpc requests for it on every stat.
2588 Frequency : Always, for filesystems larger than 2TB
2590 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2591 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2592 filesystems larger than 2TB to be mounted with mballoc enabled.
2595 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2597 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2598 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2599 filesystems larger than 2TB is mounted then the kernel will
2600 silently corrupt the start of the filesystem. CONFIG_LBD is
2601 enabled for all CFS-supported kernels, but the possibility of
2602 this happening with a modified kernel config exists.
2604 Severity : enhancement
2606 Description: add client O_DIRECT support for 2.6 kernels
2607 Details : It is now possible to do O_DIRECT reads and writes to files
2608 in the Lustre client mountpoint on 2.6 kernel clients.
2610 Severity : enhancement
2612 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2613 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2614 parallel, not waiting for response from every OST before sending
2615 a rpc to the next OST.
2620 Description: setattr vs write race when updating file timestamps
2621 Details : Client processes that update a file timestamp into the past
2622 right after writing to the file (e.g. tar) it is possible that
2623 the updated file modification time can be reset to the current
2624 time due to a race between processing the setattr and write RPC.
2626 Severity : enhancement
2628 Description: Bring 'lfs find' closer in line with regular Linux find.
2629 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2630 -print0 options and obtains all the needed info through the lustre
2633 Severity : enhancement
2635 Description: support up to 1024 configured devices on one node
2636 Details : change obd_dev array from statically allocated to dynamically
2637 allocated structs as they are first used to reduce memory usage
2642 Description: Flush dirty partially truncated pages during truncate
2643 Details : Immediatelly flush partially truncated pages in filter_setattr,
2644 this way we completely avoid having any pages in page cache on OST
2645 and can retire ugly workarounds during writes to flush such pages.
2650 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2651 Details : protect inode from truncation within vfs_unlink() context
2652 just take a reference before calling vfs_unlink() and release it
2653 when parent's i_sem is free.
2658 Description: last_id value checked outside lock on OST caused LASSERT failure
2659 Details : If there were multiple MDS->OST object precreate requests in
2660 flight, it was possible that the OST's last object id was checked
2661 outside a lock and incorrectly tripped an assertion. Move checks
2662 inside locks, and discard old precreate requests.
2665 Frequency : always, if extents are used on OSTs
2667 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2668 Details : OSTs using the extents format would not zero the high 16 bits of
2669 the index physical block number. This is not a problem for any
2670 OST filesystems smaller than 16TB, and no kernels support ext3
2671 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2672 new/modified files) and can be fixed for existing filesystems
2673 with e2fsprogs-1.39-cfs1.
2678 Description: import connection selection may be incorrect if timer wraps
2679 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2680 import connections to be ignored if the 32-bit jiffies counter
2681 wraps. Use a 64-bit jiffies counter.
2684 Frequency : very large clusters immediately after boot
2686 Description: LNET request buffers exhausted under heavy short-term load
2687 Details : If a large number of client requests are generated on a service
2688 that has previously never seen so many requests it is possible
2689 that the request buffer growth cannot keep up with the spike in
2690 demand. Instead of dropping incoming requests, they are held in
2691 the LND until the RPC service can accept more requests.
2694 Frequency : Sometimes during replay
2696 Description: Assertion failure in ll_local_open after replay.
2697 Details : If replay happened on an open request reply before we were able
2698 to set replay handler, reply will become not swabbed tripping the
2699 assertion in ll_local_open. Now we set the handler right after
2700 recognising of open request
2703 Frequency : very rare
2705 Description: kernel reports "badness in vsnprintf"
2706 Details : Reading from the "recovery_status" /proc file in small chunks
2707 may cause a negative length in lprocfs_obd_rd_recovery_status()
2708 call to vsnprintf() (which is otherwise harmless). Exit early
2709 if there is no more space in the output buffer.
2711 Severity : enhancement
2713 Description: clear OBD RPC statistics by writing to them
2714 Details : It is now possible to clear the OBD RPC statistics by writing
2715 to the "stats" file.
2720 Description: Client mtime is not the same on different clients after utimes
2721 Details : In some cases, the client was using the utimes() syscall on
2722 a file cached on another node. The clients now validate the
2723 ctime from the MDS + OSTs to determine which one is right.
2728 Description: Inability to activate failout mode
2729 Details : lconf script incorrectly assumed that in pythong string's numeric
2730 value is used in comparisons.
2733 Frequency : always with multiple stripes per file
2735 Description: Inefficient object allocation for mutli-stripe files
2736 Details : When selecting which OSTs to stripe files over, for files with
2737 a stripe count that divides evenly into the number of OSTs,
2738 the MDS is always picking the same starting OST for each file.
2739 Return the OST selection heuristic to the original design.
2744 Description: mount failures may take full timeout to return an error
2745 Details : Under some heavy load conditions it is possible that a
2746 failed mount can wait for the full obd_timeout interval,
2747 possibly several minutes, before reporting an error.
2748 Instead return an error as soon as the status is known.
2750 Frequency : quota enabled and large files being deleted
2752 Description: releasing more than 4GB of quota at once hangs OST
2753 Details : If a user deletes more than 4GB of files on a single OST it
2754 will cause the OST to spin in an infinite loop. Release
2755 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2760 Description: statfs data retrieved from /proc may be stale or zero
2761 Details : When reading per-device statfs data from /proc, in the
2762 {kbytes,files}_{total,free,avail} files, it may appear
2763 as zero or be out of date.
2766 Frequency : systems with MD RAID1 external journal devices
2768 Description: lconf's call to blkid is confused by RAID1 journal devices
2769 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2770 of returning all block devices that match the journal UUID.
2773 Frequency : always, for aggregate stripe size over 4GB
2775 Description: assertion fails when trying to use 4GB stripe size
2776 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2777 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2780 Frequency : always on ppc64
2782 Description: the first write on an ext3 filesystem with mballoc got stuck
2783 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2784 perform endianness conversion.
2786 ------------------------------------------------------------------------------
2788 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2790 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2791 INTEROPERATE with older versions automatically. Please read the
2792 user documentation before upgrading any part of a live system.
2793 * WARNING: Lustre networking configuration changes are required with
2794 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2797 * Support for kernels:
2798 2.6.9-22.0.2.EL (RHEL 4)
2799 2.6.5-7.244 (SLES 9)
2800 2.6.12.6 vanilla (kernel.org)
2803 Severity : enhancement
2804 Bugzilla : 7981/8208
2805 Description: Introduced Lustre Networking (LNET)
2806 Details : LNET is new networking infrastructure for Lustre, it includes
2807 a reorganized network configuration mode (see the user
2808 documentation for full details) as well as support for routing
2809 between different network fabrics. Lustre Networking Devices
2810 (LNDs) for the supported network fabrics have also been
2811 created for this new infrastructure.
2813 Severity : enhancement
2814 Description: Introduced Access control lists
2815 Details : clients can set ACLs on files and directories in order to have
2816 more fine-grained permissions than the standard Unix UGO+RWX.
2817 The MDS must be started with the "-o acl" mount option.
2819 Severity : enhancement
2820 Description: Introduced filesystem quotas
2821 Details : Administrators may now establish per-user quotas on the
2824 Severity : enhancement
2826 Description: Configuration change for the XT3
2827 The PTLLND is now used to run Lustre over Portals on the XT3
2828 The configure option(s) --with-cray-portals are no longer used.
2829 Rather --with-portals=<path-to-portals-includes> is used to
2830 enable building on the XT3. In addition to enable XT3 specific
2831 features the option --enable-cray-xt3 must be used.
2836 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2837 Details : A race between allocating a new llcd and re-getting the llcd_lock
2838 allowed another thread to grab newly-allocated llcd.
2840 Severity : enhancement
2842 Description: 2.6 OST async journal commit and locking fix to improve performance
2843 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2844 2.6 kernels are now async as they already were in 2.4 kernels so
2845 that they can commit concurrently with the network bulk transfer.
2846 For block-allocated files the filter allocation semaphore is held
2847 to avoid filesystem fragmentation during allocation. BKL lock
2848 removed for 2.6 xattr operations where it is no longer needed.
2853 Description: lconf incorrectly determined whether two IP networks could talk
2854 Details : In some more complicated routing and multiple-network
2855 configurations, lconf will avoid trying to make a network
2856 connection to a disjoint part of the IP space. It was doing the
2857 math incorrectly for one set of cases.
2862 Description: Fix for potential infinite loop processing records in an llog.
2863 Details : If an llog record is corrupted/zeroed, it is possible to loop
2864 forever in llog_process(). Validate the llog record length
2865 and skip the remainder of the block on error.
2868 Frequency : occasional (liblustre only)
2870 Description: liblustre could not open files whose last component is a symlink
2871 Details : sysio_path_walk() would incorrectly pass the open intent to
2872 intermediate path components.
2875 Frequency : rare (liblustre only with non-standard tuning)
2876 Bugzilla : 7201 (7350)
2877 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2878 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2879 after it has been released, only do this for referenced locks
2881 Severity : enhancement
2883 Description: specifying an (invalid) directory default stripe_size of -1
2884 would reset the directory default striping
2885 Details : stripe_size -1 was used internally to signal directory stripe
2886 removal, now use "all default" to signal dir stripe removal
2887 as a directory striping of "all default" is not useful
2890 Frequency : common for large clusters running liblustre clients
2892 Description: doing an ls when liblustre clients are running is slow
2893 Details : sending a glimpse AST to a liblustre client waits for every AST
2894 to time out, as liblustre clients will not respond. Since they
2895 cannot cache data we refresh the OST lock LVB from disk instead.
2897 Severity : enhancement
2899 Description: doing an ls at the same time as file IO can be slow
2900 Details : enqueue and other "small" requests can be blocked behind many
2901 large IO requests. Create a new OST IO portal for non-IO
2902 requests so they can be processed faster.
2905 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2907 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2908 created file with mode 000
2909 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2910 allow an ftruncate() as a non-root user to a file with mode 000.
2911 Lustre now respects this flag to disable mode checks when
2912 truncating a file owned by the user
2915 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2918 Description: Revoking locks from clients that went dead or catatonic might take
2920 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2921 cancellation of such locks instant on servers without waiting for
2922 any reply from clients. Clients drops these locks when cancel
2923 notification from server is received without replying.
2926 Frequency : liblustre-only, when liblustre client dies or becomes busy
2928 Description: Doing ls on Linux clients can take a long time with active
2930 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2931 granting such locks to it in the first place if possible. Locks
2932 are taken by proxy on the OST during the read or write and
2933 dropped immediately afterward. Add connect flags handling, do
2934 not grant locks to liblustre clients for glimpse ASTs.
2936 Severity : enhancement
2938 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2939 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2940 sequential reads and behaves badly with applications doing random
2941 reads. Improve it by reading ahead at least read region, and
2942 avoiding excessive large RPC for small reads.
2944 Severity : enhancement
2946 Description: Creating more than 1000 files for a single job may cause a load
2947 imbalance on the OSTs if there are also a large number of OSTs.
2948 Details : qos_prep_create() uses an OST index reseed value that is an
2949 even multiple of the number of available OSTs so that if the
2950 reseed happens in the middle of the object allocation it will
2951 still utilize the OSTs as uniformly as possible.
2956 Description: OST or MDS may oops in ping_evictor_main()
2957 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2958 but doesn't restart at beginning of obd_exports_timed list
2961 Severity : enhancement
2963 Description: improve by-nid export eviction on the MDS and OST
2964 Details : allow multiple exports with the same NID to be evicted at one
2965 time without re-searching the exports list.
2968 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2970 Description: MDS may oops in groups_free()
2971 Details : in rare race conditions a newly allocated group_info struct is
2972 freed again, and this can be NULL. The 2.4 compatibility code
2973 for groups_free() checked for a NULL pointer, but 2.6 did not.
2976 Frequency : common for liblustre clients doing little filesystem IO
2977 Bugzilla : 9352, 7313
2978 Description: server may evict liblustre clients accessing contended locks
2979 Details : if a client is granted a lock or receives a completion AST
2980 with a blocking AST already set it would not reply to the AST
2981 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2984 Frequency : lfs setstripe, only systems with more than 160 OSTs
2986 Description: unable to set striping with a starting offset beyond OST 160
2987 Details : llapi_create_file() incorrectly limited the starting stripe
2988 index to the maximum single-file stripe count.
2991 Frequency : LDAP users only
2993 Description: lconf did not handle in-kernel recovery with LDAP properly
2994 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2996 Severity : enhancement
2998 Description: bind OST threads to NUMA nodes to improve performance
2999 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3000 node and do their allocations there to localize memory access
3002 Severity : enhancement
3004 Description: llmount can determine client NID directly from Myrinet (GM)
3005 Details : the client NID code from gmnalnid was moved directly into
3006 llmount, removing the need to use this or specifying the
3007 client NID explicitly when mounting GM clients with zeroconf
3010 Frequency : if client is started with down MDS
3012 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3013 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3014 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3015 ptlrpc_import_delayed_req() because we didn't check if the
3016 request was interrupted, and we also didn't break out of the
3017 event loop if there was no timeout
3022 Description: data loss during non-page-aligned writes to a single file from
3023 both multiple nodes and multiple threads on one node at same time
3024 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3025 inconsistency led to false short-reads, that were cached and later
3026 used by ->prepare_write() to fill in partially written page,
3027 leading to data loss.
3030 Frequency : always, if lconf --abort_recovery used
3032 Description: lconf --abort_recovery fails with 'Operation not supported'
3033 Details : lconf was attempting to abort recovery on the MDT device and not
3036 Severity : enhancement
3038 Description: remove cleanup logs
3039 Details : replace lconf-generated cleanup logs with lustre internal
3040 cleanup routines. Eliminates the need for client-cleanup and
3043 Severity : enhancement
3045 Description: add support for EAs (user and system) on lustre filesystems
3046 Details : it is now possible to store extended attributes in the Lustre
3047 client filesystem, and with the user_xattr mount option it
3048 is possible to allow users to store EAs on their files also
3050 Severity : enhancement
3052 Description: Add possibility (config option) to show minimal available OST free
3054 Details : When compiled with --enable-mindf configure option, statfs(2)
3055 (and so, df) will return least minimal free space available from
3056 all OSTs as amount of free space on FS, instead of summary of
3057 free spaces of all OSTs.
3059 Severity : enhancement
3061 Description: do not expand extent locks acquired on OST-side
3062 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3063 by server: they are not cached anyway.
3066 Frequency : when mmap is used/binaries executed from Lustre
3068 Description: Unmmap pages before throwing them away from read cache.
3069 Details : llap_shrink cache now attempts to unmap pages before discarding
3070 them (if unmapping failed - do not discard). SLES9 kernel has
3071 extra checks that trigger if this unmapping is not done first.
3076 Description: lconf didn't resolve symlinks before checking to see whether a
3077 given mountpoint was already in use
3080 Frequency : when migrating failover services
3081 Bugzilla : 6395, 9514
3082 Description: When migrating a subset of services from a node (e.g. failback
3083 from a failover service node) the remaining services would
3084 time out and evict clients.
3085 Details : lconf --force (implied by --failover) sets the global obd_timeout
3086 to 5 seconds in order to quickly disconnect, but this caused
3087 other RPCs to time out too quickly. Do not change the global
3088 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3090 Severity : enhancement
3091 Frequency : if MDS is started with down OST
3092 Bugzilla : 9439,5706
3093 Description: Allow startup/shutdown of an MDS without depending on the
3094 availability of the OSTs.
3095 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3096 Add appropriate locking and lov-osc refcounts for safe
3097 cleaning. Add osc abort_inflight calls in case the
3098 synchronize never started.
3101 Frequency : occasional (Cray XT3 only)
3103 Description: root not authorized to access files in CRAY_PORTALS environment
3104 Details : The client process capabilities were not honoured on the MDS in
3105 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3106 been accessed by an authorized user then root was able to access
3107 the file on the local client also. The root user capabilities
3108 are now allowed on the MDS, as this environment has secure UID.
3111 Frequency : occasional
3113 Description: ldiskfs "too long searching" message happens too often
3114 Details : A debugging message (otherwise harmless) prints too often on
3115 the OST console. This has been reduced to only happen when
3116 there are fragmentation problems on the filesystem.
3121 Description: Division by zero in statfs when all OSCs are inactive
3122 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3123 lov_statfs divides by value returned by lov_get_stripecnt().
3127 Bugzilla : 9489, 3273
3128 Description: First write from each client to each OST was only 4kB in size,
3129 to initialize client writeback cache, which caused sub-optimal
3130 RPCs and poor layout on disk for the first writen file.
3131 Details : Clients now request an initial cache grant at (re)connect time
3132 and so that they can start streaming writes to the cache right
3133 away and always do full-sized RPCs if there is enough data.
3134 If the OST is rebooted the client also re-establishes its grant
3135 so that client cached writes will be honoured under the grant.
3140 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3141 Details : Now I/O RPCs go to different portal number and (presumably) fast
3142 lock requests (and glimses) and other RPCs get their own service
3143 threads pool that should be able to service those RPCs
3146 Severity : enhancement
3148 Description: Ability to exchange lustre version between client and servers and
3149 issue warnings at client side if client is too old. Also for
3150 liblustre clients there is ability to refuse connection of too old
3152 Details : New 'version' field is added to connect data structure that is
3153 filled with version info. That info is later checked by server and
3157 Frequency : rare, liblustre only.
3158 Bugzilla : 9296, 9581
3159 Description: Two simultaneous writes from liblustre at offset within same page
3160 might proceed at the same time overwriting eachother with stale
3162 Details : I/O lock withing llu_file_prwv was released too early, before data
3163 actually was hitting the wire. Extended lock-holding time until
3164 server acknowledges receiving data.
3167 Frequency : extremely rare. Never observed in practice.
3169 Description: avoid generating lustre_handle cookie of 0.
3170 Details : class_handle_hash() generates handle cookies by incrementing
3171 global counter, and can hit 0 occasionaly (this is unlikely, but
3172 not impossible, because initial value of cookie counter is
3173 selected randonly). Value of 0 is used as a sentinel meaning
3174 "unassigned handle" --- avoid it. Also coalesce two critical
3175 sections in this function into one.
3177 Severity : enhancement
3179 Description: allow liblustre clients to delegate truncate locking to OST
3180 Details : To avoid overhead of locking, liblustre client instructs OST to
3181 take extent lock in ost_punch() on client's behalf. New connection
3182 flag is added to handle backward compatibility.
3184 Severity : enhancement
3185 Bugzilla : 4928, 7341, 9758
3186 Description: allow number of OST service threads to be specified
3187 Details : a module parameter allows the number of OST service threads
3188 to be specified via "options ost ost_num_threads={N}" in the
3189 OSS's /etc/modules.conf or /etc/modprobe.conf.
3193 Bugzilla : 6146, 9635, 9895
3194 Description: servers crash with bad pointer in target_handle_connect()
3195 Details : In rare cases when a client is reconnecting it was possible that
3196 the connection request was the last reference for that export.
3197 We would temporarily drop the export reference and get a new
3198 one, but this may have been the last reference and the export
3199 was just destroyed. Get new reference before dropping old one.
3201 Severity : enhancement
3202 Frequency : if client is started with failover MDS
3204 Description: Allow multiple MDS hostnames in the mount command
3205 Details : Try to read the configuration from all specified MDS
3206 hostnames during a client mount in case the "primary"
3209 Severity : enhancement
3211 Description: Stop sending data to evicted clients as soon as possible.
3212 Details : Check if the client we are about to send or are sending data to
3213 was evicted already. (Check is done every second of waiting,
3214 for which l_wait_event interface was extended to allow checking
3215 of exit condition at specified intervals).
3218 Frequency : rare, normally only when NFS exporting is done from client
3220 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3222 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3225 Frequency : always, if srand() is called before liblustre initialization
3227 Description: Liblustre uses system PRNG disturbing its usage by user application
3228 Details : Introduce internal to lustre fast and high-quality PRNG for
3229 lustre usage and make liblustre and some other places in generic
3230 lustre code to use it.
3232 Severity : enhancement
3233 Bugzilla : 9477, 9557, 9870
3234 Description: Verify that the MDS configuration logs are updated when xml is
3235 Details : Check if the .xml configuration logs are newer than the config
3236 logs stored on the MDS and report an error if this is the case.
3237 Request --write-conf, or allow starting with --old_conf.
3239 Severity : enhancement
3241 Description: Handle symlinks in the path when checking if Lustre is mounted.
3242 Details : Resolve intermediate symlinks when checking if a client has
3243 mounted a filesystem to avoid duplicate client mounts.
3248 Description: lconf can hit an error exception but still return success.
3249 Details : The lconf command catches the Command error exception at the top
3250 level script context and will exit with the associated exit
3251 status, but doesn't ensure that this exit status is non-zero.
3256 Description: failure of ptlrpc thread startup can cause oops
3257 Details : Starting a ptlrpc service thread can fail if there are a large
3258 number of threads or the server memory is very fragmented.
3259 Handle this without oopsing.
3262 Frequency : always, only if liblustre and non-default acceptor port was used
3264 Description: liblustre cannot connect to servers with non-default acceptor port
3265 Details : tcpnal_set_default_params() was not called and was therefore
3266 ignoring the environment varaible TCPNAL_PORT, as well as other
3267 TCPNAL_ environment variables
3272 Description: two objects could be created on the same OST for a single file
3273 Details : If an OST is down, in some cases it was possible to create two
3274 objects on a single OST for a single file. No problems other
3275 than potential performance impact and spurious error messages.
3279 Bugzilla : 5681, 9562
3280 Description: Client may oops in ll_unhash_aliases
3281 Details : Client dcache may become inconsistent in race condition.
3282 In some cases "getcwd" can fail if the current directory is
3288 Description: Inode refcounting problems in NFS export code
3289 Details : link_raw functions used to call d_instantiate without obtaining
3290 extra inode reference first.
3294 Bugzilla : 9942, 9903
3295 Description: Referencing freed requests leading to crash, memleaks with NFS.
3296 Details : We used to require that call to ll_revalidate_it was always
3297 followed by ll_lookup_it. Also with revalidate_special() it is
3298 possible to call ll_revalidate_it() twice for the same dentry
3299 even if first occurence returned success. This fix changes semantic
3300 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3301 reference on a request referred from the intent.
3302 ll_intent_release() then releases such a request.
3305 Frequency : rare, normally benchmark loads only
3307 Description: unlinked inodes were kept in memory on the client
3308 Details : If a client is repeatedly creating and unlinking files it
3309 can accumulate a lot of stale inodes in the inode slab cache.
3310 If there is no other client load running this can cause the
3311 client node to run out of memory. Instead flush old inodes
3312 from client cache that have the same inode number as a new inode.
3315 Frequency : SLES9 2.6.5 kernel and long filenames only
3316 Bugzilla : 9969, 10379
3317 Description: utime reports stale NFS file handle
3318 Details : SLES9 uses out-of-dentry names in some cases, which confused
3319 the lustre dentry revalidation. Change it to always use the
3323 Frequency : rare, unless heavy write-truncate concurrency is continuous
3324 Bugzilla : 4180, 6984, 7171, 9963, 9331
3325 Description: OST becomes very slow and/or deadlocked during object unlink
3326 Details : filter_destroy() was holding onto the parent directory lock
3327 while truncating+unlinking objects. For very large objects this
3328 may block other threads for a long time and slow overall OST
3329 responsiveness. It may also be possible to get a lock ordering
3330 deadlock in this case, or run out of journal credits because of
3331 the combined truncate+unlink. Solution is to do object truncate
3332 first in one transaction without parent lock, and then do the
3333 final unlink in a new transaction with the parent lock. This
3334 reduces the lock hold time dramatically.
3337 Frequency : rare, 2.4 kernels only
3339 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3340 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3341 before unmounting their filesystem in order to avoid deadlock.
3342 The kernel_locked() function in 2.4 kernels only checks whether
3343 the kernel lock is held, not whether it is this process that is
3344 holding it as 2.6 kernels do.
3349 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3350 Details : The client ptlrpc code may be trying to reconnect to a down
3351 server before a previous connection attempt has timed out.
3352 Increase the reconnect interval to be longer than the connection
3353 timeout interval to avoid sending duplicate connections to
3357 Frequency : echo_client brw_test command
3359 Description: fix echo_client to work with OST preallocated code
3360 Details : OST preallocation code (5137) didn't take echo_client IO path
3361 into account: echo_client calls filter methods outside of any
3362 OST thread and, hence, there is no per-thread preallocated
3363 pages and buffers to use. Solution: hijack pga pages for IO. As
3364 a byproduct, this avoids unnecessary data copying.
3368 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3369 Description: Client can oops in mdc_commit_close() after open replay
3370 Details : It was possible for the MDS to return an open request with no
3371 transaction number in mds_finish_transno() if the client was
3372 evicted, but without actually returning an error. Clients
3373 would later try to replay that open and may trip an assertion
3374 Simplify the client close codepath, and always return an error
3375 from the MDS in case the open is not successful.
3378 Frequency : rare, 2.6 OSTs only
3380 Description: OST may deadlock under high load on fragmented files
3381 Details : If there was a heavy load and highly-fragmented OST filesystems
3382 it was possible to have all the OST threads deadlock waiting on
3383 allocation of biovecs, because the biovecs were not released
3384 until the entire RPC IO was completed. Instead, release biovecs
3385 as soon as they are complete to ensure forward IO progress.
3387 Severity : enhancement
3389 Description: Support for specifying external journal device at mount
3390 Details : If an OST or MDS device is formatted with an external journal
3391 device, this device major/minor is stored in the ext3 superblock
3392 and may not be valid for failover. Allow detecting and
3393 specifying the external journal at mount time.
3398 Description: Mounting an MDS with pending unlinked files may cause oops
3399 Details : target_finish_recovery() calls mds_postrecov() which returned
3400 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3401 considers this an error and immediately begins cleaning up the
3402 lov, just after starting the mds_lov process
3404 Severity : enhancement
3406 Description: Implement 'lfs df' to report actual free space on per-OST basis
3407 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3408 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3409 usage of objects; '-h' to report in human readable format.
3411 ------------------------------------------------------------------------------
3413 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3420 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3421 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3422 at startup during orphan unlinking. Instead, skip buddy update
3423 before setup, as it will be regenerated after recovery is complete.
3428 Description: If an OST is inactive, its locks might reference stale inodes.
3429 Details : lov_change_cbdata() must iterate over all namespaces, even if
3430 they are inactive to clear inode references from the lock.
3432 Severity : enhancement
3433 Frequency : occasional, if non-standard max_dirty_mb used
3435 Description: Client will block write RPCs if not enough grant
3436 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3437 then the client will block writes while waiting for another RPC
3438 to complete instead of consuming its dirty limit. With change
3439 we get improved performance when max_dirty_mb is small.
3441 Severity : enhancement
3442 Bugzilla : 3389, 6253
3443 Description: Add support for supplementary groups on the MDS.
3444 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3445 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3446 lookups for user supplementary groups into a cache.
3450 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3451 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3452 for liblustre O_EXCL is enforced by MDS.
3456 Description: readdir never returns NULL in liblustre.
3457 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3458 dentry in struct dirent.
3462 Frequency : liblustre only, depends on application IO pattern
3463 Description: liblustre clients evicted if not contacting servers
3464 Details : Don't put liblustre clients into the ping_evictor list, so
3465 they will not be evicted by the pinger ever.
3467 Severity : enhancement
3469 Description: Add ability to evict clients by NID from MDS.
3470 Details : By echoing "nid:$NID" string into
3471 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3472 $NID would be instantly evicted from this MDS and from all active
3473 OSTs connected to it.
3477 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3478 Details : lookup_it_finish() used to query file size from OSTs that was not
3483 Description: service threads change working directory to that of init
3484 Details : Starting lustre service threads may pin the working directory
3485 of the parent thread, making that filesystem busy. Threads
3486 now change to the working directory of init to avoid this.
3490 Frequency : during shutdown only
3491 Description: shutdown with a failed MDS or OST can cause unmount to hang
3492 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3497 Frequency : 2.6 only
3498 Description: chmod/chown may include an extra supplementary group
3499 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3500 supplementary group and if none is specified this is used.
3503 Bugzilla : 5479 (6816)
3505 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3506 Details : If lookup is in progress on a file that is unlinked we might try
3507 to revalidate the inode and fail in revalidate after lookup is
3508 complete and ll_file_open() enqueues the open again but
3509 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3512 Frequency : always, if lconf --abort_recovery used
3514 Description: lconf --abort_recovery fails with 'Operation not supported'
3515 Details : lconf was attempting to abort recovery on the MDT device and not
3518 ------------------------------------------------------------------------------
3520 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3525 Frequency : rare (only unsupported configurations with a node running as an
3527 Bugzilla : 6514, 5137
3528 Description: Mounting a Lustre file system on a node running as an OST could
3530 Details : OSTs now preallocates memory needed to write out data at
3531 startup, instead of when needed, to avoid having to
3532 allocate memory in possibly low memory situations.
3533 Specifically, if the file system is mounted on on OST,
3534 memory pressure could force it to try to write out data,
3535 which it needed to allocate memory to do. Due to the low
3536 memory, it would be unable to do so and the node would
3537 become unresponsive.
3539 Severity : enhancement
3541 Description: Addition of lconf --service command line option
3542 Details : lconf now accepts a '--service <arg>' option, which is
3543 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3545 Severity : enhancement
3547 Description: Failover mode is now the default for OSTs.
3548 Details : By default, OSTs will now run in failover mode. To return to
3549 the old behaviour, add '--failout' to the lmc line for OSTs.
3551 Severity : enhancement
3553 Description: Health checks are now provided for MDS and OSTs
3554 Details : Additional detailed health check information on MSD and OSTs
3555 is now provided through the procfs health_check value.
3558 Frequency : occasional, depends on IO load
3560 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3561 after numerous create/delete cycles
3562 Details : The ext3 inode allocation policy would not allocate new inodes
3563 very well on the OSTs because there are no new directories
3564 being created. Instead we look for groups with free space if
3565 the parent directories are nearly full.
3570 Description: Network or server problems during mount may cause partially
3571 mounted clients instead of returning an error.
3572 Details : The config llog parsing code may overwrite the error return
3573 code during mount error handling, returning success instead
3579 Description: MDS can fail to allocate large reply buffers
3580 Details : After long uptimes the MDS can fail to allocate large reply
3581 buffers (e.g. zconf client mount config records) due to memory
3582 fragmentation or consumption by the buffer cache. Preallocate
3583 some large reply buffers so that these replies can be sent even
3584 under memory pressure.
3588 Frequency : rare (liblustre)
3589 Description: fsx running with liblustre complained that using truncate() to
3590 extend the file doesn't work. This patch corrects that issue.
3591 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3592 ATTR_SIZE and lsm use in llu_setattr_raw.
3596 Frequency : rare, only 2.6 kernels
3597 Description: Unusual file access patterns on the MDS may result in inode
3598 data being lost in very rare circumstances.
3599 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3600 code in the 2.6 kernel caused the inode and/or EA data not to
3601 be read from disk, causing single-file corruption.
3605 Frequency : rare, only 2.6 filesystems using extents
3606 Description: Heavy concurrent write and delete load may cause data corruption.
3607 Details : It was possible under high-load situations to have an extent
3608 metadata block in the block device cache from a just-unlinked
3609 file overwrite a newly-allocated data block. We now unmap any
3610 metadata buffers that alias just-allocated data blocks.
3614 Frequency : filesystems with default stripe_count larger than 77
3615 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3616 Details : lconf specifies an inode size of 4096 bytes when the default
3617 stripe_count is larger than 77. This conflicts with the default
3618 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3619 this case to avoid pinning too much memory for large EAs.
3621 ------------------------------------------------------------------------------
3623 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3628 Frequency : rare (extremely heavy IO load with hundreds of clients)
3630 Description: Client is evicted, gets IO error writing to file
3631 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3632 caused clients to be evicted for AST timeouts. The fixes in
3633 bug 5192 mean we no longer need to have such short AST timeouts
3634 so ldlm_timeout has been increased.
3637 Frequency : occasional during --force or --failover shutdown under load
3638 Bugzilla : 5949, 4834
3639 Description: Server oops/LBUG if stopped with --force or --failover under load
3640 Details : a collection of import/export refcount and cleanup ordering
3641 issues fixed for safer force cleanup
3644 Frequency : only filesystems larger than 120 OSTs
3645 Bugzilla : 5990, 6223
3646 Description: lfs getstripe would oops on a very large filesystem
3647 Details : lov_getconfig used kfree on vmalloc'd memory
3650 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3651 Bugzilla : 6242, 6243
3652 Description: reading from files that had been truncated to a non-zero size
3653 but never opened returned no data
3654 Details : ll_file_read() reads zeros from no-object files to EOF
3659 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3660 Details : There's little harm in aborting MDS/OSS recovery and letting it
3661 try again, so I removed the LASSERT and return an error instead.
3663 Severity : enhancement
3665 Description: New debugging infrastructure for tracking down data corruption
3666 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3667 (b) cover more of the client-side code path, and (c) try to narrow
3668 down where problems occurred
3672 Bugzilla : 3819, 4364, 4397, 6313
3673 Description: Racing close and eviction MDS could cause assertion in mds_close
3674 Details : It was possible to get multiple mfd references during close and
3675 client eviction, leading to one thread referencing a freed mfd.
3677 Severity: : enhancement
3678 Bugzilla : 3262, 6359
3679 Description: Attempts to reconnect to servers are now more aggressive.
3680 Details : This builds on the enhanced upcall-less recovery that was added
3681 in 1.4.2. When trying to reconnect to servers, clients will
3682 now try each server in the failover group every 10 seconds. By
3683 default, clients would previously try one server every 25 seconds.
3688 Description: After recovery, certain operations trigger a failed
3689 assertion on a client.
3690 Details : Failing over an mds, using lconf -d --failover, while a
3691 client was doing a readdir() call would cause the client to
3692 LBUG after recovery completed and the readdir() was resent.
3694 Severity : enhancement
3696 Description: Default groups are now added by lconf
3697 Details : You can now run lconf --group <servicename> without having to
3698 manually add groups with lmc.
3701 Frequency : occasional
3703 Description: Nodes with an elan id of 0 trigger a failed assertion
3706 Frequency : always when accessing e.g. tty/console device nodes
3708 Description: tty and some other devices nodes cannot be used on lustre
3709 Details : file's private_data field is used by device data and lustre
3710 values in there got lost. New field was added to struct file to
3711 store fs-specific private data.
3714 Frequency : when exporting Lustre via NFS
3716 Description: NFSD failed occasionally when looking up a path component
3717 Details : NFSD is looking up ".." which was broken in ext3 directories
3718 that had grown large enough to become hashed.
3721 Frequency : Clusters with multiple interfaces not on the same subnet
3723 Description: Nodes will repeatedly try to reconnect to an interface which it
3724 cannot reach and report an error to the log.
3725 Details : Extra peer list entries will be created by lconf with some peers
3726 unreachable. lconf now validates the peer before adding it.
3729 Frequency : Only if a default stripe is set on the filesystem root.
3731 Description: Setting a default stripe on the filesystem root prevented the
3732 filesystem from being remounted.
3733 Details : The client was sending extra request flags in the root getattr
3734 request and did not allocate a reply buffer for the dir EA.
3737 Frequency : occasional, higher if lots of files are accessed by one client
3738 Bugzilla : 6159, 6097
3739 Description: Client trips assertion regarding lsm mismatch/magic
3740 Details : While revalidating inodes the VFS looks up inodes with ifind()
3741 and in rare cases can find an inode that is being freed.
3742 The ll_test_inode() code will free the lsm during ifind()
3743 when it finds an existing inode and then the VFS later attaches
3744 this free lsm to a new inode.
3748 Bugzilla : 6422, 7030
3749 Description: MDS deadlock between mkdir and client eviction
3750 Details : Creating a new file via mkdir or mknod (starting a transaction
3751 and getting the ns lock) can deadlock with client eviction
3752 (gets ns lock and trying to finish a synchronous transaction).
3755 Frequency : occasional
3756 Description: While starting a server, the fsfilt_ext3 module could not be
3758 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3759 kernels. Previously, lconf would still use the ext3 name
3760 when trying to load modules. Now, it will correctly use
3761 ext3 on 2.4 and ldiskfs on 2.6.
3763 Severity : enhancement
3764 Description: The default stripe count has been changed to 1
3765 Details : The interpretation of the default stripe count (0, to lfs
3766 or lmc) has been changed to mean striping across a single
3767 OST, rather than all available. For general usage we have
3768 found a stripe count of 1 or 2 works best.
3770 Severity : enhancement
3771 Description: Add support for compiling against Cray portals.
3772 Details : Conditional compiling for some areas that are different
3776 Frequency : occasional
3777 Bugzilla : 6409, 6834
3778 Description: Creating files with an explicit stripe count may lead to
3779 a failed assertion on the MDS
3780 Details : If some OSTs are full or unavailable, creating files may
3781 trigger a failed assertion on the MDS. Now, Lustre will
3782 try to use other servers or return an error to the
3786 Frequency : occasional
3788 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3789 caused lock timeout during readahead (since 1.4.2).
3790 Details : Processes doing readahead might match a lock that hasn't been
3791 granted yet if there are overlapping and conflicting lock
3792 requests. The readahead process waits on ungranted lock
3793 (original lock is CBPENDING), while OST waits for that process
3794 to cancel CBPENDING read lock and eventually evicts client.
3796 Severity : enhancement
3798 Description: Initial enabling of flock support for clients
3799 Details : Implements fcntl advisory locking and file status functions.
3800 This feature is provided as an optional mount flag (default
3801 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3802 locking are implemented yet, and those that are are not guaranteed
3803 to be completely correct in production environments.
3804 mount -t lustre -o [flock|noflock] ...
3807 Frequency : occasional
3809 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3810 assertion in the ext3 JBD (journaling) layer.
3811 Details : The b_committed_data struct is protected by the big kernel lock
3812 in 2.4 kernels, serializing journal_commit_transaction() and
3813 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3814 access to this struct. In 2.6 kernels there is finer grained
3815 locking to improve SMP performance of the JBD layer.
3819 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3824 Description: merge and cleanup kernel patches.
3827 -----------------------------------------------------------------------------
3829 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3831 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3832 versions of Lustre. Please update all servers and clients to
3833 version 1.4.2 or later at the same time. You must also run
3834 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3835 to update the configuration logs.
3837 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3838 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3839 - swab llog records of type '0' so we get proper header size/idx (5861)
3840 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3841 - fix rename of one directory over another leaking an inode (5953)
3842 - avoid SetPageDirty on 2.6 (5981)
3843 - don't re-add just-being-destroyed locks to the waiting list (5653)
3844 - when creating new directories, inherit the parent's custom
3845 striping settings if present parent (3048)
3846 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3847 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3848 - don't hold client locks on temporary worklist from l_lru (5666)
3849 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3850 - automatically evict dead clients (5921)
3851 - Update file size properly in create+truncate+fstat case (6196)
3852 - Do not unhash mountpoint dentries, do not allow removal of
3854 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3855 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3856 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3857 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3858 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3859 - Correctly handle reads of files with no objects (6243)
3860 - lctl recover will also mark a device active if deactivate used (5933)
3862 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3863 - allow --write-conf on an MDS with different nettype than client (5619)
3864 - don't write config llogs to MDS for mounts not from that MDS (5617)
3865 - lconf should create multiple TCP connections from a client (5201)
3866 - init scripts are now turned off by default; run chkconfig --on
3867 lustre and chkconfig --on lustrefs to use them
3868 - upcalls are no longer needed for clients to recover to failover
3870 - add --abort-recovery option to lconf to abort recovery on device
3872 - add support for an arbitrary number of OSTs (3026)
3873 - Quota support protocol changes.
3874 - forward compatibility changes to wire structs (6007)
3875 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3876 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3877 - improved "lustre status" script
3878 - initialize blocksize for non-regular files (6062)
3879 - added --disable-server and --disable-client configure options (5782)
3880 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3881 - Vanilla 2.4.29 support
3882 - increase maximum number of obd devices to 520 (6242)
3883 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3884 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3885 SLES 9 2.6.5 kernels
3886 - limit stripes per file to 160 (the maximum EA size) (6093)
3888 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3891 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3892 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3893 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3894 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3895 - small ext3 extents cleanups and fixes (5733)
3896 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3897 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3898 - avoid needless client->OST connect, fix handle mismatch (5317)
3899 - fix DLM error path that led to out-of-sync client, long delays (5779)
3900 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3901 - fix several locking issues related to i_size (5492,5624,5654,5672)
3902 - don't move pending lock onto export if it is already evicted (5683)
3903 - fix kernel oops when creating .foo in unlinked directory (5548)
3904 - fix deadlock in obdfilter statistics vs. object create (5811)
3905 - use time_{before,after} to avoid timer jiffies wrap (5882)
3906 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3907 - Do not leak request if server was not able to process it (5154)
3908 - If mds_open unable to find parent dir, make that negative lookup(5154)
3909 - don't create new directories with extent-mapping (5909, 5936)
3911 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3912 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3913 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3914 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3915 - support for RHEL4 kernel on i686 (5773)
3916 - provide error messages when incompatible logs are encountered (5898)
3918 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3919 * version 1.4.0.10 (1.4.1 release candidate 1)
3921 - don't keep a lock reference when lock is not granted (4238)
3922 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3923 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3924 - drop import inflight refcount on signal_completed_replay error (5255)
3925 - unlock page after async write error during send (3677)
3926 - handle missing objects in filter_preprw_read properly (5265)
3927 - no transno return for symlink open, don't save no-trasno open (3440)
3928 - don't try to complete elan receive that already failed (4012)
3929 - free RPC server reply state on error (5406)
3930 - clean up thread from ptlrpc_start_thread() on error (5160)
3931 - readahead could read extra page into cache that wasn't ejected (5388)
3932 - prevent races in class_attach/setup/cleanup/detach (5260)
3933 - don't dereference de->d_inode after l_dput of de (5458)
3934 - use "int" for stripe value returned from lock_to_stripe (5544)
3935 - mballoc allocation and error-checking fixes in 2.6 (5504)
3936 - block device patches to fix I/O request sizes in 2.6 (5482)
3937 - look up hostnames for IB nals (5602)
3938 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3939 - don't start multiple acceptors for the same port (5277)
3940 - fix incorrect LASSERT in mds_getattr_name (5635)
3941 - export a proc file for general "ping" checking (5628)
3942 - fix "lfs check" to not block when the MDS is down (5628)
3944 - service request history (4965)
3945 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3946 - create an "evict_client" /proc entry on OSTs, like the MDS has
3947 - fix mount usage message, return errors per mount(8) (5168)
3948 - change grep [] to grep "[]" in tests so they work in more UMLs
3949 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3950 - remove ancient LOV_MAGIC_V0 EA support (5047)
3951 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3952 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3953 - allow readahead allocations to fail when low on memory (5383)
3954 - mmap locking landed again, after considerable improvement (2828)
3955 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3957 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3960 - send OST transaction number in read/write reply to free req (4966)
3961 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3962 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3963 - the watchdog thread now runs as interruptible (5246)
3964 - flock/lockf fixes (but it's still disabled, pending 5135)
3965 - don't use EXT3 constants in llite code (5094)
3966 - memory shortage at startup could cause assertion (5176)
3968 - reorganization of lov code
3969 - single portals codebase
3971 - add extents/mballoc support (5025)
3972 - direct I/O reads in the obdfilter (4048)
3973 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3975 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3978 - send OST transaction number in read/write reply to free req (4966)
3979 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3980 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3981 - the watchdog thread now runs as interruptible (5246)
3982 - handle missing objects in filter_preprw_read properly (5265)
3983 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3984 - drop import inflight refcount on signal_completed_replay error (5255)
3985 - unlock page after async write error during send (3677)
3986 - return original error code on reconstructed replies (3761)
3987 - no transno return for symlink open, don't save no-trasno open (3440)
3989 - add pid to ldlm debugging output (4922)
3990 - bump the watchdog timeouts -- we can't handle 30sec yet
3991 - extra debugging for orphan dentry/inode bug (5259)
3993 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3996 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3997 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3998 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3999 - fix NULL dereference in /proc/sys/portals/routes (4827)
4000 - allow failed mdc_close() operations to be interrupted (4561)
4001 - stop precreate on OST before MDS would time out on it (4778)
4002 - don't send partial-page writes before EOF from client (4410)
4003 - discard client grant for sub-page writes on large-page clients (4520)
4004 - don't free dentries not owned by NFS code, check generation (4806)
4005 - fix lsm leak if mds_create_objects() fails (4801)
4006 - limit debug_daemon file size, always print CERROR messages (4789)
4007 - use transno after validating reply (3892)
4008 - process timed out requests if import state changes (3754)
4009 - update mtime on OST during writes, return in glimpse (4829)
4010 - add mkfsoptions to LDAP (4679)
4011 - use ->max_readahead method instead of zapping global ra (5039)
4012 - don't interrupt __l_wait_event() during strace
4014 - add software watchdogs to catch hung threads quickly (4941)
4015 - make lustrefs init script start after nfs is mounted
4016 - fix CWARN/ERROR duplication (4930)
4017 - return async write errors to application if possible (2248)
4018 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4019 - print NAL number in %x format (4645)
4020 - update barely-supported suse-2.4.21-171 series (4842)
4021 - support for sles 9 %post scripts
4022 - support for building 2.6 kernel-source packages
4023 - support for sles km_* packages
4025 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4028 - ignore -ENOENT errors in osc_destroy (3639)
4029 - notify osc create thread that OSC is being cleaned up (4600)
4030 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4031 - reconstruct ost_handle() like mds_handle() (4657)
4032 - create a new thread to do import eviction to avoid deadlock (3969)
4033 - let lconf resolve symlinked-to devices (4629)
4034 - don't unlink "objects" from directory with default EA (4554)
4035 - hold socknal file ref over connect in case target is down (4394)
4036 - allow more than 32000 subdirectories in a single directory (3244)
4037 - fix blocks count for O_DIRECT writes (3751)
4038 - OST returns ENOSPC from object create when no space left (4539)
4039 - don't send truncate RPC if file size isn't changing (4410)
4040 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4041 - bind to privileged port in socknal and tcpnal (3689)
4043 - rate limit CERROR/CWARN console message to avoid overload (4519)
4044 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4045 - basic mmap support (3918)
4046 - kernel patch series update from b1_4 (4711)
4048 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4051 - avoid crash during MDS cleanup with OST shut down (2775)
4052 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4053 - don't use bad inodes on the MDS (3744)
4054 - dynamic object preallocation to improve recovery speed (4236)
4055 - don't hold spinlock over lock dumping or change debug flags (4401)
4056 - don't zero obd_dev when it is force cleaned (3651)
4057 - print grants to console if they go negative (4431)
4058 - "lctl deactivate" will stop automatic recovery attempts (3406)
4059 - look for existing locks in ldlm_handle_enqueue() (3764)
4060 - don't resolve lock handle twice in recovery avoiding race (4401)
4061 - revalidate should check working dir is a directory (4134)
4063 - don't always mark "slow" obdfilter messages as errors (4418)
4065 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4068 - don't close LustreDB during write_conf until it is done (3860)
4069 - fix typo in lconf for_each_profile (3821)
4070 - allow dumping logs from multiple threads at one time (3820)
4071 - don't allow multiple threads in OSC recovery (3812)
4072 - fix debug_size parameters (3864)
4073 - fix mds_postrecov to initialize import for llog ctxt (3121)
4074 - replace config semaphore with spinlock (3306)
4075 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4076 - don't allow enqueue to complete on a destroyed export (3822)
4077 - down write_lock before checking llog header bitmap (3825)
4078 - recover from lock replay timeout (3764)
4079 - up llog sem before sending rpc (3652)
4080 - reduce ns lock hold times when setting kms (3267)
4081 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4082 - fix NULL deref and obd_dev leak on setup error (3312)
4083 - replace some LBUG about llog ops with error handling (3841)
4084 - don't match INVALID dentries from d_lookup and spin (3784)
4085 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4086 - fix invalid assertion in ptlrpc_set_wait (3880)
4088 - add libwrap support for the TCP acceptor (3996)
4089 - add /proc/sys/portals/routes for non-root route listing (3994)
4090 - allow setting MDS UUID in .xml (2580)
4091 - print the stack of a process that LBUGs (4228)
4093 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4096 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4097 - make sure to unset replay flag from failed open requests (3440)
4098 - if default stripe count is 0, use OST count for inode size (3636)
4099 - update parent mtime/ctime on client for create/unlink (2611)
4100 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4101 - free recovery state on server during a forced cleanup (3571)
4102 - unregister_reply for resent reqs (3063)
4103 - loop back devices mounting and status check on 2.6 (3563)
4104 - fix resource-creation race that can provoke i_size == 0 (3513)
4105 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4106 - more debugging for page-accounting assertion (3746)
4107 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4108 - avoid deadlock after precreation failure (3758)
4109 - fix race and lock order deadlock in orphan handling (3450, 3750)
4110 - add validity checks when grabbing inodes from l_ast_data (3599)
4112 - add /proc/.../recovery_status to obdfilter (3428)
4113 - lightweight CDEBUG infrastructure, debug daemon (3668)
4114 - change default OSC RPC parameters to be better on small clusters
4115 - turn off OST read cache for files smaller than 32MB
4116 - install man pages and include them in rpms (3100)
4117 - add new init script for (un)mounting lustre filesystems (2593)
4118 - run chkconfig in %post for init scripts (3701)
4119 - drop scimac NAL (unmaintained)
4121 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4124 - clean kiobufs before and after use (3485)
4125 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4126 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4127 - change init script to fail more clearly if not run as root (1528)
4128 - allow clients to reconnect during replay (1742)
4129 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4130 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4131 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4132 - fix cond_resched (3554)
4133 - only evict unfinished clients after recovery (3515)
4134 - allow bulk resend, prevent data loss (3570)
4135 - dynamic ptlrpc request buffer allocation (2102)
4136 - don't allow unlinking open directory if it isn't empty (2904)
4137 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4138 - remove extraneous obd dereference causing LASSERT failure (3334)
4139 - don't use get_cycles() when creating temp. files on the mds (3156)
4140 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4141 - handle EEXIST for set-stripe, set proper directory name (3336)
4143 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4144 - fix ksocknal_fmb_callback() error messages (2918)
4146 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4149 - don't copy lvb into (possibly NULL) reply on error (2983)
4150 - don't deref dentry after dput, don't free lvb on error (2922)
4151 - use the kms to determine writeback rpc length (2947)
4152 - increment oti_logcookies when osc is inactive (2948)
4153 - update client's i_blocks count via lvb messages (2543)
4154 - handle intent open/close of special files properly (1557)
4155 - mount MDS with errors=remount-ro, like obdfilter (2009)
4156 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4157 - don't use cancelling-locks' kms values (2947)
4158 - use highest lock extent for kms, not last one (2925)
4159 - don't dereference ERR_PTR() dentry in error handling path (3107)
4160 - fix thread race in portals_debug_dumplog() (3122)
4161 - create lprocfs device entries at setup instead of at attach (1519)
4162 - common AST error handler, don't evict client on completion race (3145)
4163 - zero nameidata in detach_mnt in 2.6 (3118)
4164 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4165 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4166 - initialize RPC timeout timer earlier for 2.6 (3219)
4167 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4168 - print nal/nid for unknown nid (3258)
4169 - additional checks for oscc recovery before doing precreate (3284)
4170 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4171 - don't crash in mdc_close for bad permissions on open (3285)
4172 - zero i_rdev for non-device files (3147)
4173 - clear page->private before handing to FS, better assertion (3119)
4174 - tune the read pipeline (3236)
4175 - fix incorrect decref of invalidated dentry (2350)
4176 - provide read-ahead stats and refine rpc in flight stats (3328)
4177 - don't hold journal transaction open across create RPC (3313)
4178 - update atime on MDS at close time (3265)
4179 - close LDAP connection when recovering to avoid server load (3315)
4180 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4181 - don't leak open file on MDS after open resend (3325)
4182 - serialize filter_precreate and filter_destroy_precreated (3329)
4183 - loop device shouldn't call sync_dev() for nul device (3092)
4184 - clear page cache after eviction (2766)
4185 - resynchronize MDS->OST in background (2824)
4186 - refuse to mount the same filesystem twice on same mountpoint (3394)
4187 - allow llmount to create routes for mounting behind routers (3320)
4188 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4189 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4190 - fix rare problem with rename on htree directories (3417)
4192 - allow default OST striping configuration per directory (1414)
4193 - fix compilation for qswnal for 2.6 kernels (3125)
4194 - increase maximum number of MDS request buffers for large systems
4195 - change liblustreapi to be useful for external progs like lfsck (3098)
4196 - increase local configuration timeout for slow disks (3353)
4197 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4199 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4202 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4203 - don't overwrite extent policy data in reply if lock was blocked (2901)
4204 - drop filter export grants atomically with removal from device (2663)
4205 - del obd_self_export from work_list in class_disconnect_exports (2908)
4206 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4207 - swab reply message in mdc_close, other PPC fixes (2464)
4208 - fix destroying of named logs (2325)
4209 - overwrite old logs when running lconf --write_conf (2264)
4210 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4211 - fix race in target_handle_connect (2898)
4212 - mds_reint_create() should take same inode create lock (2926)
4213 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4214 - don't close files for self_export to avoid uninitialized obd (2936)
4215 - allow MDS with the same name as client node (2939)
4216 - hold dentry reference for closed log files for unlink (2325)
4217 - reserve space for all logs during transactions (2059)
4218 - don't evict page beyond end of stripe extent (2925)
4219 - don't oops on a deleted current working directory (2399)
4220 - handle hard links to targets without a parent properly (2517)
4221 - don't dereference NULL lock when racing during eviction (2867)
4222 - don't grow lock extents when lots of conflicting locks (2919)
4224 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4227 - account for cache space usage on clients to avoid data loss (974)
4228 - lfsck support in lustre kernel code (2349)
4229 - reduce journal credits needed for BRW writes (2370)
4230 - orphan handling to avoid losing space on client/server crashes
4231 - ptlrpcd can be blocked, stopping ALL progress (2477)
4232 - use lock value blocks to assist in proper KMS, faster stat (1021)
4233 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4234 - recovery for initial connections (2355)
4235 - fixes for mds_cleanup_orphans (1934)
4236 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4237 - block all file creations until orphan recovery completes (1901)
4238 - client remove rq_connection from request struct (2423)
4239 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4240 - recovery timer race (2670)
4241 - mdc_close recovey bug (2532)
4242 - ptlrpc cleanup bug (2710)
4243 - mds timeout on local locks (2588)
4244 - namespace lock held during RPCs (2431)
4245 - handle interrupted sync write properly (2503)
4246 - don't try to handle a message that hasn't been replied to (2699)
4247 - client assert failure during cleanup after abort recovery (2701)
4248 - leak mdc device after failed mount (2712)
4249 - ptlrpc_check_set allows timedout requests to complete (2714)
4250 - wait for inflight reqs when ptlrpcd finishes (2710)
4251 - make sure unregistered services are removed from the srv_list
4252 - reset bulk XID's when resending them (caught by 1138 test)
4253 - unregister_bulk after timeout
4254 - fix lconf error (2694)
4255 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4256 - readahead locks pages, leaves pending causing memory pressure (2673)
4257 - increase OST request buffers to 4096 on large machines (2729)
4258 - fix up permission of existing directories in simple_mkdir (2661)
4259 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4260 - don't assign transno to errored transactions (2742)
4261 - don't delete objects on OST if given a bogus objid from MDS (2751)
4262 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4263 - if rq_no_resend, then timeout request after recovery (2432)
4264 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4265 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4266 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4267 - MDS recovery shouldn't delete objects that it has given out (2730)
4268 - if enqueue arrives after completion, don't clobber LVB (2819)
4269 - don't unlock pages twice when trigger_group_io returns error (2814)
4270 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4271 - don't write pages to disk if there was an error (1450)
4272 - don't ping imports that have recovery disabled (2676)
4273 - take buffered bytes into account when balancing socknal conn (2817)
4274 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4275 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4276 - remove little-endian swabbing of llog records (1987)
4277 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4278 - retry reposting request buffers if they fail (1191)
4279 - grow extent at grant time to avoid granting a revoked lock (2809)
4280 - lock revoke doesn't evict page if covered by a second lock (2765)
4281 - disable VM readahead to avoid reading outside lock extents (2805)
4283 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4284 - updated kernel patches for hp-2.4.20 kernel (2681)
4286 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4289 - fix truncated write corruption (2366)
4290 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4292 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4293 - ensure len(uuid) < 37 in lmc (1171)
4294 - fix ia64 OOPS in llog_test (2255)
4295 - zero end of page at obdfilter for partial page writes (2648)
4296 - don't leave stale dentries around after renames (bug 2428)
4297 - fix timeouts when evicting a client with a single lock held (2642)
4298 - set deadline for the initial HELLO message to drain (2634)
4299 - print out dotted-quad IP addresses in the socknal (2302)
4301 - additional debugging for MDS client eviction problem (2443)
4302 - fix mkfsoptions support for osts (2603, 2604)
4304 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4307 - add series for the vanilla 2.6.0 kernel
4308 - add series for the vanilla 2.4.24 kernel
4309 - add series for a cray x86/64 UL kernel drop
4310 - fix xattr patches for the vanilla 2.4.19 series
4312 - generate true UUIDs in lmc (1171)
4313 - have portals stack dumping break in UML (2466)
4314 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4315 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4316 - wake callback waiting threads on client eviction (2460)
4317 - Add --ptldebug and --subsystem to lmc (1719)
4318 - update assertion to allow safe interrupt allocation
4319 - set rq_no_resend for cancel requests (2432)
4320 - recalculate ptlrpcd timeout after resend (2494)
4321 - call vfs_rmdir when removing pending directories (2368)
4322 - fix renaming a file to itself (2429)
4323 - lmc creates a default one-stripe lov (2454)
4324 - expand procfs space to handle large clusters (2326)
4325 - increase UML stack to avoid overflow
4326 - update lconf's list of debug and subsystem masks
4327 - fix lfs find --obd (2510)
4328 - /proc tunable for disabling filter read caching (2591)
4329 - stop rpm packages from altering slapd.conf (2301)
4330 - disable nagle in the socknal under 0conf (2578)
4331 - choose mds inode size based on stripe count (2572)
4332 - fix kernel-source rpm problems (2516)
4334 - add --disable-doc to avoid pdf generation (2421)
4335 - update documentation, tests, type-os, comments
4336 - avoid format warnings on ia64
4337 - remove the TOE NAL
4338 - tiny code cleanups by removing unused fields
4340 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4343 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4344 - remove the most blatant lies from BUILDING file (2371)
4345 - change default debug level to reasonable production setting
4346 - reduce client side cache size to reduce cache flush time
4347 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4348 - make TCP zerocopy and pinger support enabled by default (2476)
4349 - sync writes completed after process exits caused crashes (2319)
4350 - maintain correct mount count on the MDS (2356)
4351 - backout 1557, because 2316 wasn't really fixed
4352 - better file I/O statistics gathering in /proc
4353 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4354 - another kernel patch to fix zero-copy TCP function export
4355 - don't take duplicate lock when processing re-sent getattr (2420)
4356 - lctl uses obd_self_export instead of creating new conn (2353)
4357 - MDS/OST recovery case which requires object creation asserted (2425)
4358 - move lfs from /usr/sbin to /usr/bin in packages
4359 - fix race between mds_client_add and mds_client_free (2417)
4360 - use kmalloc instead of slabs in portals (2430)
4361 - don't create duplicate records when a failover MDS is present (2442)
4362 - remove unnecessary mount age check (2332)
4363 - don't remove directory inodes from locks prematurely (2451)
4364 - don't break if MDS service name is the same as hostname (2103)
4365 - fix races in client write RPC generation when cache full (2482)
4367 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4370 - remove now-unused request->rq_obd (278)
4371 - if an allocation fails, print out how much memory we've used (1933)
4372 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4373 - add the "configurable stack size" patch to most series files (1256)
4374 - ability to write large log records, for 100+ OST configs (2306)
4375 - fix NULL deref when filter_prep fails (2314)
4376 - fix operator precedence error in filter_sync
4377 - dynamic allocation of socknal TX descriptors (2315)
4378 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4379 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4380 - ensure that CWARN messages in Portals always get to the syslog
4381 - __init/__exit are not for prototype decls (ldlm_init/exit)
4382 - x86-64 compile warning fixes
4383 - fix gateway LMC keyword conflict (2318)
4384 - fix MDS lock inversions in getattr/reint paths (1844)
4385 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4386 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4387 - DLM race condition prevented some lock evictions (2328)
4388 - ENOMEM detection and retry on socknal sends (2230)
4389 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4390 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4391 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4392 - better error messages when a client is rejected during recovery (1505)
4393 - avoid cancelling locks which were never granted, after failure (2330)
4394 - fix i_sem/journal inversion in mds_client_add (2333)
4395 - fix truncate/getattr lock cycle deadlock (2334)
4396 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4397 - fix two rare exit paths which could leak an l_lock() ref (2321)
4398 - fencepost error in MDS/OST orphan recovery (2226)
4399 - make log record alignment 8 bytes (1988)
4400 - lstripe now fails when requested offset > ost_count (2237)
4401 - ensure that all kernel series have a complete list.h (1607)
4402 - fix crashes in special-file operations (2316)
4403 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4405 - allow configurable automake binary, for testing new versions
4406 - small update to the lfs documentation
4408 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4410 * fix negative export reference count in fsfilt_sync (2312)
4412 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4413 * release candidate 0.9.1
4415 - orphans are moved into the PENDING directory for possible recovery
4416 - replayed opens now open by fid for orphan/rename safety (1042)
4417 - last close of an orphan inode generates a transno (683)
4418 - chdir() and mount() now pin the directory entry (1020)
4419 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4420 - discard very old requests without processing them (1502)
4421 - remove some common, well-understood CERRORs (1505)
4422 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4423 - clear "grant" flags in OST replies until OST grant code lands (1644)
4424 - fix read performance by not clobbering i_blksize on client (1598)
4425 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4426 - make LRU size a /proc tunable, clears locks when reduced (707)
4427 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4428 - server locks take references on exports now (1558)
4429 - build fixes for 2.4.20-rh trees (1663)
4430 - return an error from lov_create if all OSCs are inactive (1751)
4431 - fix import levels when a reconnect happens without a timeout (1597)
4432 - exit early from mds_open if we get a lookup error (1749)
4433 - partial page read at EOF wouldn't wait for disk before sending (1642)
4434 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4435 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4436 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4437 - write/truncate lock inversion (1639)
4438 - Don't auto-load obdclass, portals modules during cleanup (1495)
4439 - fix timestamps from jumping to "now" (1763)
4440 - extra journal assertions (1648)
4441 - add an extra multiunlink test (1771)
4442 - fix read_record/write_record API (1776)
4443 - fix leak of offset_extent, possible incorrect i_size later (1772)
4444 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4445 - Debugging for the kqswnal_get_idle_tx problems (1820)
4446 - Allow recovery to be attempted multiple times (1536)
4447 - Write out MDS last_rcvd file after it is first created (1600)
4448 - Fix tx_descriptor leak in failed transmit situations (1827)
4449 - ext3 journaling fixes for assertion failure after IO error (1871)
4450 - class_export_put() on freed export after completion AST error (1896)
4451 - Fix revalidate looping in VFS (1322)
4452 - Don't access a freed export during MDS_REINT timeout (1521)
4453 - Add open-unlink recovery support on the MDS (1673,1764)
4454 - Return an error if no MDS data was read from last_rcvd (1946)
4455 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4456 - Don't setup a disk device that doesn't match exported UUID (317)
4457 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4458 - avoid committing NULL handle in force close
4459 - local.sh is now a one-stripe LOV configuration
4460 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4461 - fix ext3 htree duplicate directory entry corruption (1516)
4462 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4463 - update to new LOV EA format (2097)
4464 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4466 2003-06-15 Phil Schwan <phil@clusterfs.com>
4469 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4470 - per-import/export recovery handling (958, 931, 959)
4471 - multiple last-rcvd slots, for serving multiple FSes (949)
4472 - connections are again shared between multiple imp/exports (963, 964)
4473 - "umount -f" would hang if any requests needed to be sent (393, 978)
4474 - avoid pinning large req buffer by copying for queued messages (989)
4475 - add "uuid" to "lctl device" command to help upcalls (991)
4476 - "open" RPCs with transnos would confuse recovery counters (1037)
4477 - do proper endian conversion of all wire messages (288, 340, 891)
4478 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4479 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4480 - fix ptlrpc_connection leak in target_handle_connect (1174)
4481 - fix import refcounting bug in OST and MDS cleanup (1134)
4482 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4483 - fix very unlikely obd_types race condition (501)
4484 - remove osc_open hack for echo_client (1187)
4485 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4486 - a failure in read_inode2 leads to deadlock (1139)
4487 - cancel ack-locks as soon as transaction is committed (1072)
4488 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4489 - make sure to commitrw after any preprw to avoid deadlock (1162)
4490 - failing to execute a file in a lustre FS would lock inode (1203)
4491 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4492 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4493 - obd_uuid2tgt would walk off the end of the list (1255)
4494 - on IA64 the osc would give portals incorrect bulk size (1258)
4495 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4496 - fix lock inversion caused by new llite matching code (1282)
4497 - limit the number of dirty pages on a client to 10MB (1286)
4498 - timed out locks were not being corrected cancelled (1289)
4499 - fix O_DIRECT above 4GB on IA-32 (1292)
4500 * major user-visible changes
4501 - fail out/fail over policy now controlled by the upcall (993)
4503 - add OBD_PING to check server availability and failure (954)
4504 - lustre messages are now sent in sending host order (288, 340, 891)
4505 - add eadatalen to MDS getattr reply (340)
4506 - OST read replies may contain second buffer, with per-page status (593)
4508 2003-03-11 Phil Schwan <phil@clusterfs.com>
4511 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4512 - failed open()s could cause deadlock; fixed (867, 869)
4513 - stop cancelling OST locks when files are closed (481)
4514 - overlapping XID spaces caused network corruption (851, 853)
4515 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4516 - setattr_raw added, to do single-RPC, server-side setattrs
4517 - lmc/lconf syntax change for OST UUIDs
4518 - fix crashy race condition between ptlrpc_free_req and osc_close
4519 - don't use request in mdc_enqueue if we hit a timeout (889)
4520 - don't set the inode i_size for regular files from the MDS (896)
4521 - handle out of order completion AST (842)
4522 - don't LBUG if a lock request times out after receiving AST (913)
4523 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4524 - if a bad lock AST arrives, send an error instead of dropping entirely
4525 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4526 - fix leak in bulk IO when only partially completed (899, 900, 926)
4527 - fix O_DIRECT for ia64 (55)
4528 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4529 - C-z after timeout could hang a process forever; fixed (977)
4531 - client-side I/O cache (678, 924, 929, 941, 970)
4533 - READPAGE and SETATTRs which don't take server-side locks get
4536 2003-02-11 Phil Schwan <phil@clusterfs.com>
4539 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4540 - Fix fsfilt_extN_readpage() to read a full page of directory
4541 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4542 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4543 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4544 - Most intent operations were removed, in favour of a new RPC mode
4545 that does a single RPC to the server and bypasses most of the VFS
4546 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4547 - Aggressively cancel local locks on DLM servers
4548 - mds_reint_unlink sends EA to the client if it's the last nlink.
4549 client uses that EA to unlink OST objects.
4550 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4551 - recursive symlinks were fixed (439)
4552 - fixed NULL deref in DEBUG_REQ
4553 - filter_update_lastobjid no longer calls sync, which annoyed extN
4554 - fixed multi-client small-writes to a single file problem (445)
4555 - fixed mtime updates during file writes (607)
4556 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4557 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4558 - fixed Linux OST size reporting problem (444, 656)
4559 - OST now updates object mtime with writes or setattr (607, 619)
4560 - client verifies file size before zeroing page past EOF (445)
4561 - OST now writes last allocated objid to disk with allocation (108)
4562 - LOV on echo now works (409)
4564 - mds_reint_unlink sends a new buffer, with the EA included. this
4565 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4566 set if a regular file was being unlinked, and it was the last link
4567 - use PtlGet from the target for bulk writes (315)
4568 - OST now updates object mtime with writes or setattr (607, 619)
4569 - LDLM now has a grant-time callback to revalidate locked items, if
4571 - Many MDS operations were reorganized to combat race conditions
4573 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4574 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4576 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4579 - Fully reactivate OST imports after reconnection (512, others)
4580 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4581 - More graceful error handling for truncating on dead OST (515)
4582 - Don't error out unless we're actually accessing dead stripes (474)
4583 - Fix garbage sizes when stripes are missing (410)
4584 - LRU counters were broken, causing constant lock purge (433, 432)
4585 - garbage on read from stripes with failed OSTs (441)
4586 - mark OSCs as active before reconnecting during recovery (438)
4587 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4588 - lfind did not preserve OST order in output (443)
4589 - symlinks cause hung clients, incorrect data (439)
4590 - stop dereferencing request after dropping refcount (457)
4591 - don't LASSERT(spin_is_locked) on non-SMP (455)
4592 - fixes for many rename() bugs
4593 - fstat didn't correctly synchronize attributes (399)
4594 - server must handle lock cancellation during blocking AST prep (487)
4595 - bulk descriptors were free()d too soon (511)
4596 - fix paths in lconf, which would load incorrect modules (451, 507)
4597 - fix confusing lconf 'host not found' error message (386)
4598 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4599 - fix race condition in mdc_blocking_ast() for inode access (526)
4600 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4601 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4602 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4603 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4604 - fix lstripe in several ways (kernel side) (527)
4605 - fix request leak in ldlm_cli_enqueue (262)
4606 - incorrect OSC was marked inactive after OST failure
4607 - call mds_fs_cleanup before unmounting filesystem (524)
4608 - fix races between taking ns_lock and ldlm_lock_change_resource
4609 - fix races updating LOV export open file list
4610 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4611 - fix recovery NULL deref in ldlm_cli_cancel_unused
4612 - fix some DLM races by using new hash table for lock handles (419)
4613 - permit the client to specify desired inodes, at replay
4614 - duplicate requests when we queue them for replay reintegration
4615 - fix last_rcvd offset calculation
4616 - sync after each recovered transaction, so we always make progress
4617 - never, not always, ERESTART requests without transnos
4618 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4620 - skip replay if the MDS says that the client is already connected
4621 - don't check for a recovery-enabled export to match lctl's UUID
4622 - don't INC_USE_COUNT for phantom exports
4623 - don't crash when cleaning up phantom exports (567)
4624 - don't double-finish or set replay data for errored mdc_open requests
4625 - abort requests when they time out, so we don't get old replies
4626 - send/receive replies for AST messages again
4627 - if the client says that it doesn't have the lock, cancel it on the
4629 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4630 mark it as destroyed, it will all work out when decref is called
4631 - fix module use counts (22, 581)
4633 - ASTs now expect a reply (server cancels lock on error reply)
4635 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4638 - fix many simultaneous client startup (392)
4639 - fix dentry->d_it clobbering
4640 - credentials weren't being shipped for readdir/getattr operations
4641 - remove invalid assertions triggered during some concurrent MD
4643 - proper Lustre versions added (336, 389)
4644 - fix memory leak for create error case (398)
4645 - fix LOV locking bug that would get cli/srv out of sync
4646 - fix echo client over LOV (409)
4647 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4648 - fix double-O_EXCL intent crash (424)
4649 - avoid sending multiple lock CANCELs (352)
4651 - MDS can do multi-client recovery (modulo bugs in new code)
4653 - many updates, edits, cleanups
4655 2002-11-18 Phil Schwan <phil@clusterfs.com>
4658 - fix null d_it dereference (346)
4659 - fix full OST/dbench hang (333)
4660 - fix permission problem with file removal (286)
4661 - fix removal of OSCs from LOV when they fail
4662 - fix NULL deref during bulk timeout (214)
4663 - fix problems related to multiple filesystems on one MDS (241)
4664 - fixed serious subtle metadata locking bugs
4665 - free locks on clients when inodes are removed due to memory
4667 - fix inode pointer in lock data (285)
4668 - partial support for multiple MDS on a single host (241)
4669 - data locks weren't cancelled at clear_inode time (290, 311)
4670 - intent locks could lead to unbounded lock growth (205)
4671 - added a maximum lock count, an LRU list, and a flusher
4672 - fix multiple rename (365)
4673 - properly abstracted the echo client
4674 - OSC locked 1 byte too many; fixed
4675 - rewrote brw callback code:
4676 - fixed recovery bugs related to LOVs (306)
4677 - fixed too-many-pages-in-one-write crash (191)
4678 - fixed (again) crash in sync_io_timeout (214)
4679 - probably fixed callback-related race (385)
4681 - Add capability to MDS protocol
4682 - LDLM cancellations and callbacks on different portals
4684 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4687 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4688 - defer open object destruction to close time (601981, 138)
4689 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4690 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4691 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4692 - don't call obd_setattr in ll_file_release for destroyed objects
4693 * protocol change to lustre_msg: move |version| and add |flags|
4694 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4695 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4696 * added replay of create, unlink, link and rename operations during
4697 MDS failover; recovery should be much more robust now
4698 * remove failed OSCs from LOVs (only lov_create uses this so far)
4699 * the lustre-HOWTO was brought (more) up to date (582544)
4701 2002-10-23 Phil Schwan <phil@clusterfs.com>
4704 - in-use dentries weren't being reused properly (617851)
4705 - prevent multiple LDLM setup (599178)
4706 - fix LOV size calculations for truncate (617853)
4707 - fix client handling of MDS intent errors (POSIX)
4708 - fix permission bug in lovstripe.c test (624321)
4709 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4710 - truncate past end of file could corrupt data
4711 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4712 - a race in recovery could return ETIMEDOUT to apps (623947)
4713 - building outside the source directory was fixed
4714 * the lustre-HOWTO was brought (more) up to date (582544)
4715 * major progress was made on recovery functionality
4717 2002-10-10 Phil Schwan <phil@clusterfs.com>
4720 - recovery deadlock fix
4721 - rm -rf causes LBUG fix (617817)
4722 - file open by multiple tasks fix (618962)
4723 - directory permissions bugs (602707 and 620007)
4724 - journal_stop fixed with locking (611313)
4725 - O_APPEND failures resolved (618273, perhaps 614459)
4726 - lconf PATH fix (619770)
4727 - IA64 build fix (621450)
4728 - RPC buffer sizes scale with amount of memory
4730 2002-10-01 Phil Schwan <phil@clusterfs.com>
4733 - locks would be cancelled without throwing away data pages,
4734 resulting in inconsistent data (605627)
4735 - inode attributes were not always being refreshed (605627, 612449)
4736 - lconf now continues to cleanup after lctl reports an error
4737 - MDS now enforces user permissions (602707)
4738 - lprocfs cleanup fixed, but not yet enabled (614157)
4739 - fixed infinite server hang, should a client not respond to an AST
4740 - avoid going into recovery if user calls readlink() with a buffer
4741 that's too small (613941)
4742 - AST RPCs no longer require replies (614867) -- this may be changed
4743 - don't crash server if client sends an IOV that's too big (611336)
4744 - fixed lock conversion deadlock (611892)
4745 - fixed the following of symlinks (614622)
4746 * recovery: the server can remove locks from a client that dies, other
4747 clients can make progress
4748 * more extN patch fixes
4749 * compile-time configurable ptlrpc buffer allocations
4751 - collaborative read cache document
4752 - Lustre Lite Performance CDR document-in-progress
4754 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4757 - fix typo in patch-2.4.18
4759 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4762 - clear ptlrpc request each time in handle_incoming_request()
4763 - unlink of files now destroys the object on the OST
4765 2002-09-19 Peter Braam <braam@clusterfs.com>
4767 * add hard link support
4768 * change obdfile creation method
4769 * kernel patch changed
4771 2002-09-19 Peter Braam <braam@clusterfs.com>
4774 - stack overflow bug in extN fixed
4776 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4778 * documentation updates
4779 - add man pages for config tools
4780 - update tests/README to describe testing with new config tools
4781 - finish metadata API descriptions
4782 * bug fixes and cleanups
4783 - statfs workaround for 16TB limit
4784 - LOV stripe allocation improved, can stripe on subset of OSTs
4785 - LOV file size/IO offset was wrong for files > 4GB in size
4786 - object EA data was being dropped, caused files to be unreadable
4787 - memory overflow with non-LOV OST caused memory corruption
4788 - fixed regression tests to work with new config tools, obdfilter
4789 - fixed bug when directory size became larger than 1 block
4790 - fixed bug (for single client case) when PWD was deleted
4791 - invalidate local directory pages when doing intent-based ops
4792 - avoid LDLM oops when lock callback contained bad data
4794 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4796 * documentation updates
4797 * bug fixes and cleanups
4798 - configuration tools
4801 - 64-bit compile warnings
4802 - 64-bit internal statfs data
4804 * test_brw on persistent OST devices
4806 * lprocfs (disabled)
4808 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4810 * documentation updates
4811 * bug fixes and cleanups
4812 * configuration tools
4814 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4817 * many small fixes to 0_5_4
4818 * io/network handling
4819 * thinkos in MDS operations
4821 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4824 * crucial basic fixes to 0.5.3
4825 * IOR, Iozone work over Elan
4828 2002-08-07 Phil Schwan <phil@clusterfs.com>
4829 * version 0_5_3, our first alpha
4830 * we use the new Portals iovs
4831 * documentation updates
4832 * bug fixes and cleanups
4833 * small changes in the DLM wire protocol
4835 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4836 * version 0_5_1 with some initial stability,
4837 * locking on MD and file I/O.
4838 * documentation updates
4839 * several bug fixes since 0.5.0
4840 * small changes in wire protocol
4842 2002-07-18 Phil Schwan <phil@clusterfs.com>
4844 * delivered as Lustre Light Alpha
4845 * fixed a crash after handling invalid MDS requests
4846 * fixed directory pages for architectures with non-4k pages sizes
4848 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4849 * release version v0_4_4
4850 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4852 * Fixes a number of interruption problems with OST operations.
4853 * Update documentation for portals header changes
4854 * Move all wire protocol structs/defines to lustre_idl.h
4855 * Fixes symlink length bug.
4856 * Add tcpdump to repository.
4858 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4859 * release version v0_4_3
4860 * Fixes statfs for inodes on extN.
4861 * Fixes bug in runtests which would delete /etc/hosts.
4862 * Use 64-bit object IDs wherever possible (not into VFS though)
4863 Remove ost_get_info, which is unused by lustre, and out of date.
4865 2002-07-03 Peter Braam <braam@clusterfs.com>
4866 * release version v0_4_2 Fixes a lookup error (type not passed)
4867 * move forward to head of Portals
4868 * move forward to latest Lustre kernel
4870 2002-06-25 Peter Braam <braam@clusterfs.com>
4871 * release version v0_4_1. Hopefully stable on single node use.