1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.21.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: 1.40.11-sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
18 Description: Using wrong pointer in osc_brw_prep_request
19 Details : Access to array[-1] can produce panic if kernel compiled with
20 CONFIG_PAGE_ALLOC enabled
24 Description: Client runs out of low memory
25 Details : Consider only lowmem when counting initial number of llap pages
29 Description: *optional* service tags registration
30 Details : if the "service tags" package is installed on a Lustre node
31 When the filesystem is mounted, a local-node service tag will
32 be created. See http://inventory.sun.com/ for more information
33 about the Service Tags asset management system.
37 Description: Kernel BUG tries to release flock
38 Details : Lustre does not destroy flock lock before last reference goes
39 away. So always drop flock locks when client is evicted and
40 perform unlock regardless of successfulness of speaking to MDS.
44 Description: add recount protection for osc callbacks, so avoid panic on shutdown
48 Description: sanity test 65a fails if stripecount of -1 is set
49 Details : handle -1 striping on filesystem in ll_dirstripe_verify
54 Description: ASSERTION(CheckWriteback(page,cmd)) failed
55 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
58 Severity : enhancement
60 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
64 Description: do not process already freed flock
65 Details : flock can possibly be freed by another thread before it reaches
66 to ldlm_flock_completion_ast.
70 Description: LBUG during stress test
71 Details : Need properly lock accesses the flock deadlock detection list.
75 Description: oops in page fault handler
76 Details : kernel page fault handler can return two special 'pages' in error case, don't
77 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
81 Description: timeout with invalidate import.
82 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
83 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
84 ptlrpc_invalidate_import and as result LASSERT.
86 Severity : enhancement
88 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
92 Description: enable MGS and MDT services start separately
93 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
94 a co-located MGS without starting the MGS, which is a complement
95 to 'nosvc' mount option.
99 Frequency : after recovery
100 Description: precreate to many object's after del orphan.
101 Details : del orphan st in oscc last_id == next_id and this triger growing
102 count of precreated objects. Set flag LOW to skip increase count
103 of precreated objects.
107 Frequency : rare, on clear nid stats
108 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
109 Details : when clean nid stats sometimes try destroy live entry,
110 and this produce panic in free.
114 Description: Stack overflow during MDS log replay
115 ease stack pressure by using a thread dealing llog_process.
119 Description: wait until IO finished before start new when do lock cancel.
120 Details : VM protocol want old IO finished before start new, in this case
121 need wait until PG_writeback is cleared until check dirty flag and
122 call writepages in lock cancel callback.
124 Severity : enhancement
126 Description: using special macro for print time and cleanup in includes.
130 Description: mds_mfd_close() ASSERTION(rc == 0)
131 Details : In mds_mfd_close(), we need protect inode's writecount change
132 within its orphan write semaphore to prevent possible races.
136 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
140 Frequency : rare, on shutdown ost
141 Description: don't hit live lock with umount ost.
142 Details : shrink_dcache_parent can be in long loop with destroy dentries,
143 use shrink_dcache_sb instead.
147 Description: don't panic with use echo client
148 Details : echo client pass NULL as client nid pointer and this produce null
153 Description: fix build on ppc32
154 Details : compile code with -m64 flag produce wrong object file for ppc32.
158 Description: add message levels for liblustreapi
162 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
163 Details : Move the 'good_osts' check before the 'total_bavail' check. This
164 will result in an -EAGAIN and in the exit call path we call
165 alloc_rr() which will with increasing aggressiveness attempt to
166 aquire precreated objects on the minimum number of required OSCs.
170 Description: Use old size assignment to avoid deadlock
171 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
172 the scheduling while holding a spinlock. We do not need locking
173 for size in ll_update_inode() because size is only updated from
174 the MDS for directories or files without objects, so there is no
175 other place to do the update, and concurrent access to such inodes
176 are protected by the inode lock.
180 Description: resolve "_IOWR redefined" build error on SLES10
184 Description: dump the memory debugging after all modules are unloaded to
185 suppress false negative in conf_sanity test 39
187 Severity : enhancement
189 Description: build kernel-ib packages for OFED 1.3 in our release cycle
194 Description: fix SLES kernel versioning
195 Details : the kernel version for our SLES 10 kernel did not include a "-"
196 before the "smp" at the end. while this was not a problem in
197 general, it did mean that software trying to use the kernel
198 version to try to detect a vendor specific kernel would fail.
199 this was most evident by the OFED build scripts.
203 Description: Don't update lov_desc members until making sure they are valid
204 Details : When updating lov_desc members via proc fs, need fix their
205 validities before doing the real update.
209 Description: don't put request into delay list while invalidate in flight.
210 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
211 import in flight. this produce timeout for invalidate and sometimes
212 can cause stale data.
216 Frequency : on ppc only
217 Description: not convert ost objects for directory because it's not exist.
218 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
222 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
224 Severity : enhancement
226 Description: Update to support 2.6.22.14 vanilla kernel.
230 Frequency : rare, on recovery
231 Description: read procfs can produce deadlock in some situation
232 Details : Holding lprocfs lock which send rpc can produce block for destroy
233 obd objects and this also block reconnect with -EALREADY. This isn't
234 fix all lprocfs bugs - but make it rare.
236 Severity : enhancement
238 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
241 Frequency : frequent on X2 node
243 Description: mdc_set_open_replay_data LBUG
244 Details : Set replay data for requests that are eligible for replay.
248 Description: lustre_mgs: operation 101 on unconnected MGS
249 Details : When MGC is disconnected from MGS long enough, MGS will evict the
250 MGC, and late on MGC cannot successfully connect to MGS and a lot
251 of the error messages complaining that MGS is not connected.
255 Frequency : on network error
256 Description: panic with double free request if network error
257 Details : mdc_finish_enqueue is finish request if any network error ocuring,
258 but it's true only for synchronus enqueue, for async enqueue
259 (via ptlrpcd) this incorrect and ptlrpcd want finish request
262 Severity : enhancement
264 Description: client-side metadata stat-ahead during readdir(directory readahead)
265 Details : perform client-side metadata stat-ahead when the client detects
266 readdir and sequential stat of dir entries therein
269 Frequency : on start mds
271 Description: Implement get_info(last_id) in obdfilter.
274 Frequency : occasional
276 Description: Correctly check stale fid, not start epoch if ost not support SOM
277 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
278 overwrited with new generated fid, but mds can anwer with one of these
279 two fids and both is not stale. setattr incorectly start epoch and
280 assume will be called done_writeting, but without SOM done_writing
284 Frequency : rare, depends on device drivers and load
286 Description: MDS or OSS nodes crash due to stack overflow
287 Details : Code changes in 1.8.0 increased the stack usage of some functions.
288 In some cases, in conjunction with device drivers that use a lot
289 of stack the MDS (or possibly OSS) service threads could overflow
290 the stack. One change which was identified to consume additional
291 stack has been reworked to avoid the extra stack usage.
294 Frequency : occasional
296 Description: Do not fail import if osc_interpret_create gets -EAGAIN
297 Details : If osc_interpret_create got -EAGAIN it immediately exits and
298 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
299 oscc_has_objects and see OSC has no objests and call
300 oscc_internal_create to resend create request.
302 Severity : enhancement
304 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
306 Severity : enhancement
308 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
311 Frequency : very rare
313 Description: Fix replay if there is an un-replied request and open
314 Details : In some cases, older replay request will revert the
315 mcd->mcd_last_xid on MDS which is used to record the client's
318 Severity : enhancement
320 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
322 Severity : enhancement
324 Description: Add rhel5 support to HEAD.
326 Serverity : enhancement
328 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
333 Description: Don't allow skipping OSTs if index has been specified.
334 Details : Don't allow skipping OSTs if index has been specified, make locking
335 in internal create lots better.
339 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
340 Details : access to bitfield in structure is always rounded to long
341 and this produce problem with not atomic change any bit.
345 Description: Lustre make rpms failed.
346 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
347 builds Lustre rpms from tarball.
350 Frequency : rare on shutdown ost
352 Description: If llog cancel was not send before clean_exports phase, this can
353 produce deadlock in llog code.
354 Details : If llog thread has last reference to obd and call class_import_put
355 this produce deadlock because llog_cleanup_commit_master wait when
356 last llog_commit_thread exited, but this never success because was
357 called from llog_commit_thread.
361 Description: allow userland application know is lost one of stripes.
362 Details : fill lvb_blocks with error code on ost and return it to
363 application if error flag found.
367 Description: NULL lov_tgts causing MDS oops
368 Details : more safe checks for NULL lov_tgts for avoid oops.
370 Severity : enhancement
372 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
376 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
377 Details : If an OST is down the MDS will hang indefinitely in
378 obd_statfs_rqset() waiting for the statfs data. While for
379 MDS QOS usage of statfs, it should not stuck in waiting.
381 Severity : enhancement
383 Description: remote_acl support
384 Details : Support ACL-based permission check for remote user.
385 Support setfacl/getfacl for remote user with the utils
386 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
387 the system "{s,g}etfacl" utils.
389 Severity : enhancement
391 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
393 Severity : enhancement
395 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
399 Description: make mgs_setparam() handle fsname containing dash
400 Details : fsname containing a dash does not work with lctl conf_param
402 Severity : enhancement
404 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
406 Severity : enhancement
408 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
412 Frequency : rare, at shutdown
413 Description: access already free / zero obd_namespace.
414 Details : if client_disconnect_export was called without force flag set,
415 and exist connect request in flight, this can produce access to
416 NULL pointer (or already free pointer) when connect_interpret
417 store ocd flags in obd_namespace.
421 Frequency : only at startup
422 Description: not alloc memory with spinlock held.
423 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
424 if any spinlock held.
426 Severity : enhancement
428 Description: make lustre randomly fail allocating memory
429 Details : Make lustre randomly failed allocating memory for testing purpose.
431 Severity : enhancement
433 Description: lost problems with lov objid file
434 Details : Fixes some scability and access to not inited memory problems
435 in work with lov objdid file.
440 Description: lfs find does not continue on file error
441 Details : Continue other files processing when a file/dir is absent.
445 Description: Inconsistent usage of lustre_pack_reply()
446 Details : Standardize the usage of lustre_pack_reply() such that it
447 always generate a CERROR on failure.
450 Frequency : occasional
452 Description: MDS hang or stay in waiting lock
453 Details : If client receive lock with CBPENDING flag ldlm need send lock
454 cancel as separate rpc, to avoid situation when cancel request
455 can't processed due all i/o threads stay in wait lock.
459 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
463 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
467 Description: Update to SLES9 kernel-2.6.5-7.287.3.
471 Description: Update to RHEL4 latest kernel.
473 Severity : enhancement
475 Description: Build SLES10 patchless client fails
476 Details : The configure was broken by run ./configure with
477 --with-linux-obj=.... argument for patchless client. When the
478 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
479 can't search header adequately. Use absolute path such as
480 -I($LINUX)/include instead.
484 Description: interrupt oig_wait produce painc on resend.
485 Details : brw_redo_request can be used for resend requests from ptlrpcd and
486 private set, and this produce situation when rq_ptlrpcd_data not
487 copyed to new allocated request and triggered LBUG on assert
488 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
489 ptlrpcd set if request is changed and can be safety changed to use
494 Description: LASSERT_{REQ,REP}SWAB macros are buggy
495 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
496 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
497 these macros and replace them with their difinitions instead.
501 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
502 Details : Update which_patch & target file for SLES10 latest kernel.
505 Description: Remove client patches from SLES 10 kernel.
506 Details : This causes SLES 10 clients to behave as patchless clients
507 even on a Lustre-patched (server) kernel.
509 Severity : enhancement
511 Description: self-adjustable client's lru lists
512 Details : use adaptive algorithm for managing client cached locks lru
513 lists according to current server load, other client's work
514 pattern, memory activities, etc. Both, server and client
515 side namespaces provide number of proc tunables for controlling
518 Severity : enhancement
520 Description: light-weight GSS support
521 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
522 and reduce performance overhead.
524 Severity : enhancement
526 Description: Linux keyring support
527 Details : Support using service of Linux keyring for Lustre GSS internal
528 context refresh/cache mechanism.
532 Description: Fix errors in lfs documentation
533 Details : Fixes man pages
537 Description: don't use GFP_* in generic Lustre code.
538 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
542 Description: obdclass is limited by single OBD_ALLOC(idarray)
543 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
547 Description: warnings with build patchless client with vanila 2.6.19 and up
548 Details : change old ctl_table style and replace ctl_table/ctl_table_header
549 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
553 Frequency : Only for SLES
554 Description: sanity_quota fail test_1
555 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
556 file, which makes TSTID[2] inunique.
560 Frequency : Only for SLES9
561 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
562 Details : Update target/ChangeLog/which_patch .
566 Frequency : Only for test_10 in sanity-quota.sh
567 Description: a bug in quota test code
568 Details : checking a test flag in a wrong place
572 Frequency : Only for test_13 in sanity-quota.sh
573 Description: a bug in quota test script
574 Details : shouldn't check overfull allocated quota
578 Description: jbd statistics
579 Details : Port older jbd statistics patch for sles10
581 Severity : enhancement
583 Description: Add printing inode info into message about error in writepage.
588 Description: reply_lock_interpret crash due to race with it and lock cancel.
589 Details : Do not replay locks that are being cancelled. Do not reference
590 locks by their address during replay, just by their handle.
594 Frequency : When flocks are used.
595 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
596 Details : Flock locks might destroy just granted lock if it could be merged
597 with another existing flock, this is done in completion handler,
598 so teach ldlm_cli_enquque_fini that this is a valid case for
604 Description: Oops in read and write path when failing to allocate lock.
605 Details : Check if lock allocation failed and return error back.
609 Description: lstripe command fails for valid OST index
610 Details : The stripe offset is compared to lov->desc.ld_tgt_count
611 instead of lov->desc.ld_active_tgt_count.
615 Description: sanity.sh failed test 103
616 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
621 Description: df doesn't work properly if diskfs blocksize != 4K
622 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
626 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
627 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
629 Severity : enhancement
631 Description: Adaptive timeouts
632 Details : RPC timeouts adapt to changing server load and network
633 conditions to reduce resend attempts and improve recovery time.
637 Description: llapi_file_create() does not allow some changes
638 Details : add llapi_file_open() that allows specifying the mode and
639 open flags, and also returns an open file handle.
643 Description: merge and cleanup kernel patches.
644 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
648 Description: Add journal checksum support.(Kernel part)
649 Details : The journal checksum feature adds two new flags i.e
650 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
651 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
652 indicates that the commit block contains the checksum for
653 the blocks described by the descriptor blocks. Now commit
654 record can be sent to disk without waiting for descriptor
655 blocks to be written to disk. This behavior is controlled
656 using JBD2_FEATURE_ASYNC_COMMIT flag.
660 Description: OSS needs mutliple precreate threads
661 Details : Add ability to start more than one create thread per OSS.
665 Description: Sanity.sh test_65k failed: setstripe should have succeeded
666 Details : execute lfs setstripe on client
670 Description: mds_obd_create error creating tmp object
671 Details : When the user sets quota on root, llog will be affected and can't
672 create files and write files.
675 Frequency : Always on ia64 patchless client, and possibly others.
677 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
678 Details : This allows the patchless client to be loaded on architectures
683 Description: RedHat Update kernel for RHEL5
684 Details : Add the kernel config file for RHEL5.
688 Description: RedHat Update kernel for RHEL5
689 Details : Modify the kernel config file more closer RHEL5.
693 Description: Build failure against Centos5 (RHEL5)
694 Details : Define PAGE_SIZE when it isn't present.
698 Description: test_7 on sanity-quota.sh can't be used on separate machines
699 Details : add facet to handle it
703 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
704 Details : with NFS, the anon dentry's parent was set to itself in
705 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
706 corresponding dentry other than use rec->ur_name.
708 Severity : enhancement
710 Description: lfs setstripe enhancement
711 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
715 Description: enable data checksumming by default
716 Details : enable checksum by default, allow --disable-checksum
717 configure option and "-o nochecksum" mount option. Checksums
718 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
722 Description: lustre support for RHEL5
723 Details : Add support for RHEL5.
727 Description: Client eviction due to failover config
728 Details : after a connection loss, the lustre client should attempt to
729 reconnect to the last active server first before trying the
730 other potential connections.
734 Description: when mds and osts use different quota unit(32bit and 64bit),
735 quota will be released repeatly.
736 Details : void sending multiple quota reqs to mds, which will keep the
737 status between the reqs.
741 Description: osts not allocated evenly to files
742 Details : change the condition to increase offset_idx
745 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
747 Description: Data corruption for OSTs that are formatted larger than 2TB
749 Details : When generating the bio request for lustre file writes the
750 sector number would overflow a temporary variable before being
751 used for the IO. The data reads correctly from Lustre (which
752 will overflow in a similar manner) but other file data or
753 filesystem metadata may be corrupted in some cases.
757 Description: Tune the kernel for good SCSI performance.
758 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
759 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
764 Description: rewrite ext2-derived code in obdclass/uuid.c
765 Details : rewrite inherited code (uuid parsing code from ext2 utils)
766 from scratch preserving functionality.
770 Description: "lfs find -obd UUID" prints directories
771 Details : "lfs find -obd UUID" will return all directory names instead
772 of just file names. It is incorrect because the directories
773 do not reside on the OSTs.
777 Description: use i_size_read and i_size_write in 2.6 port
778 Details : replace inode->i_size access with i_size_read/write()
781 Frequency : when removing large files
783 Description: scheduling issue during removal of large Lustre files
784 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
785 It causes scheduling issues when removing large files (17TB in the
789 Frequency : only with liblustre clients on XT3
791 Description: evictions taking too long
792 Details : allow llrd to evict clients directly on OSTs
795 Frequency : only on ppc
797 Description: /proc/fs/lustre/devices broken on ppc
798 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
799 We should make sure the type of 'index' is loff_t and then cast
800 explicitly as needed below. Do not assign an explicitly cast
804 Frequency : only for rhel5
806 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
807 Details : Modify the target file & which_kernel.
809 Severity : enhancement
811 Description: omit set fsid for export NFS
812 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
813 as FS_REQUIRES_DEV for avoid problems with generate fsid.
817 Frequency : Always, for kernels after 2.6.16
818 Description: Fix warning idr_remove called for id=.. which is not allocated.
819 Details : Last kernels save old s_dev before kill super and not allow
820 to restore from callback - restore it before call kill_anon_super.
824 Description: buffer overruns could theoretically occur
825 Details : llapi_semantic_traverse() modifies the "path" argument by
826 appending values to the end of the origin string, and a
827 overrun may occur. Adding buffer overrun check in liblustreapi.
831 Description: Fix error on 'ls .' at the top of the Lustre mount.
832 Details : Don't revalidate dentry if it is a root dentry.
836 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
837 Details : Modify vm-tunables-rhel4.patch.
841 Description: Kernel config for 2.6.18-vanilla.
842 Details : Modify targets/2.6-vanilla.target.in.
843 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
844 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
845 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
846 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
850 Description: improve handling recoverable errors
851 Details : if request processig with error which can be recoverable on server
852 request should be resend, otherwise page released from cache and
857 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
858 Details : Modify target file & which_patch.
859 A flaw was found in the IA32 system call emulation provided
860 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
861 value could be stored in the %RAX register, which could trigger an
862 out-of-bounds system call table access. An untrusted local user
863 could exploit this flaw to run code in the kernel
864 (ie a root privilege escalation). (CVE-2007-4573).
868 Description: O_DIRECT bypasses client statistics.
869 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
870 read_ahead_stats were updated. copied the stats section from
871 osc_send_oap_rpc() into async_internal().
875 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
879 Description: change order of libsysio includes
880 Details : '#include sysio.h' should always come before '#include xtio.h'
884 Description: handle "serious error: objid * already exists" more gracefully
885 Details : If LAST_ID value on disk is smaller than the objects existing in
886 the O/0/d* directories, it indicates disk corruption and causes an
887 LBUG(). If the object is 0-length, then we should use the existing
888 object. This will help to avoid a full fsck in most cases.
890 Severity : enhancement
892 Description: adapt the lustre_config script to support the upgrade case
893 Details : Add "-u" option for lustre_config script to support upgrading 1.4
894 server targets to 1.6 in parallel.
898 Description: To avoid grant space > avaible space when the disk is almost
899 full. Without this patch you might see the error "grant XXXX >
900 available" or some LBUG about grant, when the disk is almost
902 Details : In filter_check_grant, for non_grant cache write, we should
903 check the left space by if (*left > ungranted + bytes), instead
904 of (*left > ungranted), because only we are sure the left space
905 is enough for another "bytes", then the ungrant space should be
906 increase. In client, we should update cl_avail_grant only there
907 is OBD_MD_FLGRANT in the reply.
911 Description: Update RHEL 4 kernel to fix local root privilege escalation.
912 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
913 described in CVE-2007-4573. This problem could allow untrusted
914 local users to gain root access.
917 Frequency : when using O_DIRECT and quotas
919 Description: Incorrect file ownership on O_DIRECT output files
920 Details : block usage reported by 'lfs quota' does not take into account
921 files that have been written with O_DIRECT.
926 Description: touch file failed when fs is not full
927 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
928 otherwise we can get ENOSP while fs is not full.
932 Description: parallel lock callbacks
933 Details : Instead of sending blocking and completion callbacks as separated
934 requests, adding them to a set and sending in parallel.
937 Frequency : only for Cray XT3
938 Bugzilla : 12829/13455
939 Description: Changing primary group doesn't change the group lustre assigns to
941 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
942 overridden with the primary group provided by the group upcall,
943 whereas the supplied fsgid can be trusted if it is in the list of
944 supplementary groups returned by the group upcall.
946 Severity : enhancement
948 Description: Allow masking D_WARNING, D_ERROR messages from console
949 Details : Console messages can now be disabled via lnet.printk.
953 Description: User code with malformed file open parameter crashes client node
954 Details : Before packing join_file req, all the related reference should be
955 checked carefully in case some malformed flags cause fake
956 join_file req on client.
960 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
961 Details : ldlm_completion_ast() assumes that a lock is granted when the req
962 mode is equal to the granted mode. However, it should also check
963 that LDLM_FL_CP_REQD is not set.
967 Description: Heavy nfs access might result in deadlocks
968 Details : After ELC code landed, it is now improper to enqueue any mds
969 locks under och_sem, because enqueue might want to decide to
970 cancel open locks for same inode we are holding och_sem for.
974 Description: Client eviction while running blogbench
975 Details : A lot of unlink operations with concurrent I/O can lead to a
976 deadlock causing evictions. To address the problem, the number of
977 oustanding OST_DESTROY requests is now throttled to
978 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
979 callbacks are processed in priority.
983 Description: enable ACLs on MDS by default
984 Details : ACLs must be enabled on MDS by default.
987 Frequency : PPC/PPC64 only
989 Description: conflicts between asm-ppc64/types.h and lustre_types.h
990 Details : fix duplicated definitions between asm-ppc64/types.h and
991 lustre_types.h on PPC.
994 Frequency : PPC/PPC64 only
996 Description: asm-ppc/segment.h does not exist
997 Details : fix compile issue on PPC.
1001 Description: better handle error messages in extents code
1004 Frequency : RHEL4 only
1006 Description: mkfs is very slow on IA64/RHEL4
1007 Details : A performance regression has been discovered in the MPT Fusion
1008 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1009 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1010 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1012 Severity : enhancement
1014 Description: SNMP support enhancement
1015 Details : Adding total number of sampled request for an MDS node in snmp
1018 Severity : enhancement
1020 Description: Optimize ldlm waiting list processing for PR extent locks
1021 Details : When processing waiting list for read extent lock and meeting read
1022 lock that is same or wider to it that is not contended, skip
1023 processing rest of the list and immediatelly return current
1024 status of conflictness, since we are guaranteed there are no
1025 conflicting locks in the rest of the list.
1029 Description: Time out and refuse to reconnect
1030 Details : When the failover node is the primary node, it is possible
1031 to have two identical connections in imp_conn_list. We must
1032 compare not conn's pointers but NIDs, otherwise we can defeat
1033 connection throttling.
1037 Description: port llog fixes from b1_6 into HEAD
1038 Details : Port llog reference couting and some llog cleanups from b1_6
1039 (bug 10800) into HEAD, for protect from panic and access to already
1040 free llog structures.
1044 Description: Detect stride IO mode in read-ahead
1045 Details : When a client does stride read, read-ahead should detect that and
1046 read-ahead pages according to the detected stride pattern.
1050 Description: data checksumming impacts single node performance
1051 Details : add support for several checksum algorithm. Currently, only CRC32
1052 and Adler-32 are supported. The checksum type can be changed on
1053 the fly via /proc/fs/lustre/osc/*/checksum_type.
1057 Description: use adler32 for page checksums
1058 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1063 Description: build for x2 fails
1064 Details : fix compile issue on Cray systems.
1068 Description: Properly match for duplicate locks
1069 Details : Due to different lock order from skiplists code, we need to
1070 traverse entire list for now
1073 Frequency : only on PPC/SLES10
1075 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1076 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1077 be defined. Add a hack in mkfs_lustre.c to work around this compile
1082 Description: LASSERT on MDS when client holding flock lock dies
1083 Details : ldlm pool logic depends on number of granted locks equal to
1084 number of released locks which is not true for flock locks, so
1085 just exclude such locks from consideration.
1089 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1090 Details : Use fsfilt_sync() for both the whole filesystem sync and
1091 individual file sync to eliminate dangerous inode locking
1092 with I_LOCK that can lead to a deadlock.
1096 Description: performance in 1.6.3
1097 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1098 sure that 1MB requests are not fragmented by the block layer.
1100 Severity : enhancement
1102 Description: organize the server-side client stats on per-nid basis
1103 Details : Change the structure of stats under obdfilter and mds to
1111 The "uuid"s file would list the uuids of _active_ exports.
1112 And the clear entry is to clear all stats and stale nids.
1114 Severity : enhancement
1116 Description: eliminate client locks in face of contention
1117 Details : file contention detection and lockless i/o implementation
1118 for contended files.
1122 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1125 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1127 Description: MDS may lose file striping (and hence file data) in some cases
1128 Details : If there are additional extended attributes stored on the MDS,
1129 in particular ACLs, SELinux, or user attributes (if user_xattr
1130 is specified for the client mount options) then there is a risk
1131 of attribute loss. Additionally, the Lustre file striping
1132 needs to be larger than default (e.g. striped over all OSTs),
1133 and an additional attribute must be stored initially in the
1134 inode and then increase in size enough to be moved to the
1135 external attribute block (e.g. ACL growing in size) for file
1140 Description: skiplist implementation simplification
1141 Details : skiplists are used to group compatible locks on granted list
1142 that was implemented as tracking first and last lock of each lock group
1143 the patch changes that to using doubly linked lists
1147 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1148 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1149 because LASSERT on both the data supplied by a client, and the data
1150 on disk is dangerous and incorrect.
1152 Severity : enhancement
1154 Description: Slow truncate/writes to huge files at high offsets.
1155 Details : Directly associate cached pages to lock that protect those pages,
1156 this allows us to quickly find what pages to write and remove
1157 once lock callback is received.
1161 Description: more ldlm soft lockups
1162 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1163 starve other threads from the resource lock for a long time in
1164 case of long waiting queue, so change the debug level from
1165 D_OTHER to the less frequently used D_INFO.
1167 Severity : enhancement
1169 Description: add -gid, -group, -uid, -user options to lfs find
1173 Description: Hung threads in invalidate_inode_pages2_range
1174 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1175 one is completed. As a result, some RPCs are stuck in the queue
1180 Description: filter threads hungs on waiting journal commit
1181 Details : Cleanup filter group llog code, then only filter group llog will
1182 be only created in the MDS/OST syncing process.
1186 Description: Procfs and llog threads access destoryed import sometimes.
1187 Details : Sync the import destoryed process with procfs and llog threads by
1188 the import refcount and semaphore.
1190 Severity : enhancement
1192 Description: openlock cache of b1_6 port to HEAD
1194 --------------------------------------------------------------------------------
1196 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1198 * Support for kernels:
1199 2.6.5-7.283 (SLES 9),
1200 2.6.9-55.EL (RHEL 4),
1201 2.6.16.46-0.14 (SLES 10),
1202 2.6.18.8 vanilla (kernel.org)
1203 * Client support for unpatched kernels:
1204 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1205 2.6.16 - 2.6.22 vanilla (kernel.org)
1206 * Due to recently discovered recovery problems, we do not recommend
1207 using patchless RHEL 4 clients with this or any earlier release.
1208 * Recommended e2fsprogs version: 1.39.cfs8
1209 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1210 * Starting with this release, the ldiskfs backing filesystem required
1211 by Lustre is now in its own package, lustre-ldiskfs. This package
1212 should be installed. It is versioned separately from Lustre and
1213 may be released separately in future.
1217 Description: block reactivating mgc import until all deactivates complete
1218 Details : Fix race when failing back MDT/MGS to itself (testing)
1221 Frequency : at statup only
1223 Description: mds_lov_synchronize race leads to various problems
1224 Details : simultaneous MDT->OST connections at startup can cause the
1225 sync to abort, leaving the OSC in a bad state.
1227 Severity : enhancement
1229 Description: add optional extra BUILD_VERSION info
1230 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1231 to override the lustre version.
1234 Frequency : 2.6.18 servers only
1236 Description: ll_kern_mount() doesn't release the module reference
1237 Details : The ldiskfs module reference count never drops down to 0
1238 because ll_kern_mount() doesn't release the module reference.
1243 Description: server LBUG when using old ost_num_threads parameter
1244 Details : Accept the old ost_num_threads parameter but warn that it
1245 is deprecated, and fix an off-by-one error that caused an LBUG.
1250 Description: Transient SCSI error results in persistent IO issue
1251 Details : iobuf->dr_error is not reinitialized to 0 between two
1255 Frequency : sometimes when underlying device returns I/O errors
1257 Description: OSTs not going read-only during write failures
1258 Details : OSTs are not remounted read-only when the journal commit threads
1259 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1260 instead of the ext3 wrappers.
1264 Description: poor connect scaling with increasing client count
1265 Details : Don't run filter_grant_sanity_check for more than 100 exports
1266 to improve scaling for large numbers of clients.
1269 Frequency : SLES10 only
1271 Description: sanity-quota.sh quotacheck failed: rc = -22
1272 Details : Quotas cannot be enabled on SLES10.
1275 Frequency : liblustre clients only
1277 Description: getdirentries does not give error when run on compute nodes
1278 Details : getdirentries does not fail when the size specified as an argument
1279 is too small to contain at least one entry
1281 Severity : enhancement
1283 Description: Add LNET router traceability for debug purposes
1284 Details : If a checksum failure occurs with a router as part of the
1285 IO path, the NID of the last router that forwarded the bulk data
1286 is printed so it can be identified.
1291 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1292 Details : Due to a race condition, liblustre clients were occasionally
1293 evicted incorrectly.
1295 Severity : enhancement
1297 Description: lfs setstripe use optional parameters instead of postional
1300 Severity : enhancement
1302 Description: Nanosecond timestamp support for ldiskfs
1303 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1304 resolution timestamps. There is not yet support for this at
1305 the Lustre filesystem level.
1308 Frequency : during server recovery
1310 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1311 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1315 Frequency : nfs export on patchless client
1317 Description: connectathon hang when test nfs export over patchless client
1318 Details : Disconnected dentry cannot be found with lookup, so we do not need
1319 to unhash it or make it invalid
1322 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1323 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1324 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1329 Description: open req refcounting wrong on reconnect
1330 Details : If reconnect happened between getting open reply from server and
1331 call to mdc_set_replay_data in ll_file_open, we will schedule
1332 replay for unreferenced request that we are about to free.
1333 Subsequent close will crash in variety of ways.
1334 Check that request is still eligible for replay in
1335 mdc_set_replay_data().
1340 Description: disable writes to filesystem when reading health_check file
1341 Details : the default for reading the health_check proc file has changed
1342 to NOT do a journal transaction and write to disk, because this
1343 can cause reads of the /proc file to hang and block HA state
1344 checking on a healthy but otherwise heavily loaded system. It
1345 is possible to return to the previous behaviour during configure
1346 with --enable-health-write.
1348 Severity : enhancement
1350 Description: 64-bit inode version
1351 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1352 to the inode. This will be required for version-based recovery.
1357 Description: MDS fails to start if a duplicate client export is detected
1358 Details : in some rare cases it was possible for a client to connect to
1359 an MDS multiple times. Upon recovery the MDS would detect this
1360 and fail during startup. Handle this more gracefully.
1362 Severity : enhancement
1364 Description: Add -o localflock option to simulate old noflock
1366 Details : This will achieve local-only flock/fcntl locks
1372 Description: log_commit_thread vs filter_destroy race leads to crash
1373 Details : Take import reference before releasing llog record semaphore
1378 Description: Wrong request locking in request set processing
1379 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1380 imp_delayed_list, in this place should be used imp_lock.
1383 Frequency : when reconnection
1385 Description: Grant Leak when osc reconnect to OST
1386 Details : When osc reconnect ost, OST(filter) should check whether it
1387 should grant more space to client by comparing fed_grant and
1388 cl_avail_grant, and return the granted space to client instead
1389 of "new granted" space, because client will call osc_init_grant
1390 to update the client grant space info.
1393 Frequency : when client reconnect to OST
1395 Description: Grant Leak when osc do resend and replay bulk write
1396 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1397 bulk write request, because the grant info while be sync between
1398 OSC and OST when reconnect, and we should ignore the grant info
1399 these of resend/replay write req.
1404 Description: Grant space more than avaiable left space sometimes.
1405 Details : When then OST is about to be full, if two bulk writing from
1406 different clients came to OST. Accord the avaliable space of the
1407 OST, the first req should be permitted, and the second one
1408 should be denied by ENOSPC. But if the seconde arrived before
1409 the first one is commited. The OST might wrongly permit second
1410 writing, which will cause grant space > avaiable space.
1413 Frequency : when client is evicted
1415 Description: Grant might be wrongly erased when osc is evicted by OST
1416 Details : when the import is evicted by server, it will fork another
1417 thread ptlrpc_invalidate_import_thread to invalidate the
1418 import, where the grant will be set to 0. While the original
1419 thread will update the grant it got when connecting. So if
1420 the former happened latter, the grant will be wrongly errased
1421 because of this race.
1426 Description: Checking Stale with correct fid
1427 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1428 to check whether it is stale, because sometimes, we want the
1429 enqueue happened anyway, and op_data.fid2 will not be initialized.
1431 Severity : enhancement
1433 Description: update patchless client
1434 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1437 Frequency : only with 2.4 kernel
1439 Description: random memory corruption
1440 Details : size of struct ll_inode_info is to big for union inode.u and this
1441 can be cause of random memory corruption.
1446 Description: Memory leak in recovery
1447 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1448 It should also check obd_fail before fsfilt_start, otherwise if
1449 fsfilt_start return -EROFS,(failover mds during mds recovery).
1450 then the req will return with repmsg->transno = 0 and rc = EROFS.
1451 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1452 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1453 be freed no matter whether fsfilt_commit success or not.
1456 Frequency : only with huge count clients
1458 Description: Prevents from taking the superblock lock in llap_from_page for
1460 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1461 for llap_from_page call in ll_removepage prevents from taking the
1462 superblock lock for a soon died page.
1467 Description: Not check open intent error before release open handle
1468 Details : in some rare cases, the open intent error is not checked before
1469 release open handle, which may cause
1470 ASSERTION(open_req->rq_transno != 0), because it tries to release
1471 the failed open handle.
1476 Description: Set cat log bitmap only after create log success.
1477 Details : in some rare cases, the cat log bitmap is set too early. and it
1478 should be set only after create log success.
1482 Description: Accessing a block bevice can re-enable I/O when Lustre is
1483 tearing down a device.
1484 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1489 Description: service threads may hog cpus when there are a lot of requests
1490 Details : Insert cond_resched to give other threads a chance to use some CPU
1495 Description: the cat log was not initialized in recovery
1496 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1497 unlink log on mds will not be initialized until mds post
1498 recovery. And also in mds post recovery, the unlink log will
1499 initialization will be done asynchronausly, so there will be race
1500 between add unlink log and unlink log initialization.
1504 Description: brw_stats were being printed incorrectly
1505 Details : brw_stats were being printed as log2 but all of them were not
1506 recorded as log2. Also remove some code duplication arising from
1507 filter_tally_{read,write}.
1511 Frequency : rare, only in recovery.
1512 Description: ASSERTION(req->rq_type != LI_POISON) failed
1513 Details : imp_lock should be held while iterating over imp_sending_list for
1514 prevent destroy request after get timeout in ptlrpc_queue_wait.
1518 Description: replay-single.sh test 52 fails
1519 Details : A lock's skiplist need to be cleanup when it being unlinked
1520 from its resource list.
1524 Description: Short directio read returns full requested size rather than
1526 Details : Direct I/O operations should return actual amount of bytes
1527 transferred rather than requested size.
1529 Severity : enhancement
1531 Description: metadata RPC reduction (e.g. for rm performance)
1532 Details : decrease the amount of synchronous RPC between clients and servers
1533 by canceling conflicing lock before the operation on the client
1534 and packing thier handles into the main operation RPC to server.
1536 Severity : enhancement
1538 Description: Async OSC create to avoid the blocking unnecessarily.
1539 Details : If a OST has no remain object, system will block on the creating
1540 when need to create a new object on this OST. Now, ways use
1541 pre-created objects when available, instead of blocking on an
1542 empty osc while others are not empty. If we must block, we block
1543 for the shortest possible period of time.
1545 Severity : enhancement
1547 Description: refine locking for avoid write wrong info into lov_objid file
1548 Details : fix possible races with add new target and write/update data in
1551 --------------------------------------------------------------------------------
1553 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1558 Frequency : on some architectures
1560 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1561 Details : Uninitialized flags sometimes cause configuration commands to
1565 Frequency : patchless clients only
1567 Description: missing __iget() symbol export
1568 Details : The __iget() symbol export is missing. To avoid the need for
1569 this on patchless clients the deathrow inode reaper is turned
1570 off, and we depend on the VM to clean up old inodes. This
1571 dependency was during via the fix for bug 12181.
1573 --------------------------------------------------------------------------------
1575 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1577 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1578 INTEROPERATE with older versions automatically. In many cases a
1579 special upgrade step is needed. Please read the
1580 user documentation before upgrading any part of a live system.
1581 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1582 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1583 * WARNING: Lustre configuration and startup changes are required with
1584 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1590 Severity : enhancement
1592 Description: MountConf
1593 Details : Lustre configuration is now managed via mkfs and mount
1594 commands instead of lmc and lconf. New obd types (MGS, MGC)
1595 are added for dynamic configuration management. See
1596 https://mail.clusterfs.com/wikis/lustre/MountConf for
1599 Severity : enhancement
1601 Description: dynamic OST addition
1602 Details : OSTs can now be added to a live filesystem
1604 Severity : enhancement
1606 Description: startup order invariance
1607 Details : MDTs and OSTs can be started in any order. Clients only
1608 require the MDT to complete startup.
1610 Severity : enhancement
1612 Description: parallel, asynchronous orphan cleanup
1613 Details : orphan cleanup is now performed in separate threads for each
1614 OST, allowing parallel non-blocking operation.
1616 Severity : enhancement
1618 Description: optimized stripe assignment
1619 Details : stripe assignments are now made based on ost space available,
1620 ost previous usage, and OSS previous usage, in order to try
1621 to optimize storage space and networking resources.
1623 Severity : enhancement
1625 Description: Permanently set tunables
1626 Details : All writable /proc/fs/lustre tunables can now be permanently
1627 set on a per-server basis, at mkfs time or on a live system.
1629 Severity : enhancement
1631 Description: Lustre message v2
1632 Details : Add lustre message format v2.
1634 Severity : enhancement
1636 Description: client OST exclusion list
1637 Details : Clients can be started with a list of OSTs that should be
1638 declared "inactive" for known non-responsive OSTs.
1642 Description: SPEC SFS validation failure on NFS v2 over lustre.
1643 Details : Changes the blocksize for regular files to be 2x RPC size,
1644 and not depend on stripe size.
1646 Severity : enhancement
1648 Description: Multiple MD RPCs in flight.
1649 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1650 To avoid overly-overloading MDS, introduce a limit on number of
1651 MDS RPCs in flight for a single client and add /proc controls
1652 to adjust this limit.
1654 Severity : enhancement
1656 Description: client read/write statistics
1657 Details : Add client read/write call usage stats for performance
1658 analysis of user processes.
1659 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1660 file access. extents_stats shows chunk size distribution.
1661 extents_stats_per_process show chunk size distribution per
1664 Severity : enhancement
1666 Description: mds statistics
1667 Details : Add detailed mds operations statistics in
1668 /proc/fs/lustre/mds/*/stats.
1672 Description: Failure of copying files with lustre special EAs.
1673 Details : Client side always return success for setxattr call for lustre
1674 special xattr (currently only "trusted.lov").
1679 Description: Refcount LNET uuids
1680 Details : The global LNET uuid list grew linearly with every startup;
1681 refcount repeated list entries instead of always adding to
1684 Severity : enhancement
1686 Description: Dynamic service threads
1687 Details : Within a small range, start extra service threads
1688 automatically when the request queue builds up.
1691 Frequency : mixed-endian client/server environments
1693 Description: mixed-endian crashes
1694 Details : The new msg_v2 system had some failures in mixed-endian
1697 Severity : enhancement
1699 Description: Easy OST removal
1700 Details : OSTs can be permanently deactivated with e.g. 'lctl
1701 conf_param lustre-OST0001.osc.active=0'
1703 Severity : enhancement
1705 Description: MGS proc entries
1706 Details : Added basic proc entries for the MGS showing what filesystems
1709 Severity : enhancement
1711 Description: provide MGS failover
1712 Details : Added config lock reacquisition after MGS server failover.
1714 Severity : enhancement
1716 Description: add Linux 2.4 support
1717 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1721 Description: a large application tries to do I/O to the same resource and dies
1722 in the middle of it.
1723 Details : Check the req->rq_arrival time after the call to
1724 ost_brw_lock_get(), but before we do anything about
1725 processing it & sending the BULK transfer request. This
1726 should help move old stale pending locks off the queue as
1727 quickly as obd_timeout.
1730 Frequency : when an incorrect nid is specified during startup
1732 Description: ptlrpc connect to non-existant node causes kernel crash
1733 Details : LNET can't be re-entered from an event callback, which
1734 happened when we expire a message after the export has been
1735 cleaned up. Instead, hand the zombie cleanup off to another
1738 Severity : enhancement
1740 Description: plain/inodebits lock performance improvement
1741 Details : Grouping plain/inodebits in granted list by their request modes
1742 and bits policy, thus improving the performance of search through
1746 Frequency : only if OST filesystem is corrupted
1748 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1749 Details : for a short time RPCs with bulk IO are in the replay list,
1750 but replay of bulk IOs is unimplemented. If the OST filesystem
1751 is corrupted due to disk cache incoherency and then replay is
1752 started it is possible to trip an assertion. Avoid putting
1753 committed RPCs into the replay list at all to avoid this issue.
1756 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1758 Description: System hang on startup
1759 Details : This bug allowed the liblustre (e.g. catamount) client to
1760 return to the app before handling all startup RPCs. This
1761 could leave the node unresponsive to lustre network traffic
1762 and manifested as a server ptllnd timeout.
1764 Severity : enhancement
1766 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1767 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1768 boolean to control whether to print peer debug info when a
1769 client's RPC times out.
1772 Frequency : only for kernels with patches from Lustre below 1.4.3
1774 Description: Remove old rdonly API
1775 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1778 Frequency : only for devices with external journals
1780 Description: Set external device read-only also
1781 Details : During a commanded failover stop, we set the disk device
1782 read-only while the server shuts down. We now also set any
1783 external journal device read-only at the same time.
1786 Frequency : when upgrading from 1.4 while trying to change parameters
1788 Description: The wrong (new) MDC name was used when setting parameters for
1789 upgraded MDT's. Also allows changing of OSC (and MDC)
1790 parameters if --writeconf is specified at tunefs upgrade time.
1793 Frequency : when setting specific ost indicies
1795 Description: QOS code breaks on skipped indicies
1796 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1797 created with --index need not be sequential.
1799 Severity : enhancement
1801 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1802 Details : The uninit_groups feature works in conjunction with the kernel
1803 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1804 up the pass1 processing of e2fsck. This is a read-only feature
1805 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1806 mount filesystems that have had this feature enabled.
1808 Severity : enhancement
1810 Description: Improve multi-block allocation algorithm to avoid fragmentation
1811 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1812 allocation locality and avoid filesystem fragmentation.
1814 ------------------------------------------------------------------------------
1816 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1818 * Support for kernels:
1819 2.4.21-47.0.1.EL (RHEL 3)
1820 2.6.5-7.283 (SLES 9)
1821 2.6.9-42.0.10.EL (RHEL 4)
1822 2.6.12.6 vanilla (kernel.org)
1823 2.6.16.27-0.9 (SLES 10)
1824 * Recommended e2fsprogs version: 1.39.cfs5
1826 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1830 Frequency : occasional, depends on client load and configuration
1831 Bugzilla : 12181, 12203
1832 Description: data loss for recently-modified files
1834 Details : In some cases it is possible that recently written or created
1835 files may not be written to disk in a timely manner (this should
1836 normally be within 30s unless client IO load is very high).
1837 The problem appears as zero-length files or files that are a
1838 multiple of 1MB in size after a client crash or client eviction
1839 that are missing data at the end of the file.
1841 This problem is more likely to be hit on clients where files are
1842 repeatedly created and unlinked in the same directory, clients
1843 have a large amount of RAM, have many CPUs, the filesystem has
1844 many OSTs, the clients are rebooted frequently, and/or the files
1845 are not accessed by other nodes after being written.
1847 The presence of the problem can be detected by looking at
1848 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1849 smaller than the second (nr_unused) then dirty files will not
1850 be flushed automatically to disk. "sync; sleep 10" should be
1851 run several times on the node before unmounting it to update
1852 Lustre (this is also safe to run on nodes without this problem).
1854 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1855 that can cause this same problem, so customers using that kernel
1856 also need to update the kernel in addition to Lustre. In order
1857 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1859 It is normal that files written just before a client crash (less
1860 than 30s) may not yet have been flushed to disk, even for local
1864 Frequency : frequent on thin XT3 nodes
1866 Description: UUID collision on thin XT3 Linux nodes
1867 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1868 randomly, since we relied on an insufficiently-seeded PRNG.
1873 Description: OSS hangs after "All ost request buffers busy"
1874 Details : A deadlock between quota and journal operations caused OSS
1875 hangs after printing "All ost request buffers busy."
1878 Frequency : always on liblustre builds
1880 Description: Cleanup compiler warnings on liblustre
1883 Frequency : always on liblustre builds on XT3
1885 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1889 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1890 > 32000 subdirectories
1891 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1892 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1893 subdirectory count crosses 32000. This will aid e2fsck to
1894 correctly handle more than 32000 subdirectories.
1897 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1899 Description: System hang on startup
1900 Details : This bug allowed the liblustre (e.g. catamount) client to
1901 return to the app before handling all startup RPCs. This
1902 could leave the node unresponsive to lustre network traffic
1903 and manifested as a server ptllnd timeout.
1905 Severity : enhancement
1907 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1908 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1909 boolean to control whether to print peer debug info when a
1910 client's RPC times out.
1915 Description: make O_SYNC working on 2.6 kernels
1916 Details : 2.6 kernels use different method for mark pages for write,
1917 so need add a code to lustre for O_SYNC work.
1922 Description: Failure to close file and release space on NFS
1923 Details : Put inode details into lock acquired in ll_intent_file_open.
1924 Use mdc_intent_lock in ll_intent_open to properly
1925 detect all kind of errors unhandled by mdc_enqueue.
1930 Description: proc file read during shutdown sometimes raced obd removal,
1932 Details : Add lock to prevent obd access after proc file removal.
1935 Frequency : Only for files larger than 4GB on 32-bit clients.
1937 Description: improperly doing page alignment of locks
1938 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1939 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1944 Description: RPCs being resent when they shouldn't be
1945 Details : Some RPCs that should not be resent are being resent. This
1946 can cause inconsistencies in the RPC state machine. Do not
1947 resend such requests.
1950 Frequency : rare, only with NFS export
1952 Description: Crash on NFS re-export node
1953 Details : under very unusual load conditions an assertion is hit in
1954 ll_intent_file_open()
1957 Frequency : only if OST filesystem is corrupted
1959 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1960 Details : for a short time RPCs with bulk IO are in the replay list,
1961 but replay of bulk IOs is unimplemented. If the OST filesystem
1962 is corrupted due to disk cache incoherency and then replay is
1963 started it is possible to trip an assertion. Avoid putting
1964 committed RPCs into the replay list at all to avoid this issue.
1969 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1970 Details : Large single O_DIRECT read and write calls can fail to allocate
1971 a sufficiently large buffer to process the request. In case of
1972 allocation failure the allocation is retried with a smaller
1973 buffer and broken into smaller requests.
1975 Severity : enhancement
1977 Description: Add -o localflock option to simulate old noflock behaviour.
1978 Details : This will achieve local-only flock/fcntl locks coherentness.
1983 Description: versioning check is incomplete
1984 Details : Checking the version difference of client vs. server, report
1985 error if the gap is too big.
1990 Description: add support PG_writeback bit
1991 Details : add support for PG_writeback bit for Lustre, for more carefull
1992 work with page cache in 2.6 kernel. This also fix some deadlocks
1993 and remove hack for work O_SYNC with 2.6 kernel.
1995 Severity : enhancement
1997 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1998 Details : The uninit_groups feature works in conjunction with the kernel
1999 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2000 up the pass1 processing of e2fsck. This is a read-only feature
2001 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2002 mount filesystems that have had this feature enabled.
2004 Severity : enhancement
2006 Description: Improve multi-block allocation algorithm to avoid fragmentation
2007 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2008 allocation locality and avoid filesystem fragmentation.
2010 ------------------------------------------------------------------------------
2012 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2014 * Support for kernels:
2015 2.6.9-42.0.3.EL (RHEL 4)
2016 2.6.5-7.276 (SLES 9)
2017 2.4.21-47.0.1.EL (RHEL 3)
2018 2.6.12.6 vanilla (kernel.org)
2019 2.6.16.21-0.8 (SLES10)
2020 * Recommended e2fsprogs version: 1.39.cfs2-0
2022 * The backwards-compatible /proc/sys/portals symlink has been removed
2023 in this release. Before upgrading, please ensure that you change
2024 any configuration scripts or /etc/sysctl.conf files that access
2025 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2026 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2027 in advance of the upgrade on any system running Lustre 1.4.6 or
2028 newer, since /proc/sys/lnet was added in that version.
2029 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2033 Frequency : only when quota is used
2035 Description: avoid scanning export list for quota master
2036 Details : Change the algorithms to avoid scanning export list in order
2037 to improve the efficiency.
2040 Frequency : MDS failover only, very rarely
2042 Description: "went back in time" messages on mds failover
2043 Details : The greatest transno may be lost when the current operation
2044 finishes with an error (transno==0) and the client's last_rcvd
2045 record is over-written. Save the greatest transno in the
2046 mds_last_transno for this case.
2049 Frequency : always for specific kernels and striping counts
2051 Description: client may get "Matching packet too big" without ACL support
2052 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2053 when trying to access files in certain configurations. The
2054 clients should in fact be denied when mounting because they do
2055 not understand ACLs.
2058 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2060 Description: many clients connecting with IO in progress causes connect timeouts
2061 Details : Avoid synchronous journal commits to avoid delays caused by many
2062 clients connecting/disconnecting when bulk IO is in progress.
2063 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2064 OST_IO_PORTAL to avoid delays behind potentially many pending
2068 Frequency : occasionally with multiple writers to a single file
2070 Description: shared writes to file may result in wrong size reported by stat()
2071 Details : Allow growing of kms when extent lock is cancelled
2074 Frequency : always with random mmap IO to multi-striped file
2076 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2077 Details : Only if the hole is at the end of OST object so that kms is too
2078 small. Fix is to increase kms accordingly in ll_nopage.
2081 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2083 Description: writes to a missing object would leak memory on the OST
2084 Details : If there is an inconsistency between the MDS and OST filesystems,
2085 such that the MDS references an object that doesn't exist, writes
2086 to that object will leak memory due to incorrect cleanup in the
2087 error handling path, eventually running out of memory on the OST.
2092 Description: Creating too long symlink causes lustre errors
2093 Details : Check symlink and name lengths before sending requests to MDS.
2096 Frequency : only if flock is enabled (not on by default)
2098 Description: posix locks not released on fd closure on 2.6.9+
2099 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2100 kernels, this caused such locks not to be released on fd close and
2101 then assertions on fs unmount about still used locks.
2104 Frequency : MDS failover only, very rarely
2106 Description: clients may get ASSERTION(granted_lock != NULL)
2107 Details : When request was taking a long time, and a client was resending
2108 a getattr by name lock request. The were multiple lock requests
2109 with the same client lock handle and
2110 mds_getattr_name->fixup_handle_for_resent_request found one of the
2111 lock handles but later failed with ASSERTION(granted_lock != NULL).
2116 Description: handle->h_buffer_credits > 0, assertion failure
2117 Details : h_buffer_credits is zero after truncate, causing assertion
2118 failure. This patch extends the transaction or creates a new
2122 Frequency : NFS re-export or patchless client
2123 Bugzilla : 11179, 10796
2124 Description: Crash on NFS re-export node (__d_move)
2125 Details : We do not want to hash the dentry if we don't have a lock.
2126 But if this dentry is later used in d_move, we'd hit uninitialised
2127 list head d_hash, so we just do this to init d_hash field but
2128 leave dentry unhashed.
2131 Frequency : NFS re-export or patchless client
2133 Description: NFS exports has problem with symbolic link
2134 Details : lustre client didn't properly install dentry when re-exported
2135 to NFS or running patchless client.
2138 Frequency : NFS re-export or patchless client
2140 Description: Various nfs/patchless fixes.
2141 Details : fixes reuse disconected alias for lookup process - this fixes
2142 warning "find_exported_dentry: npd != pd",
2143 fix permission error with open files at nfs.
2144 fix apply umask when do revalidate.
2147 Frequency : occasional
2149 Description: Crash on NFS re-export node
2150 Details : calling clear_page() on the wrong pointer triggered oops in
2151 generic_mapping_read().
2154 Frequency : rarely, using O_DIRECT IO
2156 Description: unaligned directio crashes client with LASSERT
2157 Details : check for unaligned buffers before trying any requests.
2160 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2162 Description: stale data returned from RAID cache
2163 Details : If only a small amount of IO is done to the RAID device before
2164 reading it again it is possible to get stale data from the RAID
2165 cache instead of reading it from disk.
2168 Frequency : always for sles10 kernel
2170 Description: sles10 support
2171 Details : ll_follow_link: compile fixes and using of nd_set_link
2172 under newer kernels.
2175 Frequency : depends on arch, kernel and compiler version, always on sles10
2178 Description: recursive or deep enough symlinks cause stack overflow
2179 Details : getting rid of large stack-allocated variable in
2183 Frequency : depends on hardware
2185 Description: lustre write performance loss in the SLES10 kernel
2186 Details : the performance loss is caused by using of write barriers in the
2187 ext3 code. The SLES10 kernel turns barrier support on by
2188 default. The fix is to undo that change for ldiskfs.
2190 ------------------------------------------------------------------------------
2192 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2194 * Support for kernels:
2195 2.6.9-42.0.3EL (RHEL 4)
2196 2.6.5-7.276 (SLES 9)
2197 2.4.21-47.0.1.EL (RHEL 3)
2198 2.6.12.6 vanilla (kernel.org)
2202 Frequency : quota enabled and large files being deleted
2204 Description: releasing more than 4GB of quota at once hangs OST
2205 Details : If a user deletes more than 4GB of files on a single OST it
2206 will cause the OST to spin in an infinite loop. Release
2207 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2212 Description: statfs data retrieved from /proc may be stale or zero
2213 Details : When reading per-device statfs data from /proc, in the
2214 {kbytes,files}_{total,free,avail} files, it may appear
2215 as zero or be out of date.
2218 Frequency : systems with MD RAID1 external journal devices
2220 Description: lconf's call to blkid is confused by RAID1 journal devices
2221 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2222 of returning all block devices that match the journal UUID.
2225 Frequency : always, for aggregate stripe size over 4GB
2227 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2228 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2229 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2232 Frequency : always if "lfs find" used on a local file/directory
2234 Description: "lfs find" segfaults if used on a local file/directory
2235 Details : The case where a directory component was not specified wasn't
2236 handled correctly. Handle this properly.
2239 Frequency : always on ppc64
2241 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2242 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2243 perform endianness conversion.
2246 Frequency : rarely (truncate to non-zero file size after write under load)
2247 Bugzilla : 10730, 10687
2248 Description: Files padded with zeros to next 4K multiple
2249 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2250 it is possible that files that are truncated to a non-zero size
2251 immediately after being written are filled with zero bytes beyond
2252 the truncated size. No file data is lost.
2254 Severity : enhancement
2256 Description: Allow recovery/failover for liblustre clients.
2257 Details : liblustre clients were unaware of failover configurations until
2260 Severity : enhancement
2262 Description: user file locks should fail when not mounting with flock option
2263 Details : Set up an error-returning stub in ll_file_operations.lock field
2264 to prevent incorrect behaviour when client is mounted without
2265 flock option. Also, set up properly f_op->flock field for
2269 Frequency : always on ia64
2271 Description: "lfs df" loops on printing out MDS statfs information
2272 Details : The obd_ioctl_data was not initialized and in some systems
2273 this caused a failure during the ioctl that did not return
2274 an error. Initialize the struct and return an error on failure.
2277 Frequency : SLES 9 only
2279 Description: Error of copying files with lustre special EAs as root
2280 Details : Client side always return success for setxattr call for lustre
2281 special xattr (currently only "trusted.lov").
2284 Frequency : rarely on clusters with both ia64+i386 clients
2286 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2287 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2288 clients are doing concurrent writes to the same file it is
2289 possible that the ia64 clients may LASSERT because the OST
2290 extent locks are not PAGE_SIZE aligned. Ensure that grown
2291 locks are always aligned on the request boundary.
2294 Frequency : specific use, occasional
2296 Description: Overwriting in use executable truncates on-disk binary image
2297 Details : If one node attempts to overwrite an executable in use by
2298 another node, we now correctly return ETXTBSY instead of
2299 truncating the file.
2301 Severity : enhancement
2303 Description: Async OSC create to avoid the blocking unnecessarily.
2304 Details : If a OST has no remain object, system will block on the creating
2305 when need to create a new object on this OST. Now, ways use
2306 pre-created objects when available, instead of blocking on an
2307 empty osc while others are not empty. If we must block, we block
2308 for the shortest possible period of time.
2313 Description: chmod on Lustre root is propagated to other clients
2314 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2315 invalid by the follow_mount time.
2318 Frequency : liblustre clients only
2320 Description: Race in 'instant cancel' lock handling could lead to such locks
2321 never to be granted in case of SMP MDS
2322 Details : Do not destroy not yet granted but cbpending locks in
2326 Frequency : replay/resend of open
2328 Description: non null lock assetion failure in mds_intent_policy
2329 Details : Trying to replay/resend lockless open requests resulted in
2330 mds_open() returning 0 with no lock. Now it sets a flag if
2331 a lock is going to be returned.
2333 Severity : enhancement
2335 Description: Checksum enhancements
2336 Details : New checksum enhancements allow for resending RPCs that failed
2339 Severity : enhancement
2341 Description: Tunables on number of dirty pages in cacche
2342 Details : Allow to set limit on number of dirty pages cached.
2347 Description: client crash on unmount - lock still has references
2348 Details : In some error handling cases it was possible to leak a lock
2349 reference on a client while accessing a file. This was not
2350 harmful to the client during operation, but would cause the
2351 client to crash when the filesystem is unmounted.
2354 Frequency : specific case, rare
2356 Description: ETXTBSY on mds though file not in use by client
2357 Details : ETXTBSY is no longer incorrectly returned when attempting to
2358 chmod or chown a directory that the user previously tried to
2359 execute or a currently-executing binary.
2362 Frequency : extremely rare except on liblustre-based clients
2364 Description: Lustre space not freed when files are deleted
2365 Details : Clean up open-unlinked files after client eviction. Previously
2366 the unlink was skipped and the files remained as orphans.
2371 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2372 Details : In some cases it was possible to send a blocking callback to
2373 a client doing a glimpse, even though that client didn't get
2374 a lock granted. When the glimpse lock is cancelled on the OST
2375 the freed lock is left on the waiting list and corrupted the list.
2378 Frequency : all core dumps
2380 Description: Broke core dumps to lustre
2381 Details : Negative dentry may be unhashed if parent does not have UPDATE
2382 lock, but some callers, e.g. do_coredump, expect dentry to be
2383 hashed after successful create, hash it in ll_create_it.
2385 ------------------------------------------------------------------------------
2387 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2389 * Support for kernels:
2390 2.6.9-42.0.2.EL (RHEL 4)
2391 2.6.5-7.276 (SLES 9)
2392 2.4.21-40.EL (RHEL 3)
2393 2.6.12.6 vanilla (kernel.org)
2397 Frequency : always on RHEL 3
2399 Description: Number of open files grows over time
2400 Details : The number of open files grows over time, whether or not
2401 Lustre is started. This was due to a filp leak introduced
2402 by one of our kernel patches.
2404 ------------------------------------------------------------------------------
2406 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2408 * Support for kernels:
2409 2.6.9-42.EL (RHEL 4)
2410 2.6.5-7.276 (SLES 9)
2411 2.4.21-40.EL (RHEL 3)
2412 2.6.12.6 vanilla (kernel.org)
2417 Bugzilla : 5719, 9635, 9792, 9684
2418 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2419 Details : If a server is under heavy load and cannot reply to new
2420 connection requests before the client resends the (re)connect,
2421 the connection handling code can behave badly if two service
2422 threads are concurrently handing separate (re)connections from
2423 the same client. Add better locking to the connection handling
2424 code, and ensure that only a single connection will be processed
2425 for a given client UUID, even if the lock is dropped.
2427 Severity : enhancement
2429 Description: add TCP zero-copy support to kernel
2430 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2431 sends if the hardware supports scatter-gather and checksumming.
2432 This allows socklnd to do client-write and server-read more
2433 efficiently and reduce CPU utilization from skbuf copying.
2436 Frequency : only if NFS exporting from client
2438 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2439 Details : If setting attributes on a file created under NFS that had
2440 never been opened it would be possible to oops the client
2441 if the file had no objects.
2444 Frequency : always for liblustre
2446 Description: liblustre client does MDS+OSTs setattr RPC for each write
2447 Details : When doing a write from a liblustre client, the client
2448 incorrectly issued an RPC to the MDS and each OST the file was
2449 striped over in order to update the timestamps. When writing
2450 with small chunks and many clients this could overwhelm the MDS
2451 with RPCs. In all cases it would slow down the write because
2452 these RPCs are unnecessary.
2454 Severity : enhancement
2456 Description: allow number of MDS service threads to be changed at module load
2457 Details : It is now possible to change the number of MDS service threads
2458 running. Adding "options mds mds_num_threads={N}" to the MDS's
2459 /etc/modprobe.conf will set the number of threads for the next
2460 time Lustre is restarted (assuming the "mds" module is also
2461 reloaded at that time). The default number of threads will
2462 stay the same, 32 for most systems.
2467 Description: OST crash if filesystem is unformatted or corrupt
2468 Details : If an OST is started on a device that has never been formatted
2469 or if the filesystem is corrupt and cannot even mount then the
2470 error handling cleanup routines would dereference a NULL pointer.
2475 Description: NULL pointer deref in llap_from_page.
2476 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2477 incorrect) mapping if the page was truncated/reclaimed while it was
2478 searched for. Check for this condition and skip such pages when
2479 doing readahead. Introduce extra check to llap_from_page() to
2480 verify page->mapping->host is non-NULL (so page is not anonymous).
2483 Frequency : Sometimes when using sys_sendfile
2485 Description: "page not covered by a lock" warnings from ll_readpage
2486 Details : sendfile called ll_readpage without right page locks present.
2487 Now we introduced ll_file_sendfile that does necessary locking
2488 around call to generic_file_sendfile() much like we do in
2492 Frequency : with certain MDS communication failures at client mount time
2494 Description: NULL pointer deref after failed client mount
2495 Details : a client connection request may delayed by the network layer
2496 and not be sent until after the PTLRPC layer has timed out the
2497 request. If the client fails the mount immediately it will try
2498 to clean up before the network times out the request. Add a
2499 reference from the request import to the obd device and delay
2500 the cleanup until the network drops the request.
2503 Frequency : occasionally during client (re)connect
2505 Description: assertion failure during client (re)connect
2506 Details : processing a client connection request may be delayed by the
2507 client or server longer than the client connect timeout. This
2508 causes the client to resend the connection request. If the
2509 original connection request is replied in this interval, the
2510 client may trip an assertion failure in ptlrpc_connect_interpret()
2511 which thought it would be the only running connect process.
2514 Frequency : only with obd_echo servers and clients that are rebooted
2516 Description: kernel BUG accessing uninitialized data structure
2517 Details : When running an obd_echo server it did not start the ping_evictor
2518 thread, and when a client was evicted an uninitialized data
2519 structure was accessed. Start the ping_evictor in the RPC
2520 service startup instead of the OBD startup.
2522 Severity : enhancement
2523 Bugzilla : 10193 (patchless)
2524 Description: Remove dependency on various unexported kernel interfaces.
2525 Details : No longer need reparent_to_init, exit_mm, exit_files,
2526 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2529 Frequency : rare (only users of deprecated and unsupported LDAP config)
2531 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2532 Details : LDAP apparently contains 'lustreName' attributes instead of
2533 'name'. A simple remapping of the name is sufficient.
2536 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2538 Description: waiting_locks_callback trips kernel BUG if client is evicted
2539 Details : Running with the dump_on_timeout debug flag turned on makes
2540 it possible that the waiting_locks_callback() can try to dump
2541 the Lustre kernel debug logs from an interrupt handler. Defer
2542 this log dumping to the expired_lock_main() thread.
2544 Severity : enhancement
2546 Description: Support NFS exporting on 2.6 kernels.
2547 Details : Implement non-rawops metadata methods for NFS server to use without
2548 changing NFS server code.
2551 Frequency : very rare (synthetic metadata workload only)
2553 Description: two racing renames might cause an MDS thread to deadlock
2554 Details : Running the "racer" program may cause one MDS thread to rename
2555 a file from being the source of a rename to being the target of
2556 a rename at exactly the same time that another thread is doing
2557 so, and the second thread has already enqueued these locks after
2558 doing a lookup of the target and is trying to relock them in
2559 order. Ensure that we don't try to re-lock the same resource.
2562 Frequency : only very large systems with liblustre clients
2564 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2565 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2566 all OSTs in parallel. This allows the eviction of stale liblustre
2567 clients to proceed much faster than if they were done in series,
2568 and also offers similar improvements for other set_info RPCs.
2573 Description: excessive CPU usage during initial read phase on client
2574 Details : During the initial read phase on a client, it would agressively
2575 retry readahead on the file, consuming too much CPU and impacting
2576 performance (since 1.4.5.8). Improve the readahead algorithm
2577 to avoid this, and also improve some other common cases (read
2578 of small files in particular, where "small" is files smaller than
2579 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2584 Description: MDS crash when receiving packet with unknown intent.
2585 Details : Do not LBUG in unknown intent case, just return -EFAULT
2587 Severity : enhancement
2588 Bugzilla : 9293, 9385
2589 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2590 Details : Do not serialize getattr (non-intent version) and statfs.
2593 Frequency : occasional, when OST network is overloaded/intermittent
2595 Description: client evicted by OST after bulk IO timeout
2596 Details : If a client sends a bulk IO request (read or write) the OST
2597 may evict the client if it is unresposive to its data GET/PUT
2598 request. This is incorrect if the network is overloaded (takes
2599 too long to transfer the RPC data) or dropped the OST GET/PUT
2600 request. There is no need to evict the client at all, since
2601 the pinger and/or lock callbacks will handle this, and the
2602 client can restart the bulk request.
2605 Frequency : Always when mmapping file with no objects
2607 Description: client crashes when mmapping file with no objects
2608 Details : Check that we actually have objects in a file before doing any
2609 operations on objects in ll_vm_open, ll_vm_close and
2615 Description: Request leak when working with deleted CWD
2616 Details : Introduce advanced request refcount tracking for requests
2617 referenced from lustre intent.
2619 Severity : Enhancement
2621 Description: Cache open file handles on client.
2622 Details : MDS now will return special lock along with openhandle, if
2623 requested and client is allowed to hold openhandle, even if unused,
2624 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2625 closing files for every read/write openration.
2627 Severity : Enhancement
2629 Description: Cache open negative dentries on client when possible.
2630 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2631 negative dentries on lock revocation.
2636 Description: Remounting a client read-only wasn't possible with a zconf mount
2637 Details : It wasn't possible to remount a client read-only with llmount.
2639 Severity : enhancement
2640 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2641 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2642 MPICH is included to add Lustre-specific ADIO interfaces.
2643 This is based closely on the UFS ADIO layer and only differs
2644 in file creation, in order to allow the OST striping to be set.
2645 This is user-contributed code and not supported by CFS.
2650 Description: extended inode attributes (immutable, append-only) work improperly
2651 when 2.4 and 2.6 kernels are used on client/server or vice versa
2652 Details : Introduce kernel-independent values for these flags.
2654 Severity : enhancement
2657 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2658 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2659 granularity. Now, Fractional MB granularity is be supported,
2660 this is very useful for low memory system.
2662 Severity : enhancement
2664 Description: Getattr by fid
2665 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2666 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2667 have locks and avoids getattr rpc requests for it on every stat.
2670 Frequency : Always, for filesystems larger than 2TB
2672 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2673 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2674 filesystems larger than 2TB to be mounted with mballoc enabled.
2677 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2679 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2680 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2681 filesystems larger than 2TB is mounted then the kernel will
2682 silently corrupt the start of the filesystem. CONFIG_LBD is
2683 enabled for all CFS-supported kernels, but the possibility of
2684 this happening with a modified kernel config exists.
2686 Severity : enhancement
2688 Description: add client O_DIRECT support for 2.6 kernels
2689 Details : It is now possible to do O_DIRECT reads and writes to files
2690 in the Lustre client mountpoint on 2.6 kernel clients.
2692 Severity : enhancement
2694 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2695 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2696 parallel, not waiting for response from every OST before sending
2697 a rpc to the next OST.
2702 Description: setattr vs write race when updating file timestamps
2703 Details : Client processes that update a file timestamp into the past
2704 right after writing to the file (e.g. tar) it is possible that
2705 the updated file modification time can be reset to the current
2706 time due to a race between processing the setattr and write RPC.
2708 Severity : enhancement
2710 Description: Bring 'lfs find' closer in line with regular Linux find.
2711 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2712 -print0 options and obtains all the needed info through the lustre
2715 Severity : enhancement
2717 Description: support up to 1024 configured devices on one node
2718 Details : change obd_dev array from statically allocated to dynamically
2719 allocated structs as they are first used to reduce memory usage
2724 Description: Flush dirty partially truncated pages during truncate
2725 Details : Immediatelly flush partially truncated pages in filter_setattr,
2726 this way we completely avoid having any pages in page cache on OST
2727 and can retire ugly workarounds during writes to flush such pages.
2732 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2733 Details : protect inode from truncation within vfs_unlink() context
2734 just take a reference before calling vfs_unlink() and release it
2735 when parent's i_sem is free.
2740 Description: last_id value checked outside lock on OST caused LASSERT failure
2741 Details : If there were multiple MDS->OST object precreate requests in
2742 flight, it was possible that the OST's last object id was checked
2743 outside a lock and incorrectly tripped an assertion. Move checks
2744 inside locks, and discard old precreate requests.
2747 Frequency : always, if extents are used on OSTs
2749 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2750 Details : OSTs using the extents format would not zero the high 16 bits of
2751 the index physical block number. This is not a problem for any
2752 OST filesystems smaller than 16TB, and no kernels support ext3
2753 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2754 new/modified files) and can be fixed for existing filesystems
2755 with e2fsprogs-1.39-cfs1.
2760 Description: import connection selection may be incorrect if timer wraps
2761 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2762 import connections to be ignored if the 32-bit jiffies counter
2763 wraps. Use a 64-bit jiffies counter.
2766 Frequency : very large clusters immediately after boot
2768 Description: LNET request buffers exhausted under heavy short-term load
2769 Details : If a large number of client requests are generated on a service
2770 that has previously never seen so many requests it is possible
2771 that the request buffer growth cannot keep up with the spike in
2772 demand. Instead of dropping incoming requests, they are held in
2773 the LND until the RPC service can accept more requests.
2776 Frequency : Sometimes during replay
2778 Description: Assertion failure in ll_local_open after replay.
2779 Details : If replay happened on an open request reply before we were able
2780 to set replay handler, reply will become not swabbed tripping the
2781 assertion in ll_local_open. Now we set the handler right after
2782 recognising of open request
2785 Frequency : very rare
2787 Description: kernel reports "badness in vsnprintf"
2788 Details : Reading from the "recovery_status" /proc file in small chunks
2789 may cause a negative length in lprocfs_obd_rd_recovery_status()
2790 call to vsnprintf() (which is otherwise harmless). Exit early
2791 if there is no more space in the output buffer.
2793 Severity : enhancement
2795 Description: clear OBD RPC statistics by writing to them
2796 Details : It is now possible to clear the OBD RPC statistics by writing
2797 to the "stats" file.
2802 Description: Client mtime is not the same on different clients after utimes
2803 Details : In some cases, the client was using the utimes() syscall on
2804 a file cached on another node. The clients now validate the
2805 ctime from the MDS + OSTs to determine which one is right.
2810 Description: Inability to activate failout mode
2811 Details : lconf script incorrectly assumed that in pythong string's numeric
2812 value is used in comparisons.
2815 Frequency : always with multiple stripes per file
2817 Description: Inefficient object allocation for mutli-stripe files
2818 Details : When selecting which OSTs to stripe files over, for files with
2819 a stripe count that divides evenly into the number of OSTs,
2820 the MDS is always picking the same starting OST for each file.
2821 Return the OST selection heuristic to the original design.
2826 Description: mount failures may take full timeout to return an error
2827 Details : Under some heavy load conditions it is possible that a
2828 failed mount can wait for the full obd_timeout interval,
2829 possibly several minutes, before reporting an error.
2830 Instead return an error as soon as the status is known.
2832 Frequency : quota enabled and large files being deleted
2834 Description: releasing more than 4GB of quota at once hangs OST
2835 Details : If a user deletes more than 4GB of files on a single OST it
2836 will cause the OST to spin in an infinite loop. Release
2837 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2842 Description: statfs data retrieved from /proc may be stale or zero
2843 Details : When reading per-device statfs data from /proc, in the
2844 {kbytes,files}_{total,free,avail} files, it may appear
2845 as zero or be out of date.
2848 Frequency : systems with MD RAID1 external journal devices
2850 Description: lconf's call to blkid is confused by RAID1 journal devices
2851 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2852 of returning all block devices that match the journal UUID.
2855 Frequency : always, for aggregate stripe size over 4GB
2857 Description: assertion fails when trying to use 4GB stripe size
2858 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2859 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2862 Frequency : always on ppc64
2864 Description: the first write on an ext3 filesystem with mballoc got stuck
2865 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2866 perform endianness conversion.
2868 ------------------------------------------------------------------------------
2870 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2872 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2873 INTEROPERATE with older versions automatically. Please read the
2874 user documentation before upgrading any part of a live system.
2875 * WARNING: Lustre networking configuration changes are required with
2876 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2879 * Support for kernels:
2880 2.6.9-22.0.2.EL (RHEL 4)
2881 2.6.5-7.244 (SLES 9)
2882 2.6.12.6 vanilla (kernel.org)
2885 Severity : enhancement
2886 Bugzilla : 7981/8208
2887 Description: Introduced Lustre Networking (LNET)
2888 Details : LNET is new networking infrastructure for Lustre, it includes
2889 a reorganized network configuration mode (see the user
2890 documentation for full details) as well as support for routing
2891 between different network fabrics. Lustre Networking Devices
2892 (LNDs) for the supported network fabrics have also been
2893 created for this new infrastructure.
2895 Severity : enhancement
2896 Description: Introduced Access control lists
2897 Details : clients can set ACLs on files and directories in order to have
2898 more fine-grained permissions than the standard Unix UGO+RWX.
2899 The MDS must be started with the "-o acl" mount option.
2901 Severity : enhancement
2902 Description: Introduced filesystem quotas
2903 Details : Administrators may now establish per-user quotas on the
2906 Severity : enhancement
2908 Description: Configuration change for the XT3
2909 The PTLLND is now used to run Lustre over Portals on the XT3
2910 The configure option(s) --with-cray-portals are no longer used.
2911 Rather --with-portals=<path-to-portals-includes> is used to
2912 enable building on the XT3. In addition to enable XT3 specific
2913 features the option --enable-cray-xt3 must be used.
2918 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2919 Details : A race between allocating a new llcd and re-getting the llcd_lock
2920 allowed another thread to grab newly-allocated llcd.
2922 Severity : enhancement
2924 Description: 2.6 OST async journal commit and locking fix to improve performance
2925 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2926 2.6 kernels are now async as they already were in 2.4 kernels so
2927 that they can commit concurrently with the network bulk transfer.
2928 For block-allocated files the filter allocation semaphore is held
2929 to avoid filesystem fragmentation during allocation. BKL lock
2930 removed for 2.6 xattr operations where it is no longer needed.
2935 Description: lconf incorrectly determined whether two IP networks could talk
2936 Details : In some more complicated routing and multiple-network
2937 configurations, lconf will avoid trying to make a network
2938 connection to a disjoint part of the IP space. It was doing the
2939 math incorrectly for one set of cases.
2944 Description: Fix for potential infinite loop processing records in an llog.
2945 Details : If an llog record is corrupted/zeroed, it is possible to loop
2946 forever in llog_process(). Validate the llog record length
2947 and skip the remainder of the block on error.
2950 Frequency : occasional (liblustre only)
2952 Description: liblustre could not open files whose last component is a symlink
2953 Details : sysio_path_walk() would incorrectly pass the open intent to
2954 intermediate path components.
2957 Frequency : rare (liblustre only with non-standard tuning)
2958 Bugzilla : 7201 (7350)
2959 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2960 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2961 after it has been released, only do this for referenced locks
2963 Severity : enhancement
2965 Description: specifying an (invalid) directory default stripe_size of -1
2966 would reset the directory default striping
2967 Details : stripe_size -1 was used internally to signal directory stripe
2968 removal, now use "all default" to signal dir stripe removal
2969 as a directory striping of "all default" is not useful
2972 Frequency : common for large clusters running liblustre clients
2974 Description: doing an ls when liblustre clients are running is slow
2975 Details : sending a glimpse AST to a liblustre client waits for every AST
2976 to time out, as liblustre clients will not respond. Since they
2977 cannot cache data we refresh the OST lock LVB from disk instead.
2979 Severity : enhancement
2981 Description: doing an ls at the same time as file IO can be slow
2982 Details : enqueue and other "small" requests can be blocked behind many
2983 large IO requests. Create a new OST IO portal for non-IO
2984 requests so they can be processed faster.
2987 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2989 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2990 created file with mode 000
2991 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2992 allow an ftruncate() as a non-root user to a file with mode 000.
2993 Lustre now respects this flag to disable mode checks when
2994 truncating a file owned by the user
2997 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3000 Description: Revoking locks from clients that went dead or catatonic might take
3002 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3003 cancellation of such locks instant on servers without waiting for
3004 any reply from clients. Clients drops these locks when cancel
3005 notification from server is received without replying.
3008 Frequency : liblustre-only, when liblustre client dies or becomes busy
3010 Description: Doing ls on Linux clients can take a long time with active
3012 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3013 granting such locks to it in the first place if possible. Locks
3014 are taken by proxy on the OST during the read or write and
3015 dropped immediately afterward. Add connect flags handling, do
3016 not grant locks to liblustre clients for glimpse ASTs.
3018 Severity : enhancement
3020 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3021 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3022 sequential reads and behaves badly with applications doing random
3023 reads. Improve it by reading ahead at least read region, and
3024 avoiding excessive large RPC for small reads.
3026 Severity : enhancement
3028 Description: Creating more than 1000 files for a single job may cause a load
3029 imbalance on the OSTs if there are also a large number of OSTs.
3030 Details : qos_prep_create() uses an OST index reseed value that is an
3031 even multiple of the number of available OSTs so that if the
3032 reseed happens in the middle of the object allocation it will
3033 still utilize the OSTs as uniformly as possible.
3038 Description: OST or MDS may oops in ping_evictor_main()
3039 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3040 but doesn't restart at beginning of obd_exports_timed list
3043 Severity : enhancement
3045 Description: improve by-nid export eviction on the MDS and OST
3046 Details : allow multiple exports with the same NID to be evicted at one
3047 time without re-searching the exports list.
3050 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3052 Description: MDS may oops in groups_free()
3053 Details : in rare race conditions a newly allocated group_info struct is
3054 freed again, and this can be NULL. The 2.4 compatibility code
3055 for groups_free() checked for a NULL pointer, but 2.6 did not.
3058 Frequency : common for liblustre clients doing little filesystem IO
3059 Bugzilla : 9352, 7313
3060 Description: server may evict liblustre clients accessing contended locks
3061 Details : if a client is granted a lock or receives a completion AST
3062 with a blocking AST already set it would not reply to the AST
3063 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3066 Frequency : lfs setstripe, only systems with more than 160 OSTs
3068 Description: unable to set striping with a starting offset beyond OST 160
3069 Details : llapi_create_file() incorrectly limited the starting stripe
3070 index to the maximum single-file stripe count.
3073 Frequency : LDAP users only
3075 Description: lconf did not handle in-kernel recovery with LDAP properly
3076 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3078 Severity : enhancement
3080 Description: bind OST threads to NUMA nodes to improve performance
3081 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3082 node and do their allocations there to localize memory access
3084 Severity : enhancement
3086 Description: llmount can determine client NID directly from Myrinet (GM)
3087 Details : the client NID code from gmnalnid was moved directly into
3088 llmount, removing the need to use this or specifying the
3089 client NID explicitly when mounting GM clients with zeroconf
3092 Frequency : if client is started with down MDS
3094 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3095 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3096 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3097 ptlrpc_import_delayed_req() because we didn't check if the
3098 request was interrupted, and we also didn't break out of the
3099 event loop if there was no timeout
3104 Description: data loss during non-page-aligned writes to a single file from
3105 both multiple nodes and multiple threads on one node at same time
3106 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3107 inconsistency led to false short-reads, that were cached and later
3108 used by ->prepare_write() to fill in partially written page,
3109 leading to data loss.
3112 Frequency : always, if lconf --abort_recovery used
3114 Description: lconf --abort_recovery fails with 'Operation not supported'
3115 Details : lconf was attempting to abort recovery on the MDT device and not
3118 Severity : enhancement
3120 Description: remove cleanup logs
3121 Details : replace lconf-generated cleanup logs with lustre internal
3122 cleanup routines. Eliminates the need for client-cleanup and
3125 Severity : enhancement
3127 Description: add support for EAs (user and system) on lustre filesystems
3128 Details : it is now possible to store extended attributes in the Lustre
3129 client filesystem, and with the user_xattr mount option it
3130 is possible to allow users to store EAs on their files also
3132 Severity : enhancement
3134 Description: Add possibility (config option) to show minimal available OST free
3136 Details : When compiled with --enable-mindf configure option, statfs(2)
3137 (and so, df) will return least minimal free space available from
3138 all OSTs as amount of free space on FS, instead of summary of
3139 free spaces of all OSTs.
3141 Severity : enhancement
3143 Description: do not expand extent locks acquired on OST-side
3144 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3145 by server: they are not cached anyway.
3148 Frequency : when mmap is used/binaries executed from Lustre
3150 Description: Unmmap pages before throwing them away from read cache.
3151 Details : llap_shrink cache now attempts to unmap pages before discarding
3152 them (if unmapping failed - do not discard). SLES9 kernel has
3153 extra checks that trigger if this unmapping is not done first.
3158 Description: lconf didn't resolve symlinks before checking to see whether a
3159 given mountpoint was already in use
3162 Frequency : when migrating failover services
3163 Bugzilla : 6395, 9514
3164 Description: When migrating a subset of services from a node (e.g. failback
3165 from a failover service node) the remaining services would
3166 time out and evict clients.
3167 Details : lconf --force (implied by --failover) sets the global obd_timeout
3168 to 5 seconds in order to quickly disconnect, but this caused
3169 other RPCs to time out too quickly. Do not change the global
3170 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3172 Severity : enhancement
3173 Frequency : if MDS is started with down OST
3174 Bugzilla : 9439,5706
3175 Description: Allow startup/shutdown of an MDS without depending on the
3176 availability of the OSTs.
3177 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3178 Add appropriate locking and lov-osc refcounts for safe
3179 cleaning. Add osc abort_inflight calls in case the
3180 synchronize never started.
3183 Frequency : occasional (Cray XT3 only)
3185 Description: root not authorized to access files in CRAY_PORTALS environment
3186 Details : The client process capabilities were not honoured on the MDS in
3187 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3188 been accessed by an authorized user then root was able to access
3189 the file on the local client also. The root user capabilities
3190 are now allowed on the MDS, as this environment has secure UID.
3193 Frequency : occasional
3195 Description: ldiskfs "too long searching" message happens too often
3196 Details : A debugging message (otherwise harmless) prints too often on
3197 the OST console. This has been reduced to only happen when
3198 there are fragmentation problems on the filesystem.
3203 Description: Division by zero in statfs when all OSCs are inactive
3204 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3205 lov_statfs divides by value returned by lov_get_stripecnt().
3209 Bugzilla : 9489, 3273
3210 Description: First write from each client to each OST was only 4kB in size,
3211 to initialize client writeback cache, which caused sub-optimal
3212 RPCs and poor layout on disk for the first writen file.
3213 Details : Clients now request an initial cache grant at (re)connect time
3214 and so that they can start streaming writes to the cache right
3215 away and always do full-sized RPCs if there is enough data.
3216 If the OST is rebooted the client also re-establishes its grant
3217 so that client cached writes will be honoured under the grant.
3222 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3223 Details : Now I/O RPCs go to different portal number and (presumably) fast
3224 lock requests (and glimses) and other RPCs get their own service
3225 threads pool that should be able to service those RPCs
3228 Severity : enhancement
3230 Description: Ability to exchange lustre version between client and servers and
3231 issue warnings at client side if client is too old. Also for
3232 liblustre clients there is ability to refuse connection of too old
3234 Details : New 'version' field is added to connect data structure that is
3235 filled with version info. That info is later checked by server and
3239 Frequency : rare, liblustre only.
3240 Bugzilla : 9296, 9581
3241 Description: Two simultaneous writes from liblustre at offset within same page
3242 might proceed at the same time overwriting eachother with stale
3244 Details : I/O lock withing llu_file_prwv was released too early, before data
3245 actually was hitting the wire. Extended lock-holding time until
3246 server acknowledges receiving data.
3249 Frequency : extremely rare. Never observed in practice.
3251 Description: avoid generating lustre_handle cookie of 0.
3252 Details : class_handle_hash() generates handle cookies by incrementing
3253 global counter, and can hit 0 occasionaly (this is unlikely, but
3254 not impossible, because initial value of cookie counter is
3255 selected randonly). Value of 0 is used as a sentinel meaning
3256 "unassigned handle" --- avoid it. Also coalesce two critical
3257 sections in this function into one.
3259 Severity : enhancement
3261 Description: allow liblustre clients to delegate truncate locking to OST
3262 Details : To avoid overhead of locking, liblustre client instructs OST to
3263 take extent lock in ost_punch() on client's behalf. New connection
3264 flag is added to handle backward compatibility.
3266 Severity : enhancement
3267 Bugzilla : 4928, 7341, 9758
3268 Description: allow number of OST service threads to be specified
3269 Details : a module parameter allows the number of OST service threads
3270 to be specified via "options ost ost_num_threads={N}" in the
3271 OSS's /etc/modules.conf or /etc/modprobe.conf.
3275 Bugzilla : 6146, 9635, 9895
3276 Description: servers crash with bad pointer in target_handle_connect()
3277 Details : In rare cases when a client is reconnecting it was possible that
3278 the connection request was the last reference for that export.
3279 We would temporarily drop the export reference and get a new
3280 one, but this may have been the last reference and the export
3281 was just destroyed. Get new reference before dropping old one.
3283 Severity : enhancement
3284 Frequency : if client is started with failover MDS
3286 Description: Allow multiple MDS hostnames in the mount command
3287 Details : Try to read the configuration from all specified MDS
3288 hostnames during a client mount in case the "primary"
3291 Severity : enhancement
3293 Description: Stop sending data to evicted clients as soon as possible.
3294 Details : Check if the client we are about to send or are sending data to
3295 was evicted already. (Check is done every second of waiting,
3296 for which l_wait_event interface was extended to allow checking
3297 of exit condition at specified intervals).
3300 Frequency : rare, normally only when NFS exporting is done from client
3302 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3304 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3307 Frequency : always, if srand() is called before liblustre initialization
3309 Description: Liblustre uses system PRNG disturbing its usage by user application
3310 Details : Introduce internal to lustre fast and high-quality PRNG for
3311 lustre usage and make liblustre and some other places in generic
3312 lustre code to use it.
3314 Severity : enhancement
3315 Bugzilla : 9477, 9557, 9870
3316 Description: Verify that the MDS configuration logs are updated when xml is
3317 Details : Check if the .xml configuration logs are newer than the config
3318 logs stored on the MDS and report an error if this is the case.
3319 Request --write-conf, or allow starting with --old_conf.
3321 Severity : enhancement
3323 Description: Handle symlinks in the path when checking if Lustre is mounted.
3324 Details : Resolve intermediate symlinks when checking if a client has
3325 mounted a filesystem to avoid duplicate client mounts.
3330 Description: lconf can hit an error exception but still return success.
3331 Details : The lconf command catches the Command error exception at the top
3332 level script context and will exit with the associated exit
3333 status, but doesn't ensure that this exit status is non-zero.
3338 Description: failure of ptlrpc thread startup can cause oops
3339 Details : Starting a ptlrpc service thread can fail if there are a large
3340 number of threads or the server memory is very fragmented.
3341 Handle this without oopsing.
3344 Frequency : always, only if liblustre and non-default acceptor port was used
3346 Description: liblustre cannot connect to servers with non-default acceptor port
3347 Details : tcpnal_set_default_params() was not called and was therefore
3348 ignoring the environment varaible TCPNAL_PORT, as well as other
3349 TCPNAL_ environment variables
3354 Description: two objects could be created on the same OST for a single file
3355 Details : If an OST is down, in some cases it was possible to create two
3356 objects on a single OST for a single file. No problems other
3357 than potential performance impact and spurious error messages.
3361 Bugzilla : 5681, 9562
3362 Description: Client may oops in ll_unhash_aliases
3363 Details : Client dcache may become inconsistent in race condition.
3364 In some cases "getcwd" can fail if the current directory is
3370 Description: Inode refcounting problems in NFS export code
3371 Details : link_raw functions used to call d_instantiate without obtaining
3372 extra inode reference first.
3376 Bugzilla : 9942, 9903
3377 Description: Referencing freed requests leading to crash, memleaks with NFS.
3378 Details : We used to require that call to ll_revalidate_it was always
3379 followed by ll_lookup_it. Also with revalidate_special() it is
3380 possible to call ll_revalidate_it() twice for the same dentry
3381 even if first occurence returned success. This fix changes semantic
3382 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3383 reference on a request referred from the intent.
3384 ll_intent_release() then releases such a request.
3387 Frequency : rare, normally benchmark loads only
3389 Description: unlinked inodes were kept in memory on the client
3390 Details : If a client is repeatedly creating and unlinking files it
3391 can accumulate a lot of stale inodes in the inode slab cache.
3392 If there is no other client load running this can cause the
3393 client node to run out of memory. Instead flush old inodes
3394 from client cache that have the same inode number as a new inode.
3397 Frequency : SLES9 2.6.5 kernel and long filenames only
3398 Bugzilla : 9969, 10379
3399 Description: utime reports stale NFS file handle
3400 Details : SLES9 uses out-of-dentry names in some cases, which confused
3401 the lustre dentry revalidation. Change it to always use the
3405 Frequency : rare, unless heavy write-truncate concurrency is continuous
3406 Bugzilla : 4180, 6984, 7171, 9963, 9331
3407 Description: OST becomes very slow and/or deadlocked during object unlink
3408 Details : filter_destroy() was holding onto the parent directory lock
3409 while truncating+unlinking objects. For very large objects this
3410 may block other threads for a long time and slow overall OST
3411 responsiveness. It may also be possible to get a lock ordering
3412 deadlock in this case, or run out of journal credits because of
3413 the combined truncate+unlink. Solution is to do object truncate
3414 first in one transaction without parent lock, and then do the
3415 final unlink in a new transaction with the parent lock. This
3416 reduces the lock hold time dramatically.
3419 Frequency : rare, 2.4 kernels only
3421 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3422 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3423 before unmounting their filesystem in order to avoid deadlock.
3424 The kernel_locked() function in 2.4 kernels only checks whether
3425 the kernel lock is held, not whether it is this process that is
3426 holding it as 2.6 kernels do.
3431 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3432 Details : The client ptlrpc code may be trying to reconnect to a down
3433 server before a previous connection attempt has timed out.
3434 Increase the reconnect interval to be longer than the connection
3435 timeout interval to avoid sending duplicate connections to
3439 Frequency : echo_client brw_test command
3441 Description: fix echo_client to work with OST preallocated code
3442 Details : OST preallocation code (5137) didn't take echo_client IO path
3443 into account: echo_client calls filter methods outside of any
3444 OST thread and, hence, there is no per-thread preallocated
3445 pages and buffers to use. Solution: hijack pga pages for IO. As
3446 a byproduct, this avoids unnecessary data copying.
3450 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3451 Description: Client can oops in mdc_commit_close() after open replay
3452 Details : It was possible for the MDS to return an open request with no
3453 transaction number in mds_finish_transno() if the client was
3454 evicted, but without actually returning an error. Clients
3455 would later try to replay that open and may trip an assertion
3456 Simplify the client close codepath, and always return an error
3457 from the MDS in case the open is not successful.
3460 Frequency : rare, 2.6 OSTs only
3462 Description: OST may deadlock under high load on fragmented files
3463 Details : If there was a heavy load and highly-fragmented OST filesystems
3464 it was possible to have all the OST threads deadlock waiting on
3465 allocation of biovecs, because the biovecs were not released
3466 until the entire RPC IO was completed. Instead, release biovecs
3467 as soon as they are complete to ensure forward IO progress.
3469 Severity : enhancement
3471 Description: Support for specifying external journal device at mount
3472 Details : If an OST or MDS device is formatted with an external journal
3473 device, this device major/minor is stored in the ext3 superblock
3474 and may not be valid for failover. Allow detecting and
3475 specifying the external journal at mount time.
3480 Description: Mounting an MDS with pending unlinked files may cause oops
3481 Details : target_finish_recovery() calls mds_postrecov() which returned
3482 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3483 considers this an error and immediately begins cleaning up the
3484 lov, just after starting the mds_lov process
3486 Severity : enhancement
3488 Description: Implement 'lfs df' to report actual free space on per-OST basis
3489 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3490 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3491 usage of objects; '-h' to report in human readable format.
3493 ------------------------------------------------------------------------------
3495 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3502 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3503 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3504 at startup during orphan unlinking. Instead, skip buddy update
3505 before setup, as it will be regenerated after recovery is complete.
3510 Description: If an OST is inactive, its locks might reference stale inodes.
3511 Details : lov_change_cbdata() must iterate over all namespaces, even if
3512 they are inactive to clear inode references from the lock.
3514 Severity : enhancement
3515 Frequency : occasional, if non-standard max_dirty_mb used
3517 Description: Client will block write RPCs if not enough grant
3518 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3519 then the client will block writes while waiting for another RPC
3520 to complete instead of consuming its dirty limit. With change
3521 we get improved performance when max_dirty_mb is small.
3523 Severity : enhancement
3524 Bugzilla : 3389, 6253
3525 Description: Add support for supplementary groups on the MDS.
3526 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3527 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3528 lookups for user supplementary groups into a cache.
3532 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3533 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3534 for liblustre O_EXCL is enforced by MDS.
3538 Description: readdir never returns NULL in liblustre.
3539 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3540 dentry in struct dirent.
3544 Frequency : liblustre only, depends on application IO pattern
3545 Description: liblustre clients evicted if not contacting servers
3546 Details : Don't put liblustre clients into the ping_evictor list, so
3547 they will not be evicted by the pinger ever.
3549 Severity : enhancement
3551 Description: Add ability to evict clients by NID from MDS.
3552 Details : By echoing "nid:$NID" string into
3553 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3554 $NID would be instantly evicted from this MDS and from all active
3555 OSTs connected to it.
3559 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3560 Details : lookup_it_finish() used to query file size from OSTs that was not
3565 Description: service threads change working directory to that of init
3566 Details : Starting lustre service threads may pin the working directory
3567 of the parent thread, making that filesystem busy. Threads
3568 now change to the working directory of init to avoid this.
3572 Frequency : during shutdown only
3573 Description: shutdown with a failed MDS or OST can cause unmount to hang
3574 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3579 Frequency : 2.6 only
3580 Description: chmod/chown may include an extra supplementary group
3581 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3582 supplementary group and if none is specified this is used.
3585 Bugzilla : 5479 (6816)
3587 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3588 Details : If lookup is in progress on a file that is unlinked we might try
3589 to revalidate the inode and fail in revalidate after lookup is
3590 complete and ll_file_open() enqueues the open again but
3591 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3594 Frequency : always, if lconf --abort_recovery used
3596 Description: lconf --abort_recovery fails with 'Operation not supported'
3597 Details : lconf was attempting to abort recovery on the MDT device and not
3600 ------------------------------------------------------------------------------
3602 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3607 Frequency : rare (only unsupported configurations with a node running as an
3609 Bugzilla : 6514, 5137
3610 Description: Mounting a Lustre file system on a node running as an OST could
3612 Details : OSTs now preallocates memory needed to write out data at
3613 startup, instead of when needed, to avoid having to
3614 allocate memory in possibly low memory situations.
3615 Specifically, if the file system is mounted on on OST,
3616 memory pressure could force it to try to write out data,
3617 which it needed to allocate memory to do. Due to the low
3618 memory, it would be unable to do so and the node would
3619 become unresponsive.
3621 Severity : enhancement
3623 Description: Addition of lconf --service command line option
3624 Details : lconf now accepts a '--service <arg>' option, which is
3625 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3627 Severity : enhancement
3629 Description: Failover mode is now the default for OSTs.
3630 Details : By default, OSTs will now run in failover mode. To return to
3631 the old behaviour, add '--failout' to the lmc line for OSTs.
3633 Severity : enhancement
3635 Description: Health checks are now provided for MDS and OSTs
3636 Details : Additional detailed health check information on MSD and OSTs
3637 is now provided through the procfs health_check value.
3640 Frequency : occasional, depends on IO load
3642 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3643 after numerous create/delete cycles
3644 Details : The ext3 inode allocation policy would not allocate new inodes
3645 very well on the OSTs because there are no new directories
3646 being created. Instead we look for groups with free space if
3647 the parent directories are nearly full.
3652 Description: Network or server problems during mount may cause partially
3653 mounted clients instead of returning an error.
3654 Details : The config llog parsing code may overwrite the error return
3655 code during mount error handling, returning success instead
3661 Description: MDS can fail to allocate large reply buffers
3662 Details : After long uptimes the MDS can fail to allocate large reply
3663 buffers (e.g. zconf client mount config records) due to memory
3664 fragmentation or consumption by the buffer cache. Preallocate
3665 some large reply buffers so that these replies can be sent even
3666 under memory pressure.
3670 Frequency : rare (liblustre)
3671 Description: fsx running with liblustre complained that using truncate() to
3672 extend the file doesn't work. This patch corrects that issue.
3673 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3674 ATTR_SIZE and lsm use in llu_setattr_raw.
3678 Frequency : rare, only 2.6 kernels
3679 Description: Unusual file access patterns on the MDS may result in inode
3680 data being lost in very rare circumstances.
3681 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3682 code in the 2.6 kernel caused the inode and/or EA data not to
3683 be read from disk, causing single-file corruption.
3687 Frequency : rare, only 2.6 filesystems using extents
3688 Description: Heavy concurrent write and delete load may cause data corruption.
3689 Details : It was possible under high-load situations to have an extent
3690 metadata block in the block device cache from a just-unlinked
3691 file overwrite a newly-allocated data block. We now unmap any
3692 metadata buffers that alias just-allocated data blocks.
3696 Frequency : filesystems with default stripe_count larger than 77
3697 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3698 Details : lconf specifies an inode size of 4096 bytes when the default
3699 stripe_count is larger than 77. This conflicts with the default
3700 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3701 this case to avoid pinning too much memory for large EAs.
3703 ------------------------------------------------------------------------------
3705 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3710 Frequency : rare (extremely heavy IO load with hundreds of clients)
3712 Description: Client is evicted, gets IO error writing to file
3713 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3714 caused clients to be evicted for AST timeouts. The fixes in
3715 bug 5192 mean we no longer need to have such short AST timeouts
3716 so ldlm_timeout has been increased.
3719 Frequency : occasional during --force or --failover shutdown under load
3720 Bugzilla : 5949, 4834
3721 Description: Server oops/LBUG if stopped with --force or --failover under load
3722 Details : a collection of import/export refcount and cleanup ordering
3723 issues fixed for safer force cleanup
3726 Frequency : only filesystems larger than 120 OSTs
3727 Bugzilla : 5990, 6223
3728 Description: lfs getstripe would oops on a very large filesystem
3729 Details : lov_getconfig used kfree on vmalloc'd memory
3732 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3733 Bugzilla : 6242, 6243
3734 Description: reading from files that had been truncated to a non-zero size
3735 but never opened returned no data
3736 Details : ll_file_read() reads zeros from no-object files to EOF
3741 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3742 Details : There's little harm in aborting MDS/OSS recovery and letting it
3743 try again, so I removed the LASSERT and return an error instead.
3745 Severity : enhancement
3747 Description: New debugging infrastructure for tracking down data corruption
3748 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3749 (b) cover more of the client-side code path, and (c) try to narrow
3750 down where problems occurred
3754 Bugzilla : 3819, 4364, 4397, 6313
3755 Description: Racing close and eviction MDS could cause assertion in mds_close
3756 Details : It was possible to get multiple mfd references during close and
3757 client eviction, leading to one thread referencing a freed mfd.
3759 Severity: : enhancement
3760 Bugzilla : 3262, 6359
3761 Description: Attempts to reconnect to servers are now more aggressive.
3762 Details : This builds on the enhanced upcall-less recovery that was added
3763 in 1.4.2. When trying to reconnect to servers, clients will
3764 now try each server in the failover group every 10 seconds. By
3765 default, clients would previously try one server every 25 seconds.
3770 Description: After recovery, certain operations trigger a failed
3771 assertion on a client.
3772 Details : Failing over an mds, using lconf -d --failover, while a
3773 client was doing a readdir() call would cause the client to
3774 LBUG after recovery completed and the readdir() was resent.
3776 Severity : enhancement
3778 Description: Default groups are now added by lconf
3779 Details : You can now run lconf --group <servicename> without having to
3780 manually add groups with lmc.
3783 Frequency : occasional
3785 Description: Nodes with an elan id of 0 trigger a failed assertion
3788 Frequency : always when accessing e.g. tty/console device nodes
3790 Description: tty and some other devices nodes cannot be used on lustre
3791 Details : file's private_data field is used by device data and lustre
3792 values in there got lost. New field was added to struct file to
3793 store fs-specific private data.
3796 Frequency : when exporting Lustre via NFS
3798 Description: NFSD failed occasionally when looking up a path component
3799 Details : NFSD is looking up ".." which was broken in ext3 directories
3800 that had grown large enough to become hashed.
3803 Frequency : Clusters with multiple interfaces not on the same subnet
3805 Description: Nodes will repeatedly try to reconnect to an interface which it
3806 cannot reach and report an error to the log.
3807 Details : Extra peer list entries will be created by lconf with some peers
3808 unreachable. lconf now validates the peer before adding it.
3811 Frequency : Only if a default stripe is set on the filesystem root.
3813 Description: Setting a default stripe on the filesystem root prevented the
3814 filesystem from being remounted.
3815 Details : The client was sending extra request flags in the root getattr
3816 request and did not allocate a reply buffer for the dir EA.
3819 Frequency : occasional, higher if lots of files are accessed by one client
3820 Bugzilla : 6159, 6097
3821 Description: Client trips assertion regarding lsm mismatch/magic
3822 Details : While revalidating inodes the VFS looks up inodes with ifind()
3823 and in rare cases can find an inode that is being freed.
3824 The ll_test_inode() code will free the lsm during ifind()
3825 when it finds an existing inode and then the VFS later attaches
3826 this free lsm to a new inode.
3830 Bugzilla : 6422, 7030
3831 Description: MDS deadlock between mkdir and client eviction
3832 Details : Creating a new file via mkdir or mknod (starting a transaction
3833 and getting the ns lock) can deadlock with client eviction
3834 (gets ns lock and trying to finish a synchronous transaction).
3837 Frequency : occasional
3838 Description: While starting a server, the fsfilt_ext3 module could not be
3840 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3841 kernels. Previously, lconf would still use the ext3 name
3842 when trying to load modules. Now, it will correctly use
3843 ext3 on 2.4 and ldiskfs on 2.6.
3845 Severity : enhancement
3846 Description: The default stripe count has been changed to 1
3847 Details : The interpretation of the default stripe count (0, to lfs
3848 or lmc) has been changed to mean striping across a single
3849 OST, rather than all available. For general usage we have
3850 found a stripe count of 1 or 2 works best.
3852 Severity : enhancement
3853 Description: Add support for compiling against Cray portals.
3854 Details : Conditional compiling for some areas that are different
3858 Frequency : occasional
3859 Bugzilla : 6409, 6834
3860 Description: Creating files with an explicit stripe count may lead to
3861 a failed assertion on the MDS
3862 Details : If some OSTs are full or unavailable, creating files may
3863 trigger a failed assertion on the MDS. Now, Lustre will
3864 try to use other servers or return an error to the
3868 Frequency : occasional
3870 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3871 caused lock timeout during readahead (since 1.4.2).
3872 Details : Processes doing readahead might match a lock that hasn't been
3873 granted yet if there are overlapping and conflicting lock
3874 requests. The readahead process waits on ungranted lock
3875 (original lock is CBPENDING), while OST waits for that process
3876 to cancel CBPENDING read lock and eventually evicts client.
3878 Severity : enhancement
3880 Description: Initial enabling of flock support for clients
3881 Details : Implements fcntl advisory locking and file status functions.
3882 This feature is provided as an optional mount flag (default
3883 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3884 locking are implemented yet, and those that are are not guaranteed
3885 to be completely correct in production environments.
3886 mount -t lustre -o [flock|noflock] ...
3889 Frequency : occasional
3891 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3892 assertion in the ext3 JBD (journaling) layer.
3893 Details : The b_committed_data struct is protected by the big kernel lock
3894 in 2.4 kernels, serializing journal_commit_transaction() and
3895 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3896 access to this struct. In 2.6 kernels there is finer grained
3897 locking to improve SMP performance of the JBD layer.
3901 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3906 Description: merge and cleanup kernel patches.
3909 -----------------------------------------------------------------------------
3911 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3913 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3914 versions of Lustre. Please update all servers and clients to
3915 version 1.4.2 or later at the same time. You must also run
3916 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3917 to update the configuration logs.
3919 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3920 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3921 - swab llog records of type '0' so we get proper header size/idx (5861)
3922 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3923 - fix rename of one directory over another leaking an inode (5953)
3924 - avoid SetPageDirty on 2.6 (5981)
3925 - don't re-add just-being-destroyed locks to the waiting list (5653)
3926 - when creating new directories, inherit the parent's custom
3927 striping settings if present parent (3048)
3928 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3929 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3930 - don't hold client locks on temporary worklist from l_lru (5666)
3931 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3932 - automatically evict dead clients (5921)
3933 - Update file size properly in create+truncate+fstat case (6196)
3934 - Do not unhash mountpoint dentries, do not allow removal of
3936 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3937 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3938 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3939 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3940 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3941 - Correctly handle reads of files with no objects (6243)
3942 - lctl recover will also mark a device active if deactivate used (5933)
3944 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3945 - allow --write-conf on an MDS with different nettype than client (5619)
3946 - don't write config llogs to MDS for mounts not from that MDS (5617)
3947 - lconf should create multiple TCP connections from a client (5201)
3948 - init scripts are now turned off by default; run chkconfig --on
3949 lustre and chkconfig --on lustrefs to use them
3950 - upcalls are no longer needed for clients to recover to failover
3952 - add --abort-recovery option to lconf to abort recovery on device
3954 - add support for an arbitrary number of OSTs (3026)
3955 - Quota support protocol changes.
3956 - forward compatibility changes to wire structs (6007)
3957 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3958 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3959 - improved "lustre status" script
3960 - initialize blocksize for non-regular files (6062)
3961 - added --disable-server and --disable-client configure options (5782)
3962 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3963 - Vanilla 2.4.29 support
3964 - increase maximum number of obd devices to 520 (6242)
3965 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3966 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3967 SLES 9 2.6.5 kernels
3968 - limit stripes per file to 160 (the maximum EA size) (6093)
3970 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3973 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3974 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3975 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3976 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3977 - small ext3 extents cleanups and fixes (5733)
3978 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3979 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3980 - avoid needless client->OST connect, fix handle mismatch (5317)
3981 - fix DLM error path that led to out-of-sync client, long delays (5779)
3982 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3983 - fix several locking issues related to i_size (5492,5624,5654,5672)
3984 - don't move pending lock onto export if it is already evicted (5683)
3985 - fix kernel oops when creating .foo in unlinked directory (5548)
3986 - fix deadlock in obdfilter statistics vs. object create (5811)
3987 - use time_{before,after} to avoid timer jiffies wrap (5882)
3988 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3989 - Do not leak request if server was not able to process it (5154)
3990 - If mds_open unable to find parent dir, make that negative lookup(5154)
3991 - don't create new directories with extent-mapping (5909, 5936)
3993 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3994 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3995 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3996 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3997 - support for RHEL4 kernel on i686 (5773)
3998 - provide error messages when incompatible logs are encountered (5898)
4000 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4001 * version 1.4.0.10 (1.4.1 release candidate 1)
4003 - don't keep a lock reference when lock is not granted (4238)
4004 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4005 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4006 - drop import inflight refcount on signal_completed_replay error (5255)
4007 - unlock page after async write error during send (3677)
4008 - handle missing objects in filter_preprw_read properly (5265)
4009 - no transno return for symlink open, don't save no-trasno open (3440)
4010 - don't try to complete elan receive that already failed (4012)
4011 - free RPC server reply state on error (5406)
4012 - clean up thread from ptlrpc_start_thread() on error (5160)
4013 - readahead could read extra page into cache that wasn't ejected (5388)
4014 - prevent races in class_attach/setup/cleanup/detach (5260)
4015 - don't dereference de->d_inode after l_dput of de (5458)
4016 - use "int" for stripe value returned from lock_to_stripe (5544)
4017 - mballoc allocation and error-checking fixes in 2.6 (5504)
4018 - block device patches to fix I/O request sizes in 2.6 (5482)
4019 - look up hostnames for IB nals (5602)
4020 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4021 - don't start multiple acceptors for the same port (5277)
4022 - fix incorrect LASSERT in mds_getattr_name (5635)
4023 - export a proc file for general "ping" checking (5628)
4024 - fix "lfs check" to not block when the MDS is down (5628)
4026 - service request history (4965)
4027 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4028 - create an "evict_client" /proc entry on OSTs, like the MDS has
4029 - fix mount usage message, return errors per mount(8) (5168)
4030 - change grep [] to grep "[]" in tests so they work in more UMLs
4031 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4032 - remove ancient LOV_MAGIC_V0 EA support (5047)
4033 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4034 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4035 - allow readahead allocations to fail when low on memory (5383)
4036 - mmap locking landed again, after considerable improvement (2828)
4037 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4039 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4042 - send OST transaction number in read/write reply to free req (4966)
4043 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4044 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4045 - the watchdog thread now runs as interruptible (5246)
4046 - flock/lockf fixes (but it's still disabled, pending 5135)
4047 - don't use EXT3 constants in llite code (5094)
4048 - memory shortage at startup could cause assertion (5176)
4050 - reorganization of lov code
4051 - single portals codebase
4053 - add extents/mballoc support (5025)
4054 - direct I/O reads in the obdfilter (4048)
4055 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4057 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4060 - send OST transaction number in read/write reply to free req (4966)
4061 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4062 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4063 - the watchdog thread now runs as interruptible (5246)
4064 - handle missing objects in filter_preprw_read properly (5265)
4065 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4066 - drop import inflight refcount on signal_completed_replay error (5255)
4067 - unlock page after async write error during send (3677)
4068 - return original error code on reconstructed replies (3761)
4069 - no transno return for symlink open, don't save no-trasno open (3440)
4071 - add pid to ldlm debugging output (4922)
4072 - bump the watchdog timeouts -- we can't handle 30sec yet
4073 - extra debugging for orphan dentry/inode bug (5259)
4075 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4078 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4079 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4080 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4081 - fix NULL dereference in /proc/sys/portals/routes (4827)
4082 - allow failed mdc_close() operations to be interrupted (4561)
4083 - stop precreate on OST before MDS would time out on it (4778)
4084 - don't send partial-page writes before EOF from client (4410)
4085 - discard client grant for sub-page writes on large-page clients (4520)
4086 - don't free dentries not owned by NFS code, check generation (4806)
4087 - fix lsm leak if mds_create_objects() fails (4801)
4088 - limit debug_daemon file size, always print CERROR messages (4789)
4089 - use transno after validating reply (3892)
4090 - process timed out requests if import state changes (3754)
4091 - update mtime on OST during writes, return in glimpse (4829)
4092 - add mkfsoptions to LDAP (4679)
4093 - use ->max_readahead method instead of zapping global ra (5039)
4094 - don't interrupt __l_wait_event() during strace
4096 - add software watchdogs to catch hung threads quickly (4941)
4097 - make lustrefs init script start after nfs is mounted
4098 - fix CWARN/ERROR duplication (4930)
4099 - return async write errors to application if possible (2248)
4100 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4101 - print NAL number in %x format (4645)
4102 - update barely-supported suse-2.4.21-171 series (4842)
4103 - support for sles 9 %post scripts
4104 - support for building 2.6 kernel-source packages
4105 - support for sles km_* packages
4107 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4110 - ignore -ENOENT errors in osc_destroy (3639)
4111 - notify osc create thread that OSC is being cleaned up (4600)
4112 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4113 - reconstruct ost_handle() like mds_handle() (4657)
4114 - create a new thread to do import eviction to avoid deadlock (3969)
4115 - let lconf resolve symlinked-to devices (4629)
4116 - don't unlink "objects" from directory with default EA (4554)
4117 - hold socknal file ref over connect in case target is down (4394)
4118 - allow more than 32000 subdirectories in a single directory (3244)
4119 - fix blocks count for O_DIRECT writes (3751)
4120 - OST returns ENOSPC from object create when no space left (4539)
4121 - don't send truncate RPC if file size isn't changing (4410)
4122 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4123 - bind to privileged port in socknal and tcpnal (3689)
4125 - rate limit CERROR/CWARN console message to avoid overload (4519)
4126 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4127 - basic mmap support (3918)
4128 - kernel patch series update from b1_4 (4711)
4130 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4133 - avoid crash during MDS cleanup with OST shut down (2775)
4134 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4135 - don't use bad inodes on the MDS (3744)
4136 - dynamic object preallocation to improve recovery speed (4236)
4137 - don't hold spinlock over lock dumping or change debug flags (4401)
4138 - don't zero obd_dev when it is force cleaned (3651)
4139 - print grants to console if they go negative (4431)
4140 - "lctl deactivate" will stop automatic recovery attempts (3406)
4141 - look for existing locks in ldlm_handle_enqueue() (3764)
4142 - don't resolve lock handle twice in recovery avoiding race (4401)
4143 - revalidate should check working dir is a directory (4134)
4145 - don't always mark "slow" obdfilter messages as errors (4418)
4147 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4150 - don't close LustreDB during write_conf until it is done (3860)
4151 - fix typo in lconf for_each_profile (3821)
4152 - allow dumping logs from multiple threads at one time (3820)
4153 - don't allow multiple threads in OSC recovery (3812)
4154 - fix debug_size parameters (3864)
4155 - fix mds_postrecov to initialize import for llog ctxt (3121)
4156 - replace config semaphore with spinlock (3306)
4157 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4158 - don't allow enqueue to complete on a destroyed export (3822)
4159 - down write_lock before checking llog header bitmap (3825)
4160 - recover from lock replay timeout (3764)
4161 - up llog sem before sending rpc (3652)
4162 - reduce ns lock hold times when setting kms (3267)
4163 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4164 - fix NULL deref and obd_dev leak on setup error (3312)
4165 - replace some LBUG about llog ops with error handling (3841)
4166 - don't match INVALID dentries from d_lookup and spin (3784)
4167 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4168 - fix invalid assertion in ptlrpc_set_wait (3880)
4170 - add libwrap support for the TCP acceptor (3996)
4171 - add /proc/sys/portals/routes for non-root route listing (3994)
4172 - allow setting MDS UUID in .xml (2580)
4173 - print the stack of a process that LBUGs (4228)
4175 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4178 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4179 - make sure to unset replay flag from failed open requests (3440)
4180 - if default stripe count is 0, use OST count for inode size (3636)
4181 - update parent mtime/ctime on client for create/unlink (2611)
4182 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4183 - free recovery state on server during a forced cleanup (3571)
4184 - unregister_reply for resent reqs (3063)
4185 - loop back devices mounting and status check on 2.6 (3563)
4186 - fix resource-creation race that can provoke i_size == 0 (3513)
4187 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4188 - more debugging for page-accounting assertion (3746)
4189 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4190 - avoid deadlock after precreation failure (3758)
4191 - fix race and lock order deadlock in orphan handling (3450, 3750)
4192 - add validity checks when grabbing inodes from l_ast_data (3599)
4194 - add /proc/.../recovery_status to obdfilter (3428)
4195 - lightweight CDEBUG infrastructure, debug daemon (3668)
4196 - change default OSC RPC parameters to be better on small clusters
4197 - turn off OST read cache for files smaller than 32MB
4198 - install man pages and include them in rpms (3100)
4199 - add new init script for (un)mounting lustre filesystems (2593)
4200 - run chkconfig in %post for init scripts (3701)
4201 - drop scimac NAL (unmaintained)
4203 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4206 - clean kiobufs before and after use (3485)
4207 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4208 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4209 - change init script to fail more clearly if not run as root (1528)
4210 - allow clients to reconnect during replay (1742)
4211 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4212 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4213 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4214 - fix cond_resched (3554)
4215 - only evict unfinished clients after recovery (3515)
4216 - allow bulk resend, prevent data loss (3570)
4217 - dynamic ptlrpc request buffer allocation (2102)
4218 - don't allow unlinking open directory if it isn't empty (2904)
4219 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4220 - remove extraneous obd dereference causing LASSERT failure (3334)
4221 - don't use get_cycles() when creating temp. files on the mds (3156)
4222 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4223 - handle EEXIST for set-stripe, set proper directory name (3336)
4225 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4226 - fix ksocknal_fmb_callback() error messages (2918)
4228 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4231 - don't copy lvb into (possibly NULL) reply on error (2983)
4232 - don't deref dentry after dput, don't free lvb on error (2922)
4233 - use the kms to determine writeback rpc length (2947)
4234 - increment oti_logcookies when osc is inactive (2948)
4235 - update client's i_blocks count via lvb messages (2543)
4236 - handle intent open/close of special files properly (1557)
4237 - mount MDS with errors=remount-ro, like obdfilter (2009)
4238 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4239 - don't use cancelling-locks' kms values (2947)
4240 - use highest lock extent for kms, not last one (2925)
4241 - don't dereference ERR_PTR() dentry in error handling path (3107)
4242 - fix thread race in portals_debug_dumplog() (3122)
4243 - create lprocfs device entries at setup instead of at attach (1519)
4244 - common AST error handler, don't evict client on completion race (3145)
4245 - zero nameidata in detach_mnt in 2.6 (3118)
4246 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4247 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4248 - initialize RPC timeout timer earlier for 2.6 (3219)
4249 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4250 - print nal/nid for unknown nid (3258)
4251 - additional checks for oscc recovery before doing precreate (3284)
4252 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4253 - don't crash in mdc_close for bad permissions on open (3285)
4254 - zero i_rdev for non-device files (3147)
4255 - clear page->private before handing to FS, better assertion (3119)
4256 - tune the read pipeline (3236)
4257 - fix incorrect decref of invalidated dentry (2350)
4258 - provide read-ahead stats and refine rpc in flight stats (3328)
4259 - don't hold journal transaction open across create RPC (3313)
4260 - update atime on MDS at close time (3265)
4261 - close LDAP connection when recovering to avoid server load (3315)
4262 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4263 - don't leak open file on MDS after open resend (3325)
4264 - serialize filter_precreate and filter_destroy_precreated (3329)
4265 - loop device shouldn't call sync_dev() for nul device (3092)
4266 - clear page cache after eviction (2766)
4267 - resynchronize MDS->OST in background (2824)
4268 - refuse to mount the same filesystem twice on same mountpoint (3394)
4269 - allow llmount to create routes for mounting behind routers (3320)
4270 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4271 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4272 - fix rare problem with rename on htree directories (3417)
4274 - allow default OST striping configuration per directory (1414)
4275 - fix compilation for qswnal for 2.6 kernels (3125)
4276 - increase maximum number of MDS request buffers for large systems
4277 - change liblustreapi to be useful for external progs like lfsck (3098)
4278 - increase local configuration timeout for slow disks (3353)
4279 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4281 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4284 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4285 - don't overwrite extent policy data in reply if lock was blocked (2901)
4286 - drop filter export grants atomically with removal from device (2663)
4287 - del obd_self_export from work_list in class_disconnect_exports (2908)
4288 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4289 - swab reply message in mdc_close, other PPC fixes (2464)
4290 - fix destroying of named logs (2325)
4291 - overwrite old logs when running lconf --write_conf (2264)
4292 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4293 - fix race in target_handle_connect (2898)
4294 - mds_reint_create() should take same inode create lock (2926)
4295 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4296 - don't close files for self_export to avoid uninitialized obd (2936)
4297 - allow MDS with the same name as client node (2939)
4298 - hold dentry reference for closed log files for unlink (2325)
4299 - reserve space for all logs during transactions (2059)
4300 - don't evict page beyond end of stripe extent (2925)
4301 - don't oops on a deleted current working directory (2399)
4302 - handle hard links to targets without a parent properly (2517)
4303 - don't dereference NULL lock when racing during eviction (2867)
4304 - don't grow lock extents when lots of conflicting locks (2919)
4306 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4309 - account for cache space usage on clients to avoid data loss (974)
4310 - lfsck support in lustre kernel code (2349)
4311 - reduce journal credits needed for BRW writes (2370)
4312 - orphan handling to avoid losing space on client/server crashes
4313 - ptlrpcd can be blocked, stopping ALL progress (2477)
4314 - use lock value blocks to assist in proper KMS, faster stat (1021)
4315 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4316 - recovery for initial connections (2355)
4317 - fixes for mds_cleanup_orphans (1934)
4318 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4319 - block all file creations until orphan recovery completes (1901)
4320 - client remove rq_connection from request struct (2423)
4321 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4322 - recovery timer race (2670)
4323 - mdc_close recovey bug (2532)
4324 - ptlrpc cleanup bug (2710)
4325 - mds timeout on local locks (2588)
4326 - namespace lock held during RPCs (2431)
4327 - handle interrupted sync write properly (2503)
4328 - don't try to handle a message that hasn't been replied to (2699)
4329 - client assert failure during cleanup after abort recovery (2701)
4330 - leak mdc device after failed mount (2712)
4331 - ptlrpc_check_set allows timedout requests to complete (2714)
4332 - wait for inflight reqs when ptlrpcd finishes (2710)
4333 - make sure unregistered services are removed from the srv_list
4334 - reset bulk XID's when resending them (caught by 1138 test)
4335 - unregister_bulk after timeout
4336 - fix lconf error (2694)
4337 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4338 - readahead locks pages, leaves pending causing memory pressure (2673)
4339 - increase OST request buffers to 4096 on large machines (2729)
4340 - fix up permission of existing directories in simple_mkdir (2661)
4341 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4342 - don't assign transno to errored transactions (2742)
4343 - don't delete objects on OST if given a bogus objid from MDS (2751)
4344 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4345 - if rq_no_resend, then timeout request after recovery (2432)
4346 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4347 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4348 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4349 - MDS recovery shouldn't delete objects that it has given out (2730)
4350 - if enqueue arrives after completion, don't clobber LVB (2819)
4351 - don't unlock pages twice when trigger_group_io returns error (2814)
4352 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4353 - don't write pages to disk if there was an error (1450)
4354 - don't ping imports that have recovery disabled (2676)
4355 - take buffered bytes into account when balancing socknal conn (2817)
4356 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4357 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4358 - remove little-endian swabbing of llog records (1987)
4359 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4360 - retry reposting request buffers if they fail (1191)
4361 - grow extent at grant time to avoid granting a revoked lock (2809)
4362 - lock revoke doesn't evict page if covered by a second lock (2765)
4363 - disable VM readahead to avoid reading outside lock extents (2805)
4365 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4366 - updated kernel patches for hp-2.4.20 kernel (2681)
4368 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4371 - fix truncated write corruption (2366)
4372 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4374 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4375 - ensure len(uuid) < 37 in lmc (1171)
4376 - fix ia64 OOPS in llog_test (2255)
4377 - zero end of page at obdfilter for partial page writes (2648)
4378 - don't leave stale dentries around after renames (bug 2428)
4379 - fix timeouts when evicting a client with a single lock held (2642)
4380 - set deadline for the initial HELLO message to drain (2634)
4381 - print out dotted-quad IP addresses in the socknal (2302)
4383 - additional debugging for MDS client eviction problem (2443)
4384 - fix mkfsoptions support for osts (2603, 2604)
4386 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4389 - add series for the vanilla 2.6.0 kernel
4390 - add series for the vanilla 2.4.24 kernel
4391 - add series for a cray x86/64 UL kernel drop
4392 - fix xattr patches for the vanilla 2.4.19 series
4394 - generate true UUIDs in lmc (1171)
4395 - have portals stack dumping break in UML (2466)
4396 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4397 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4398 - wake callback waiting threads on client eviction (2460)
4399 - Add --ptldebug and --subsystem to lmc (1719)
4400 - update assertion to allow safe interrupt allocation
4401 - set rq_no_resend for cancel requests (2432)
4402 - recalculate ptlrpcd timeout after resend (2494)
4403 - call vfs_rmdir when removing pending directories (2368)
4404 - fix renaming a file to itself (2429)
4405 - lmc creates a default one-stripe lov (2454)
4406 - expand procfs space to handle large clusters (2326)
4407 - increase UML stack to avoid overflow
4408 - update lconf's list of debug and subsystem masks
4409 - fix lfs find --obd (2510)
4410 - /proc tunable for disabling filter read caching (2591)
4411 - stop rpm packages from altering slapd.conf (2301)
4412 - disable nagle in the socknal under 0conf (2578)
4413 - choose mds inode size based on stripe count (2572)
4414 - fix kernel-source rpm problems (2516)
4416 - add --disable-doc to avoid pdf generation (2421)
4417 - update documentation, tests, type-os, comments
4418 - avoid format warnings on ia64
4419 - remove the TOE NAL
4420 - tiny code cleanups by removing unused fields
4422 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4425 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4426 - remove the most blatant lies from BUILDING file (2371)
4427 - change default debug level to reasonable production setting
4428 - reduce client side cache size to reduce cache flush time
4429 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4430 - make TCP zerocopy and pinger support enabled by default (2476)
4431 - sync writes completed after process exits caused crashes (2319)
4432 - maintain correct mount count on the MDS (2356)
4433 - backout 1557, because 2316 wasn't really fixed
4434 - better file I/O statistics gathering in /proc
4435 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4436 - another kernel patch to fix zero-copy TCP function export
4437 - don't take duplicate lock when processing re-sent getattr (2420)
4438 - lctl uses obd_self_export instead of creating new conn (2353)
4439 - MDS/OST recovery case which requires object creation asserted (2425)
4440 - move lfs from /usr/sbin to /usr/bin in packages
4441 - fix race between mds_client_add and mds_client_free (2417)
4442 - use kmalloc instead of slabs in portals (2430)
4443 - don't create duplicate records when a failover MDS is present (2442)
4444 - remove unnecessary mount age check (2332)
4445 - don't remove directory inodes from locks prematurely (2451)
4446 - don't break if MDS service name is the same as hostname (2103)
4447 - fix races in client write RPC generation when cache full (2482)
4449 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4452 - remove now-unused request->rq_obd (278)
4453 - if an allocation fails, print out how much memory we've used (1933)
4454 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4455 - add the "configurable stack size" patch to most series files (1256)
4456 - ability to write large log records, for 100+ OST configs (2306)
4457 - fix NULL deref when filter_prep fails (2314)
4458 - fix operator precedence error in filter_sync
4459 - dynamic allocation of socknal TX descriptors (2315)
4460 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4461 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4462 - ensure that CWARN messages in Portals always get to the syslog
4463 - __init/__exit are not for prototype decls (ldlm_init/exit)
4464 - x86-64 compile warning fixes
4465 - fix gateway LMC keyword conflict (2318)
4466 - fix MDS lock inversions in getattr/reint paths (1844)
4467 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4468 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4469 - DLM race condition prevented some lock evictions (2328)
4470 - ENOMEM detection and retry on socknal sends (2230)
4471 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4472 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4473 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4474 - better error messages when a client is rejected during recovery (1505)
4475 - avoid cancelling locks which were never granted, after failure (2330)
4476 - fix i_sem/journal inversion in mds_client_add (2333)
4477 - fix truncate/getattr lock cycle deadlock (2334)
4478 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4479 - fix two rare exit paths which could leak an l_lock() ref (2321)
4480 - fencepost error in MDS/OST orphan recovery (2226)
4481 - make log record alignment 8 bytes (1988)
4482 - lstripe now fails when requested offset > ost_count (2237)
4483 - ensure that all kernel series have a complete list.h (1607)
4484 - fix crashes in special-file operations (2316)
4485 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4487 - allow configurable automake binary, for testing new versions
4488 - small update to the lfs documentation
4490 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4492 * fix negative export reference count in fsfilt_sync (2312)
4494 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4495 * release candidate 0.9.1
4497 - orphans are moved into the PENDING directory for possible recovery
4498 - replayed opens now open by fid for orphan/rename safety (1042)
4499 - last close of an orphan inode generates a transno (683)
4500 - chdir() and mount() now pin the directory entry (1020)
4501 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4502 - discard very old requests without processing them (1502)
4503 - remove some common, well-understood CERRORs (1505)
4504 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4505 - clear "grant" flags in OST replies until OST grant code lands (1644)
4506 - fix read performance by not clobbering i_blksize on client (1598)
4507 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4508 - make LRU size a /proc tunable, clears locks when reduced (707)
4509 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4510 - server locks take references on exports now (1558)
4511 - build fixes for 2.4.20-rh trees (1663)
4512 - return an error from lov_create if all OSCs are inactive (1751)
4513 - fix import levels when a reconnect happens without a timeout (1597)
4514 - exit early from mds_open if we get a lookup error (1749)
4515 - partial page read at EOF wouldn't wait for disk before sending (1642)
4516 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4517 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4518 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4519 - write/truncate lock inversion (1639)
4520 - Don't auto-load obdclass, portals modules during cleanup (1495)
4521 - fix timestamps from jumping to "now" (1763)
4522 - extra journal assertions (1648)
4523 - add an extra multiunlink test (1771)
4524 - fix read_record/write_record API (1776)
4525 - fix leak of offset_extent, possible incorrect i_size later (1772)
4526 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4527 - Debugging for the kqswnal_get_idle_tx problems (1820)
4528 - Allow recovery to be attempted multiple times (1536)
4529 - Write out MDS last_rcvd file after it is first created (1600)
4530 - Fix tx_descriptor leak in failed transmit situations (1827)
4531 - ext3 journaling fixes for assertion failure after IO error (1871)
4532 - class_export_put() on freed export after completion AST error (1896)
4533 - Fix revalidate looping in VFS (1322)
4534 - Don't access a freed export during MDS_REINT timeout (1521)
4535 - Add open-unlink recovery support on the MDS (1673,1764)
4536 - Return an error if no MDS data was read from last_rcvd (1946)
4537 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4538 - Don't setup a disk device that doesn't match exported UUID (317)
4539 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4540 - avoid committing NULL handle in force close
4541 - local.sh is now a one-stripe LOV configuration
4542 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4543 - fix ext3 htree duplicate directory entry corruption (1516)
4544 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4545 - update to new LOV EA format (2097)
4546 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4548 2003-06-15 Phil Schwan <phil@clusterfs.com>
4551 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4552 - per-import/export recovery handling (958, 931, 959)
4553 - multiple last-rcvd slots, for serving multiple FSes (949)
4554 - connections are again shared between multiple imp/exports (963, 964)
4555 - "umount -f" would hang if any requests needed to be sent (393, 978)
4556 - avoid pinning large req buffer by copying for queued messages (989)
4557 - add "uuid" to "lctl device" command to help upcalls (991)
4558 - "open" RPCs with transnos would confuse recovery counters (1037)
4559 - do proper endian conversion of all wire messages (288, 340, 891)
4560 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4561 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4562 - fix ptlrpc_connection leak in target_handle_connect (1174)
4563 - fix import refcounting bug in OST and MDS cleanup (1134)
4564 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4565 - fix very unlikely obd_types race condition (501)
4566 - remove osc_open hack for echo_client (1187)
4567 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4568 - a failure in read_inode2 leads to deadlock (1139)
4569 - cancel ack-locks as soon as transaction is committed (1072)
4570 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4571 - make sure to commitrw after any preprw to avoid deadlock (1162)
4572 - failing to execute a file in a lustre FS would lock inode (1203)
4573 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4574 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4575 - obd_uuid2tgt would walk off the end of the list (1255)
4576 - on IA64 the osc would give portals incorrect bulk size (1258)
4577 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4578 - fix lock inversion caused by new llite matching code (1282)
4579 - limit the number of dirty pages on a client to 10MB (1286)
4580 - timed out locks were not being corrected cancelled (1289)
4581 - fix O_DIRECT above 4GB on IA-32 (1292)
4582 * major user-visible changes
4583 - fail out/fail over policy now controlled by the upcall (993)
4585 - add OBD_PING to check server availability and failure (954)
4586 - lustre messages are now sent in sending host order (288, 340, 891)
4587 - add eadatalen to MDS getattr reply (340)
4588 - OST read replies may contain second buffer, with per-page status (593)
4590 2003-03-11 Phil Schwan <phil@clusterfs.com>
4593 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4594 - failed open()s could cause deadlock; fixed (867, 869)
4595 - stop cancelling OST locks when files are closed (481)
4596 - overlapping XID spaces caused network corruption (851, 853)
4597 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4598 - setattr_raw added, to do single-RPC, server-side setattrs
4599 - lmc/lconf syntax change for OST UUIDs
4600 - fix crashy race condition between ptlrpc_free_req and osc_close
4601 - don't use request in mdc_enqueue if we hit a timeout (889)
4602 - don't set the inode i_size for regular files from the MDS (896)
4603 - handle out of order completion AST (842)
4604 - don't LBUG if a lock request times out after receiving AST (913)
4605 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4606 - if a bad lock AST arrives, send an error instead of dropping entirely
4607 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4608 - fix leak in bulk IO when only partially completed (899, 900, 926)
4609 - fix O_DIRECT for ia64 (55)
4610 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4611 - C-z after timeout could hang a process forever; fixed (977)
4613 - client-side I/O cache (678, 924, 929, 941, 970)
4615 - READPAGE and SETATTRs which don't take server-side locks get
4618 2003-02-11 Phil Schwan <phil@clusterfs.com>
4621 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4622 - Fix fsfilt_extN_readpage() to read a full page of directory
4623 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4624 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4625 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4626 - Most intent operations were removed, in favour of a new RPC mode
4627 that does a single RPC to the server and bypasses most of the VFS
4628 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4629 - Aggressively cancel local locks on DLM servers
4630 - mds_reint_unlink sends EA to the client if it's the last nlink.
4631 client uses that EA to unlink OST objects.
4632 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4633 - recursive symlinks were fixed (439)
4634 - fixed NULL deref in DEBUG_REQ
4635 - filter_update_lastobjid no longer calls sync, which annoyed extN
4636 - fixed multi-client small-writes to a single file problem (445)
4637 - fixed mtime updates during file writes (607)
4638 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4639 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4640 - fixed Linux OST size reporting problem (444, 656)
4641 - OST now updates object mtime with writes or setattr (607, 619)
4642 - client verifies file size before zeroing page past EOF (445)
4643 - OST now writes last allocated objid to disk with allocation (108)
4644 - LOV on echo now works (409)
4646 - mds_reint_unlink sends a new buffer, with the EA included. this
4647 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4648 set if a regular file was being unlinked, and it was the last link
4649 - use PtlGet from the target for bulk writes (315)
4650 - OST now updates object mtime with writes or setattr (607, 619)
4651 - LDLM now has a grant-time callback to revalidate locked items, if
4653 - Many MDS operations were reorganized to combat race conditions
4655 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4656 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4658 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4661 - Fully reactivate OST imports after reconnection (512, others)
4662 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4663 - More graceful error handling for truncating on dead OST (515)
4664 - Don't error out unless we're actually accessing dead stripes (474)
4665 - Fix garbage sizes when stripes are missing (410)
4666 - LRU counters were broken, causing constant lock purge (433, 432)
4667 - garbage on read from stripes with failed OSTs (441)
4668 - mark OSCs as active before reconnecting during recovery (438)
4669 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4670 - lfind did not preserve OST order in output (443)
4671 - symlinks cause hung clients, incorrect data (439)
4672 - stop dereferencing request after dropping refcount (457)
4673 - don't LASSERT(spin_is_locked) on non-SMP (455)
4674 - fixes for many rename() bugs
4675 - fstat didn't correctly synchronize attributes (399)
4676 - server must handle lock cancellation during blocking AST prep (487)
4677 - bulk descriptors were free()d too soon (511)
4678 - fix paths in lconf, which would load incorrect modules (451, 507)
4679 - fix confusing lconf 'host not found' error message (386)
4680 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4681 - fix race condition in mdc_blocking_ast() for inode access (526)
4682 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4683 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4684 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4685 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4686 - fix lstripe in several ways (kernel side) (527)
4687 - fix request leak in ldlm_cli_enqueue (262)
4688 - incorrect OSC was marked inactive after OST failure
4689 - call mds_fs_cleanup before unmounting filesystem (524)
4690 - fix races between taking ns_lock and ldlm_lock_change_resource
4691 - fix races updating LOV export open file list
4692 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4693 - fix recovery NULL deref in ldlm_cli_cancel_unused
4694 - fix some DLM races by using new hash table for lock handles (419)
4695 - permit the client to specify desired inodes, at replay
4696 - duplicate requests when we queue them for replay reintegration
4697 - fix last_rcvd offset calculation
4698 - sync after each recovered transaction, so we always make progress
4699 - never, not always, ERESTART requests without transnos
4700 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4702 - skip replay if the MDS says that the client is already connected
4703 - don't check for a recovery-enabled export to match lctl's UUID
4704 - don't INC_USE_COUNT for phantom exports
4705 - don't crash when cleaning up phantom exports (567)
4706 - don't double-finish or set replay data for errored mdc_open requests
4707 - abort requests when they time out, so we don't get old replies
4708 - send/receive replies for AST messages again
4709 - if the client says that it doesn't have the lock, cancel it on the
4711 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4712 mark it as destroyed, it will all work out when decref is called
4713 - fix module use counts (22, 581)
4715 - ASTs now expect a reply (server cancels lock on error reply)
4717 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4720 - fix many simultaneous client startup (392)
4721 - fix dentry->d_it clobbering
4722 - credentials weren't being shipped for readdir/getattr operations
4723 - remove invalid assertions triggered during some concurrent MD
4725 - proper Lustre versions added (336, 389)
4726 - fix memory leak for create error case (398)
4727 - fix LOV locking bug that would get cli/srv out of sync
4728 - fix echo client over LOV (409)
4729 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4730 - fix double-O_EXCL intent crash (424)
4731 - avoid sending multiple lock CANCELs (352)
4733 - MDS can do multi-client recovery (modulo bugs in new code)
4735 - many updates, edits, cleanups
4737 2002-11-18 Phil Schwan <phil@clusterfs.com>
4740 - fix null d_it dereference (346)
4741 - fix full OST/dbench hang (333)
4742 - fix permission problem with file removal (286)
4743 - fix removal of OSCs from LOV when they fail
4744 - fix NULL deref during bulk timeout (214)
4745 - fix problems related to multiple filesystems on one MDS (241)
4746 - fixed serious subtle metadata locking bugs
4747 - free locks on clients when inodes are removed due to memory
4749 - fix inode pointer in lock data (285)
4750 - partial support for multiple MDS on a single host (241)
4751 - data locks weren't cancelled at clear_inode time (290, 311)
4752 - intent locks could lead to unbounded lock growth (205)
4753 - added a maximum lock count, an LRU list, and a flusher
4754 - fix multiple rename (365)
4755 - properly abstracted the echo client
4756 - OSC locked 1 byte too many; fixed
4757 - rewrote brw callback code:
4758 - fixed recovery bugs related to LOVs (306)
4759 - fixed too-many-pages-in-one-write crash (191)
4760 - fixed (again) crash in sync_io_timeout (214)
4761 - probably fixed callback-related race (385)
4763 - Add capability to MDS protocol
4764 - LDLM cancellations and callbacks on different portals
4766 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4769 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4770 - defer open object destruction to close time (601981, 138)
4771 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4772 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4773 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4774 - don't call obd_setattr in ll_file_release for destroyed objects
4775 * protocol change to lustre_msg: move |version| and add |flags|
4776 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4777 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4778 * added replay of create, unlink, link and rename operations during
4779 MDS failover; recovery should be much more robust now
4780 * remove failed OSCs from LOVs (only lov_create uses this so far)
4781 * the lustre-HOWTO was brought (more) up to date (582544)
4783 2002-10-23 Phil Schwan <phil@clusterfs.com>
4786 - in-use dentries weren't being reused properly (617851)
4787 - prevent multiple LDLM setup (599178)
4788 - fix LOV size calculations for truncate (617853)
4789 - fix client handling of MDS intent errors (POSIX)
4790 - fix permission bug in lovstripe.c test (624321)
4791 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4792 - truncate past end of file could corrupt data
4793 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4794 - a race in recovery could return ETIMEDOUT to apps (623947)
4795 - building outside the source directory was fixed
4796 * the lustre-HOWTO was brought (more) up to date (582544)
4797 * major progress was made on recovery functionality
4799 2002-10-10 Phil Schwan <phil@clusterfs.com>
4802 - recovery deadlock fix
4803 - rm -rf causes LBUG fix (617817)
4804 - file open by multiple tasks fix (618962)
4805 - directory permissions bugs (602707 and 620007)
4806 - journal_stop fixed with locking (611313)
4807 - O_APPEND failures resolved (618273, perhaps 614459)
4808 - lconf PATH fix (619770)
4809 - IA64 build fix (621450)
4810 - RPC buffer sizes scale with amount of memory
4812 2002-10-01 Phil Schwan <phil@clusterfs.com>
4815 - locks would be cancelled without throwing away data pages,
4816 resulting in inconsistent data (605627)
4817 - inode attributes were not always being refreshed (605627, 612449)
4818 - lconf now continues to cleanup after lctl reports an error
4819 - MDS now enforces user permissions (602707)
4820 - lprocfs cleanup fixed, but not yet enabled (614157)
4821 - fixed infinite server hang, should a client not respond to an AST
4822 - avoid going into recovery if user calls readlink() with a buffer
4823 that's too small (613941)
4824 - AST RPCs no longer require replies (614867) -- this may be changed
4825 - don't crash server if client sends an IOV that's too big (611336)
4826 - fixed lock conversion deadlock (611892)
4827 - fixed the following of symlinks (614622)
4828 * recovery: the server can remove locks from a client that dies, other
4829 clients can make progress
4830 * more extN patch fixes
4831 * compile-time configurable ptlrpc buffer allocations
4833 - collaborative read cache document
4834 - Lustre Lite Performance CDR document-in-progress
4836 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4839 - fix typo in patch-2.4.18
4841 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4844 - clear ptlrpc request each time in handle_incoming_request()
4845 - unlink of files now destroys the object on the OST
4847 2002-09-19 Peter Braam <braam@clusterfs.com>
4849 * add hard link support
4850 * change obdfile creation method
4851 * kernel patch changed
4853 2002-09-19 Peter Braam <braam@clusterfs.com>
4856 - stack overflow bug in extN fixed
4858 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4860 * documentation updates
4861 - add man pages for config tools
4862 - update tests/README to describe testing with new config tools
4863 - finish metadata API descriptions
4864 * bug fixes and cleanups
4865 - statfs workaround for 16TB limit
4866 - LOV stripe allocation improved, can stripe on subset of OSTs
4867 - LOV file size/IO offset was wrong for files > 4GB in size
4868 - object EA data was being dropped, caused files to be unreadable
4869 - memory overflow with non-LOV OST caused memory corruption
4870 - fixed regression tests to work with new config tools, obdfilter
4871 - fixed bug when directory size became larger than 1 block
4872 - fixed bug (for single client case) when PWD was deleted
4873 - invalidate local directory pages when doing intent-based ops
4874 - avoid LDLM oops when lock callback contained bad data
4876 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4878 * documentation updates
4879 * bug fixes and cleanups
4880 - configuration tools
4883 - 64-bit compile warnings
4884 - 64-bit internal statfs data
4886 * test_brw on persistent OST devices
4888 * lprocfs (disabled)
4890 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4892 * documentation updates
4893 * bug fixes and cleanups
4894 * configuration tools
4896 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4899 * many small fixes to 0_5_4
4900 * io/network handling
4901 * thinkos in MDS operations
4903 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4906 * crucial basic fixes to 0.5.3
4907 * IOR, Iozone work over Elan
4910 2002-08-07 Phil Schwan <phil@clusterfs.com>
4911 * version 0_5_3, our first alpha
4912 * we use the new Portals iovs
4913 * documentation updates
4914 * bug fixes and cleanups
4915 * small changes in the DLM wire protocol
4917 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4918 * version 0_5_1 with some initial stability,
4919 * locking on MD and file I/O.
4920 * documentation updates
4921 * several bug fixes since 0.5.0
4922 * small changes in wire protocol
4924 2002-07-18 Phil Schwan <phil@clusterfs.com>
4926 * delivered as Lustre Light Alpha
4927 * fixed a crash after handling invalid MDS requests
4928 * fixed directory pages for architectures with non-4k pages sizes
4930 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4931 * release version v0_4_4
4932 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4934 * Fixes a number of interruption problems with OST operations.
4935 * Update documentation for portals header changes
4936 * Move all wire protocol structs/defines to lustre_idl.h
4937 * Fixes symlink length bug.
4938 * Add tcpdump to repository.
4940 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4941 * release version v0_4_3
4942 * Fixes statfs for inodes on extN.
4943 * Fixes bug in runtests which would delete /etc/hosts.
4944 * Use 64-bit object IDs wherever possible (not into VFS though)
4945 Remove ost_get_info, which is unused by lustre, and out of date.
4947 2002-07-03 Peter Braam <braam@clusterfs.com>
4948 * release version v0_4_2 Fixes a lookup error (type not passed)
4949 * move forward to head of Portals
4950 * move forward to latest Lustre kernel
4952 2002-06-25 Peter Braam <braam@clusterfs.com>
4953 * release version v0_4_1. Hopefully stable on single node use.