1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.23 (SLES 10),
5 2.6.18-92.1.6.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).
15 Severity : enhancement
17 Description: Upcall on Lustre log has been dumped
18 Details : Allow for a user mode script to be called once a Lustre log has
19 been dumped. It passes the filename of the dumped log to the
20 script, the location of the script can be specified via
21 /proc/sys/lnet/debug_log_upcall.
26 Description: avoid messages about idr_remove called for id which is not allocated.
27 Details : Move assigment s_dev for clustered nfs to end of initialization, for
28 avoid problem with error handling.
33 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
34 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
35 CONN_USED_HASH and this prodice warning when put connection again
38 Severity : enhancement
40 Description: Export bytes_read/bytes_write count on OSC/OST.
44 Description: Early reply size mismatch, MGC loses connection
45 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
46 the connect flags are properly negotiated.
50 Description: Properly propagate oinfo flags from lov to osc for statfs
51 Details : restore missing copy oi_flags to lov requests.
53 Severity : enhancement
55 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
56 : l_getgroups(8), lst(8), routerstat(8)
57 Details : included man pages for llobdstat(8), llstat(8),
58 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
60 Severity : enhancement
62 Description: configure's --enable-quota should check the
63 : kernel .config for CONFIG_QUOTA
64 Details : configure is terminated if --enable-quota is passed but
65 : no quota support is in kernel
69 Description: Remove portals compatibility
70 Details : Remove portals compatibility, not interoperable with releases
75 Description: Resolve device initialization race
76 Details : Prevent proc handler from accessing devices added to the
77 obd_devs array but yet be intialized.
79 Severity : enhancement
81 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
83 Severity : enhancement
85 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
90 Description: Using wrong pointer in osc_brw_prep_request
91 Details : Access to array[-1] can produce panic if kernel compiled with
92 CONFIG_PAGE_ALLOC enabled
96 Description: Client runs out of low memory
97 Details : Consider only lowmem when counting initial number of llap pages
101 Description: *optional* service tags registration
102 Details : if the "service tags" package is installed on a Lustre node
103 When the filesystem is mounted, a local-node service tag will
104 be created. See http://inventory.sun.com/ for more information
105 about the Service Tags asset management system.
109 Description: Kernel BUG tries to release flock
110 Details : Lustre does not destroy flock lock before last reference goes
111 away. So always drop flock locks when client is evicted and
112 perform unlock regardless of successfulness of speaking to MDS.
116 Description: add recount protection for osc callbacks, so avoid panic on shutdown
120 Description: sanity test 65a fails if stripecount of -1 is set
121 Details : handle -1 striping on filesystem in ll_dirstripe_verify
126 Description: ASSERTION(CheckWriteback(page,cmd)) failed
127 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
130 Severity : enhancement
132 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
136 Description: do not process already freed flock
137 Details : flock can possibly be freed by another thread before it reaches
138 to ldlm_flock_completion_ast.
142 Description: LBUG during stress test
143 Details : Need properly lock accesses the flock deadlock detection list.
147 Description: oops in page fault handler
148 Details : kernel page fault handler can return two special 'pages' in error case, don't
149 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
153 Description: timeout with invalidate import.
154 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
155 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
156 ptlrpc_invalidate_import and as result LASSERT.
158 Severity : enhancement
160 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
164 Description: enable MGS and MDT services start separately
165 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
166 a co-located MGS without starting the MGS, which is a complement
167 to 'nosvc' mount option.
171 Frequency : after recovery
172 Description: precreate to many object's after del orphan.
173 Details : del orphan st in oscc last_id == next_id and this triger growing
174 count of precreated objects. Set flag LOW to skip increase count
175 of precreated objects.
179 Frequency : rare, on clear nid stats
180 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
181 Details : when clean nid stats sometimes try destroy live entry,
182 and this produce panic in free.
186 Description: Stack overflow during MDS log replay
187 ease stack pressure by using a thread dealing llog_process.
191 Description: wait until IO finished before start new when do lock cancel.
192 Details : VM protocol want old IO finished before start new, in this case
193 need wait until PG_writeback is cleared until check dirty flag and
194 call writepages in lock cancel callback.
196 Severity : enhancement
198 Description: using special macro for print time and cleanup in includes.
202 Description: mds_mfd_close() ASSERTION(rc == 0)
203 Details : In mds_mfd_close(), we need protect inode's writecount change
204 within its orphan write semaphore to prevent possible races.
208 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
212 Frequency : rare, on shutdown ost
213 Description: don't hit live lock with umount ost.
214 Details : shrink_dcache_parent can be in long loop with destroy dentries,
215 use shrink_dcache_sb instead.
219 Description: don't panic with use echo client
220 Details : echo client pass NULL as client nid pointer and this produce null
225 Description: fix build on ppc32
226 Details : compile code with -m64 flag produce wrong object file for ppc32.
230 Description: add message levels for liblustreapi
234 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
235 Details : Move the 'good_osts' check before the 'total_bavail' check. This
236 will result in an -EAGAIN and in the exit call path we call
237 alloc_rr() which will with increasing aggressiveness attempt to
238 aquire precreated objects on the minimum number of required OSCs.
242 Description: Use old size assignment to avoid deadlock
243 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
244 the scheduling while holding a spinlock. We do not need locking
245 for size in ll_update_inode() because size is only updated from
246 the MDS for directories or files without objects, so there is no
247 other place to do the update, and concurrent access to such inodes
248 are protected by the inode lock.
252 Description: resolve "_IOWR redefined" build error on SLES10
256 Description: dump the memory debugging after all modules are unloaded to
257 suppress false negative in conf_sanity test 39
259 Severity : enhancement
261 Description: build kernel-ib packages for OFED 1.3 in our release cycle
266 Description: fix SLES kernel versioning
267 Details : the kernel version for our SLES 10 kernel did not include a "-"
268 before the "smp" at the end. while this was not a problem in
269 general, it did mean that software trying to use the kernel
270 version to try to detect a vendor specific kernel would fail.
271 this was most evident by the OFED build scripts.
275 Description: Don't update lov_desc members until making sure they are valid
276 Details : When updating lov_desc members via proc fs, need fix their
277 validities before doing the real update.
281 Description: don't put request into delay list while invalidate in flight.
282 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
283 import in flight. this produce timeout for invalidate and sometimes
284 can cause stale data.
288 Frequency : on ppc only
289 Description: not convert ost objects for directory because it's not exist.
290 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
294 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
296 Severity : enhancement
298 Description: Update to support 2.6.22.14 vanilla kernel.
302 Frequency : rare, on recovery
303 Description: read procfs can produce deadlock in some situation
304 Details : Holding lprocfs lock which send rpc can produce block for destroy
305 obd objects and this also block reconnect with -EALREADY. This isn't
306 fix all lprocfs bugs - but make it rare.
308 Severity : enhancement
310 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
313 Frequency : frequent on X2 node
315 Description: mdc_set_open_replay_data LBUG
316 Details : Set replay data for requests that are eligible for replay.
320 Description: lustre_mgs: operation 101 on unconnected MGS
321 Details : When MGC is disconnected from MGS long enough, MGS will evict the
322 MGC, and late on MGC cannot successfully connect to MGS and a lot
323 of the error messages complaining that MGS is not connected.
327 Frequency : on network error
328 Description: panic with double free request if network error
329 Details : mdc_finish_enqueue is finish request if any network error ocuring,
330 but it's true only for synchronus enqueue, for async enqueue
331 (via ptlrpcd) this incorrect and ptlrpcd want finish request
334 Severity : enhancement
336 Description: client-side metadata stat-ahead during readdir(directory readahead)
337 Details : perform client-side metadata stat-ahead when the client detects
338 readdir and sequential stat of dir entries therein
341 Frequency : on start mds
343 Description: Implement get_info(last_id) in obdfilter.
346 Frequency : occasional
348 Description: Correctly check stale fid, not start epoch if ost not support SOM
349 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
350 overwrited with new generated fid, but mds can anwer with one of these
351 two fids and both is not stale. setattr incorectly start epoch and
352 assume will be called done_writeting, but without SOM done_writing
356 Frequency : rare, depends on device drivers and load
358 Description: MDS or OSS nodes crash due to stack overflow
359 Details : Code changes in 1.8.0 increased the stack usage of some functions.
360 In some cases, in conjunction with device drivers that use a lot
361 of stack the MDS (or possibly OSS) service threads could overflow
362 the stack. One change which was identified to consume additional
363 stack has been reworked to avoid the extra stack usage.
366 Frequency : occasional
368 Description: Do not fail import if osc_interpret_create gets -EAGAIN
369 Details : If osc_interpret_create got -EAGAIN it immediately exits and
370 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
371 oscc_has_objects and see OSC has no objests and call
372 oscc_internal_create to resend create request.
374 Severity : enhancement
376 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
378 Severity : enhancement
380 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
383 Frequency : very rare
385 Description: Fix replay if there is an un-replied request and open
386 Details : In some cases, older replay request will revert the
387 mcd->mcd_last_xid on MDS which is used to record the client's
390 Severity : enhancement
392 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
394 Severity : enhancement
396 Description: Add rhel5 support to HEAD.
398 Serverity : enhancement
400 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
405 Description: Don't allow skipping OSTs if index has been specified.
406 Details : Don't allow skipping OSTs if index has been specified, make locking
407 in internal create lots better.
411 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
412 Details : access to bitfield in structure is always rounded to long
413 and this produce problem with not atomic change any bit.
417 Description: Lustre make rpms failed.
418 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
419 builds Lustre rpms from tarball.
422 Frequency : rare on shutdown ost
424 Description: If llog cancel was not send before clean_exports phase, this can
425 produce deadlock in llog code.
426 Details : If llog thread has last reference to obd and call class_import_put
427 this produce deadlock because llog_cleanup_commit_master wait when
428 last llog_commit_thread exited, but this never success because was
429 called from llog_commit_thread.
433 Description: allow userland application know is lost one of stripes.
434 Details : fill lvb_blocks with error code on ost and return it to
435 application if error flag found.
439 Description: NULL lov_tgts causing MDS oops
440 Details : more safe checks for NULL lov_tgts for avoid oops.
442 Severity : enhancement
444 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
448 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
449 Details : If an OST is down the MDS will hang indefinitely in
450 obd_statfs_rqset() waiting for the statfs data. While for
451 MDS QOS usage of statfs, it should not stuck in waiting.
453 Severity : enhancement
455 Description: remote_acl support
456 Details : Support ACL-based permission check for remote user.
457 Support setfacl/getfacl for remote user with the utils
458 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
459 the system "{s,g}etfacl" utils.
461 Severity : enhancement
463 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
465 Severity : enhancement
467 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
471 Description: make mgs_setparam() handle fsname containing dash
472 Details : fsname containing a dash does not work with lctl conf_param
474 Severity : enhancement
476 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
478 Severity : enhancement
480 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
484 Frequency : rare, at shutdown
485 Description: access already free / zero obd_namespace.
486 Details : if client_disconnect_export was called without force flag set,
487 and exist connect request in flight, this can produce access to
488 NULL pointer (or already free pointer) when connect_interpret
489 store ocd flags in obd_namespace.
493 Frequency : only at startup
494 Description: not alloc memory with spinlock held.
495 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
496 if any spinlock held.
498 Severity : enhancement
500 Description: make lustre randomly fail allocating memory
501 Details : Make lustre randomly failed allocating memory for testing purpose.
503 Severity : enhancement
505 Description: lost problems with lov objid file
506 Details : Fixes some scability and access to not inited memory problems
507 in work with lov objdid file.
512 Description: lfs find does not continue on file error
513 Details : Continue other files processing when a file/dir is absent.
517 Description: Inconsistent usage of lustre_pack_reply()
518 Details : Standardize the usage of lustre_pack_reply() such that it
519 always generate a CERROR on failure.
522 Frequency : occasional
524 Description: MDS hang or stay in waiting lock
525 Details : If client receive lock with CBPENDING flag ldlm need send lock
526 cancel as separate rpc, to avoid situation when cancel request
527 can't processed due all i/o threads stay in wait lock.
531 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
535 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
539 Description: Update to SLES9 kernel-2.6.5-7.287.3.
543 Description: Update to RHEL4 latest kernel.
545 Severity : enhancement
547 Description: Build SLES10 patchless client fails
548 Details : The configure was broken by run ./configure with
549 --with-linux-obj=.... argument for patchless client. When the
550 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
551 can't search header adequately. Use absolute path such as
552 -I($LINUX)/include instead.
556 Description: interrupt oig_wait produce painc on resend.
557 Details : brw_redo_request can be used for resend requests from ptlrpcd and
558 private set, and this produce situation when rq_ptlrpcd_data not
559 copyed to new allocated request and triggered LBUG on assert
560 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
561 ptlrpcd set if request is changed and can be safety changed to use
566 Description: LASSERT_{REQ,REP}SWAB macros are buggy
567 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
568 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
569 these macros and replace them with their difinitions instead.
573 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
574 Details : Update which_patch & target file for SLES10 latest kernel.
577 Description: Remove client patches from SLES 10 kernel.
578 Details : This causes SLES 10 clients to behave as patchless clients
579 even on a Lustre-patched (server) kernel.
581 Severity : enhancement
583 Description: self-adjustable client's lru lists
584 Details : use adaptive algorithm for managing client cached locks lru
585 lists according to current server load, other client's work
586 pattern, memory activities, etc. Both, server and client
587 side namespaces provide number of proc tunables for controlling
590 Severity : enhancement
592 Description: light-weight GSS support
593 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
594 and reduce performance overhead.
596 Severity : enhancement
598 Description: Linux keyring support
599 Details : Support using service of Linux keyring for Lustre GSS internal
600 context refresh/cache mechanism.
604 Description: Fix errors in lfs documentation
605 Details : Fixes man pages
609 Description: don't use GFP_* in generic Lustre code.
610 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
614 Description: obdclass is limited by single OBD_ALLOC(idarray)
615 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
619 Description: warnings with build patchless client with vanila 2.6.19 and up
620 Details : change old ctl_table style and replace ctl_table/ctl_table_header
621 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
625 Frequency : Only for SLES
626 Description: sanity_quota fail test_1
627 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
628 file, which makes TSTID[2] inunique.
632 Frequency : Only for SLES9
633 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
634 Details : Update target/ChangeLog/which_patch .
638 Frequency : Only for test_10 in sanity-quota.sh
639 Description: a bug in quota test code
640 Details : checking a test flag in a wrong place
644 Frequency : Only for test_13 in sanity-quota.sh
645 Description: a bug in quota test script
646 Details : shouldn't check overfull allocated quota
650 Description: jbd statistics
651 Details : Port older jbd statistics patch for sles10
653 Severity : enhancement
655 Description: Add printing inode info into message about error in writepage.
660 Description: reply_lock_interpret crash due to race with it and lock cancel.
661 Details : Do not replay locks that are being cancelled. Do not reference
662 locks by their address during replay, just by their handle.
666 Frequency : When flocks are used.
667 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
668 Details : Flock locks might destroy just granted lock if it could be merged
669 with another existing flock, this is done in completion handler,
670 so teach ldlm_cli_enquque_fini that this is a valid case for
676 Description: Oops in read and write path when failing to allocate lock.
677 Details : Check if lock allocation failed and return error back.
681 Description: lstripe command fails for valid OST index
682 Details : The stripe offset is compared to lov->desc.ld_tgt_count
683 instead of lov->desc.ld_active_tgt_count.
687 Description: sanity.sh failed test 103
688 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
693 Description: df doesn't work properly if diskfs blocksize != 4K
694 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
698 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
699 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
701 Severity : enhancement
703 Description: Adaptive timeouts
704 Details : RPC timeouts adapt to changing server load and network
705 conditions to reduce resend attempts and improve recovery time.
709 Description: llapi_file_create() does not allow some changes
710 Details : add llapi_file_open() that allows specifying the mode and
711 open flags, and also returns an open file handle.
715 Description: merge and cleanup kernel patches.
716 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
720 Description: Add journal checksum support.(Kernel part)
721 Details : The journal checksum feature adds two new flags i.e
722 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
723 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
724 indicates that the commit block contains the checksum for
725 the blocks described by the descriptor blocks. Now commit
726 record can be sent to disk without waiting for descriptor
727 blocks to be written to disk. This behavior is controlled
728 using JBD2_FEATURE_ASYNC_COMMIT flag.
732 Description: OSS needs mutliple precreate threads
733 Details : Add ability to start more than one create thread per OSS.
737 Description: Sanity.sh test_65k failed: setstripe should have succeeded
738 Details : execute lfs setstripe on client
742 Description: mds_obd_create error creating tmp object
743 Details : When the user sets quota on root, llog will be affected and can't
744 create files and write files.
747 Frequency : Always on ia64 patchless client, and possibly others.
749 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
750 Details : This allows the patchless client to be loaded on architectures
755 Description: RedHat Update kernel for RHEL5
756 Details : Add the kernel config file for RHEL5.
760 Description: RedHat Update kernel for RHEL5
761 Details : Modify the kernel config file more closer RHEL5.
765 Description: Build failure against Centos5 (RHEL5)
766 Details : Define PAGE_SIZE when it isn't present.
770 Description: test_7 on sanity-quota.sh can't be used on separate machines
771 Details : add facet to handle it
775 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
776 Details : with NFS, the anon dentry's parent was set to itself in
777 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
778 corresponding dentry other than use rec->ur_name.
780 Severity : enhancement
782 Description: lfs setstripe enhancement
783 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
787 Description: enable data checksumming by default
788 Details : enable checksum by default, allow --disable-checksum
789 configure option and "-o nochecksum" mount option. Checksums
790 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
794 Description: lustre support for RHEL5
795 Details : Add support for RHEL5.
799 Description: Client eviction due to failover config
800 Details : after a connection loss, the lustre client should attempt to
801 reconnect to the last active server first before trying the
802 other potential connections.
806 Description: when mds and osts use different quota unit(32bit and 64bit),
807 quota will be released repeatly.
808 Details : void sending multiple quota reqs to mds, which will keep the
809 status between the reqs.
813 Description: osts not allocated evenly to files
814 Details : change the condition to increase offset_idx
817 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
819 Description: Data corruption for OSTs that are formatted larger than 2TB
821 Details : When generating the bio request for lustre file writes the
822 sector number would overflow a temporary variable before being
823 used for the IO. The data reads correctly from Lustre (which
824 will overflow in a similar manner) but other file data or
825 filesystem metadata may be corrupted in some cases.
829 Description: Tune the kernel for good SCSI performance.
830 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
831 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
836 Description: rewrite ext2-derived code in obdclass/uuid.c
837 Details : rewrite inherited code (uuid parsing code from ext2 utils)
838 from scratch preserving functionality.
842 Description: "lfs find -obd UUID" prints directories
843 Details : "lfs find -obd UUID" will return all directory names instead
844 of just file names. It is incorrect because the directories
845 do not reside on the OSTs.
849 Description: use i_size_read and i_size_write in 2.6 port
850 Details : replace inode->i_size access with i_size_read/write()
853 Frequency : when removing large files
855 Description: scheduling issue during removal of large Lustre files
856 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
857 It causes scheduling issues when removing large files (17TB in the
861 Frequency : only with liblustre clients on XT3
863 Description: evictions taking too long
864 Details : allow llrd to evict clients directly on OSTs
867 Frequency : only on ppc
869 Description: /proc/fs/lustre/devices broken on ppc
870 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
871 We should make sure the type of 'index' is loff_t and then cast
872 explicitly as needed below. Do not assign an explicitly cast
876 Frequency : only for rhel5
878 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
879 Details : Modify the target file & which_kernel.
881 Severity : enhancement
883 Description: omit set fsid for export NFS
884 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
885 as FS_REQUIRES_DEV for avoid problems with generate fsid.
889 Frequency : Always, for kernels after 2.6.16
890 Description: Fix warning idr_remove called for id=.. which is not allocated.
891 Details : Last kernels save old s_dev before kill super and not allow
892 to restore from callback - restore it before call kill_anon_super.
896 Description: buffer overruns could theoretically occur
897 Details : llapi_semantic_traverse() modifies the "path" argument by
898 appending values to the end of the origin string, and a
899 overrun may occur. Adding buffer overrun check in liblustreapi.
903 Description: Fix error on 'ls .' at the top of the Lustre mount.
904 Details : Don't revalidate dentry if it is a root dentry.
908 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
909 Details : Modify vm-tunables-rhel4.patch.
913 Description: Kernel config for 2.6.18-vanilla.
914 Details : Modify targets/2.6-vanilla.target.in.
915 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
916 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
917 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
918 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
922 Description: improve handling recoverable errors
923 Details : if request processig with error which can be recoverable on server
924 request should be resend, otherwise page released from cache and
929 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
930 Details : Modify target file & which_patch.
931 A flaw was found in the IA32 system call emulation provided
932 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
933 value could be stored in the %RAX register, which could trigger an
934 out-of-bounds system call table access. An untrusted local user
935 could exploit this flaw to run code in the kernel
936 (ie a root privilege escalation). (CVE-2007-4573).
940 Description: O_DIRECT bypasses client statistics.
941 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
942 read_ahead_stats were updated. copied the stats section from
943 osc_send_oap_rpc() into async_internal().
947 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
951 Description: change order of libsysio includes
952 Details : '#include sysio.h' should always come before '#include xtio.h'
956 Description: handle "serious error: objid * already exists" more gracefully
957 Details : If LAST_ID value on disk is smaller than the objects existing in
958 the O/0/d* directories, it indicates disk corruption and causes an
959 LBUG(). If the object is 0-length, then we should use the existing
960 object. This will help to avoid a full fsck in most cases.
962 Severity : enhancement
964 Description: adapt the lustre_config script to support the upgrade case
965 Details : Add "-u" option for lustre_config script to support upgrading 1.4
966 server targets to 1.6 in parallel.
970 Description: To avoid grant space > avaible space when the disk is almost
971 full. Without this patch you might see the error "grant XXXX >
972 available" or some LBUG about grant, when the disk is almost
974 Details : In filter_check_grant, for non_grant cache write, we should
975 check the left space by if (*left > ungranted + bytes), instead
976 of (*left > ungranted), because only we are sure the left space
977 is enough for another "bytes", then the ungrant space should be
978 increase. In client, we should update cl_avail_grant only there
979 is OBD_MD_FLGRANT in the reply.
983 Description: Update RHEL 4 kernel to fix local root privilege escalation.
984 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
985 described in CVE-2007-4573. This problem could allow untrusted
986 local users to gain root access.
989 Frequency : when using O_DIRECT and quotas
991 Description: Incorrect file ownership on O_DIRECT output files
992 Details : block usage reported by 'lfs quota' does not take into account
993 files that have been written with O_DIRECT.
998 Description: touch file failed when fs is not full
999 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1000 otherwise we can get ENOSP while fs is not full.
1004 Description: parallel lock callbacks
1005 Details : Instead of sending blocking and completion callbacks as separated
1006 requests, adding them to a set and sending in parallel.
1009 Frequency : only for Cray XT3
1010 Bugzilla : 12829/13455
1011 Description: Changing primary group doesn't change the group lustre assigns to
1013 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1014 overridden with the primary group provided by the group upcall,
1015 whereas the supplied fsgid can be trusted if it is in the list of
1016 supplementary groups returned by the group upcall.
1018 Severity : enhancement
1020 Description: Allow masking D_WARNING, D_ERROR messages from console
1021 Details : Console messages can now be disabled via lnet.printk.
1025 Description: User code with malformed file open parameter crashes client node
1026 Details : Before packing join_file req, all the related reference should be
1027 checked carefully in case some malformed flags cause fake
1028 join_file req on client.
1032 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1033 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1034 mode is equal to the granted mode. However, it should also check
1035 that LDLM_FL_CP_REQD is not set.
1039 Description: Heavy nfs access might result in deadlocks
1040 Details : After ELC code landed, it is now improper to enqueue any mds
1041 locks under och_sem, because enqueue might want to decide to
1042 cancel open locks for same inode we are holding och_sem for.
1046 Description: Client eviction while running blogbench
1047 Details : A lot of unlink operations with concurrent I/O can lead to a
1048 deadlock causing evictions. To address the problem, the number of
1049 oustanding OST_DESTROY requests is now throttled to
1050 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1051 callbacks are processed in priority.
1055 Description: enable ACLs on MDS by default
1056 Details : ACLs must be enabled on MDS by default.
1059 Frequency : PPC/PPC64 only
1061 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1062 Details : fix duplicated definitions between asm-ppc64/types.h and
1063 lustre_types.h on PPC.
1066 Frequency : PPC/PPC64 only
1068 Description: asm-ppc/segment.h does not exist
1069 Details : fix compile issue on PPC.
1073 Description: better handle error messages in extents code
1076 Frequency : RHEL4 only
1078 Description: mkfs is very slow on IA64/RHEL4
1079 Details : A performance regression has been discovered in the MPT Fusion
1080 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1081 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1082 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1084 Severity : enhancement
1086 Description: SNMP support enhancement
1087 Details : Adding total number of sampled request for an MDS node in snmp
1090 Severity : enhancement
1092 Description: Optimize ldlm waiting list processing for PR extent locks
1093 Details : When processing waiting list for read extent lock and meeting read
1094 lock that is same or wider to it that is not contended, skip
1095 processing rest of the list and immediatelly return current
1096 status of conflictness, since we are guaranteed there are no
1097 conflicting locks in the rest of the list.
1101 Description: Time out and refuse to reconnect
1102 Details : When the failover node is the primary node, it is possible
1103 to have two identical connections in imp_conn_list. We must
1104 compare not conn's pointers but NIDs, otherwise we can defeat
1105 connection throttling.
1109 Description: port llog fixes from b1_6 into HEAD
1110 Details : Port llog reference couting and some llog cleanups from b1_6
1111 (bug 10800) into HEAD, for protect from panic and access to already
1112 free llog structures.
1116 Description: Detect stride IO mode in read-ahead
1117 Details : When a client does stride read, read-ahead should detect that and
1118 read-ahead pages according to the detected stride pattern.
1122 Description: data checksumming impacts single node performance
1123 Details : add support for several checksum algorithm. Currently, only CRC32
1124 and Adler-32 are supported. The checksum type can be changed on
1125 the fly via /proc/fs/lustre/osc/*/checksum_type.
1129 Description: use adler32 for page checksums
1130 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1135 Description: build for x2 fails
1136 Details : fix compile issue on Cray systems.
1140 Description: Properly match for duplicate locks
1141 Details : Due to different lock order from skiplists code, we need to
1142 traverse entire list for now
1145 Frequency : only on PPC/SLES10
1147 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1148 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1149 be defined. Add a hack in mkfs_lustre.c to work around this compile
1154 Description: LASSERT on MDS when client holding flock lock dies
1155 Details : ldlm pool logic depends on number of granted locks equal to
1156 number of released locks which is not true for flock locks, so
1157 just exclude such locks from consideration.
1161 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1162 Details : Use fsfilt_sync() for both the whole filesystem sync and
1163 individual file sync to eliminate dangerous inode locking
1164 with I_LOCK that can lead to a deadlock.
1168 Description: performance in 1.6.3
1169 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1170 sure that 1MB requests are not fragmented by the block layer.
1172 Severity : enhancement
1174 Description: organize the server-side client stats on per-nid basis
1175 Details : Change the structure of stats under obdfilter and mds to
1183 The "uuid"s file would list the uuids of _active_ exports.
1184 And the clear entry is to clear all stats and stale nids.
1186 Severity : enhancement
1188 Description: eliminate client locks in face of contention
1189 Details : file contention detection and lockless i/o implementation
1190 for contended files.
1194 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1197 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1199 Description: MDS may lose file striping (and hence file data) in some cases
1200 Details : If there are additional extended attributes stored on the MDS,
1201 in particular ACLs, SELinux, or user attributes (if user_xattr
1202 is specified for the client mount options) then there is a risk
1203 of attribute loss. Additionally, the Lustre file striping
1204 needs to be larger than default (e.g. striped over all OSTs),
1205 and an additional attribute must be stored initially in the
1206 inode and then increase in size enough to be moved to the
1207 external attribute block (e.g. ACL growing in size) for file
1212 Description: skiplist implementation simplification
1213 Details : skiplists are used to group compatible locks on granted list
1214 that was implemented as tracking first and last lock of each lock group
1215 the patch changes that to using doubly linked lists
1219 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1220 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1221 because LASSERT on both the data supplied by a client, and the data
1222 on disk is dangerous and incorrect.
1224 Severity : enhancement
1226 Description: Slow truncate/writes to huge files at high offsets.
1227 Details : Directly associate cached pages to lock that protect those pages,
1228 this allows us to quickly find what pages to write and remove
1229 once lock callback is received.
1233 Description: more ldlm soft lockups
1234 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1235 starve other threads from the resource lock for a long time in
1236 case of long waiting queue, so change the debug level from
1237 D_OTHER to the less frequently used D_INFO.
1239 Severity : enhancement
1241 Description: add -gid, -group, -uid, -user options to lfs find
1245 Description: Hung threads in invalidate_inode_pages2_range
1246 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1247 one is completed. As a result, some RPCs are stuck in the queue
1252 Description: filter threads hungs on waiting journal commit
1253 Details : Cleanup filter group llog code, then only filter group llog will
1254 be only created in the MDS/OST syncing process.
1258 Description: Procfs and llog threads access destoryed import sometimes.
1259 Details : Sync the import destoryed process with procfs and llog threads by
1260 the import refcount and semaphore.
1262 Severity : enhancement
1264 Description: openlock cache of b1_6 port to HEAD
1269 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1270 Details : If insertion of an extent fails, then discard the inode
1271 preallocation and free data blocks else it can lead to duplicate
1276 Description: don't always update ctime in ext3_xattr_set_handle()
1277 Details : Current xattr code updates the inode ctime in ext3_xattr_set_handle.
1278 In some cases the ctime should not be updated, for example for
1279 2.0->1.8 compatibility it is necessary to delete an xattr and it
1280 should not update the ctime.
1284 Bugzilla : 15713/16362
1285 Description: Assertion in iopen_connect_dentry in 1.6.3
1286 Details : looking up an inode via iopen with the wrong generation number can
1287 populate the dcache with a disconneced dentry while the inode
1288 number is in the process of being reallocated. This causes an
1289 assertion failure in iopen since the inode's dentry list contains
1290 both a connected and disconnected dentry.
1294 Description: assertion failure in ldlm_handle2lock()
1295 Details : fix a race between class_handle_unhash() and class_handle2object()
1296 introduced in lustre 1.6.5 by bug 13622.
1301 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1302 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1307 Description: oops in sd_iostats_seq_show()
1308 Details : unloading/reloading the scsi low level driver triggers a kernel
1309 bug when trying to access the sd iostat file.
1314 Description: Kernel panics during QLogic driver reload
1315 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1316 patch, causing memory corruption.
1321 Description: journal_dev option does not work in b1_6
1322 Details : pass mount option during pre-mount.
1324 Severity : enhancement
1326 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1327 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1328 extent information of a file. It can be used to map logical blocks
1329 in a file to physical blocks in the block device.
1333 Description: LDLM soft lockups - improvement
1334 Details : It is be possible to send the lock handle along with each read
1335 or write request because the client is already doing a lock match
1336 itself so there isn't any reason the OST should have to re-do that
1339 --------------------------------------------------------------------------------
1341 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1343 * Support for kernels:
1344 2.6.5-7.283 (SLES 9),
1345 2.6.9-55.EL (RHEL 4),
1346 2.6.16.46-0.14 (SLES 10),
1347 2.6.18.8 vanilla (kernel.org)
1348 * Client support for unpatched kernels:
1349 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1350 2.6.16 - 2.6.22 vanilla (kernel.org)
1351 * Due to recently discovered recovery problems, we do not recommend
1352 using patchless RHEL 4 clients with this or any earlier release.
1353 * Recommended e2fsprogs version: 1.39.cfs8
1354 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1355 * Starting with this release, the ldiskfs backing filesystem required
1356 by Lustre is now in its own package, lustre-ldiskfs. This package
1357 should be installed. It is versioned separately from Lustre and
1358 may be released separately in future.
1362 Description: block reactivating mgc import until all deactivates complete
1363 Details : Fix race when failing back MDT/MGS to itself (testing)
1366 Frequency : at statup only
1368 Description: mds_lov_synchronize race leads to various problems
1369 Details : simultaneous MDT->OST connections at startup can cause the
1370 sync to abort, leaving the OSC in a bad state.
1372 Severity : enhancement
1374 Description: add optional extra BUILD_VERSION info
1375 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1376 to override the lustre version.
1379 Frequency : 2.6.18 servers only
1381 Description: ll_kern_mount() doesn't release the module reference
1382 Details : The ldiskfs module reference count never drops down to 0
1383 because ll_kern_mount() doesn't release the module reference.
1388 Description: server LBUG when using old ost_num_threads parameter
1389 Details : Accept the old ost_num_threads parameter but warn that it
1390 is deprecated, and fix an off-by-one error that caused an LBUG.
1395 Description: Transient SCSI error results in persistent IO issue
1396 Details : iobuf->dr_error is not reinitialized to 0 between two
1400 Frequency : sometimes when underlying device returns I/O errors
1402 Description: OSTs not going read-only during write failures
1403 Details : OSTs are not remounted read-only when the journal commit threads
1404 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1405 instead of the ext3 wrappers.
1409 Description: poor connect scaling with increasing client count
1410 Details : Don't run filter_grant_sanity_check for more than 100 exports
1411 to improve scaling for large numbers of clients.
1414 Frequency : SLES10 only
1416 Description: sanity-quota.sh quotacheck failed: rc = -22
1417 Details : Quotas cannot be enabled on SLES10.
1420 Frequency : liblustre clients only
1422 Description: getdirentries does not give error when run on compute nodes
1423 Details : getdirentries does not fail when the size specified as an argument
1424 is too small to contain at least one entry
1426 Severity : enhancement
1428 Description: Add LNET router traceability for debug purposes
1429 Details : If a checksum failure occurs with a router as part of the
1430 IO path, the NID of the last router that forwarded the bulk data
1431 is printed so it can be identified.
1436 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1437 Details : Due to a race condition, liblustre clients were occasionally
1438 evicted incorrectly.
1440 Severity : enhancement
1442 Description: lfs setstripe use optional parameters instead of postional
1445 Severity : enhancement
1447 Description: Nanosecond timestamp support for ldiskfs
1448 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1449 resolution timestamps. There is not yet support for this at
1450 the Lustre filesystem level.
1453 Frequency : during server recovery
1455 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1456 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1460 Frequency : nfs export on patchless client
1462 Description: connectathon hang when test nfs export over patchless client
1463 Details : Disconnected dentry cannot be found with lookup, so we do not need
1464 to unhash it or make it invalid
1467 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1468 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1469 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1474 Description: open req refcounting wrong on reconnect
1475 Details : If reconnect happened between getting open reply from server and
1476 call to mdc_set_replay_data in ll_file_open, we will schedule
1477 replay for unreferenced request that we are about to free.
1478 Subsequent close will crash in variety of ways.
1479 Check that request is still eligible for replay in
1480 mdc_set_replay_data().
1485 Description: disable writes to filesystem when reading health_check file
1486 Details : the default for reading the health_check proc file has changed
1487 to NOT do a journal transaction and write to disk, because this
1488 can cause reads of the /proc file to hang and block HA state
1489 checking on a healthy but otherwise heavily loaded system. It
1490 is possible to return to the previous behaviour during configure
1491 with --enable-health-write.
1493 Severity : enhancement
1495 Description: 64-bit inode version
1496 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1497 to the inode. This will be required for version-based recovery.
1502 Description: MDS fails to start if a duplicate client export is detected
1503 Details : in some rare cases it was possible for a client to connect to
1504 an MDS multiple times. Upon recovery the MDS would detect this
1505 and fail during startup. Handle this more gracefully.
1507 Severity : enhancement
1509 Description: Add -o localflock option to simulate old noflock
1511 Details : This will achieve local-only flock/fcntl locks
1517 Description: log_commit_thread vs filter_destroy race leads to crash
1518 Details : Take import reference before releasing llog record semaphore
1523 Description: Wrong request locking in request set processing
1524 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1525 imp_delayed_list, in this place should be used imp_lock.
1528 Frequency : when reconnection
1530 Description: Grant Leak when osc reconnect to OST
1531 Details : When osc reconnect ost, OST(filter) should check whether it
1532 should grant more space to client by comparing fed_grant and
1533 cl_avail_grant, and return the granted space to client instead
1534 of "new granted" space, because client will call osc_init_grant
1535 to update the client grant space info.
1538 Frequency : when client reconnect to OST
1540 Description: Grant Leak when osc do resend and replay bulk write
1541 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1542 bulk write request, because the grant info while be sync between
1543 OSC and OST when reconnect, and we should ignore the grant info
1544 these of resend/replay write req.
1549 Description: Grant space more than avaiable left space sometimes.
1550 Details : When then OST is about to be full, if two bulk writing from
1551 different clients came to OST. Accord the avaliable space of the
1552 OST, the first req should be permitted, and the second one
1553 should be denied by ENOSPC. But if the seconde arrived before
1554 the first one is commited. The OST might wrongly permit second
1555 writing, which will cause grant space > avaiable space.
1558 Frequency : when client is evicted
1560 Description: Grant might be wrongly erased when osc is evicted by OST
1561 Details : when the import is evicted by server, it will fork another
1562 thread ptlrpc_invalidate_import_thread to invalidate the
1563 import, where the grant will be set to 0. While the original
1564 thread will update the grant it got when connecting. So if
1565 the former happened latter, the grant will be wrongly errased
1566 because of this race.
1571 Description: Checking Stale with correct fid
1572 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1573 to check whether it is stale, because sometimes, we want the
1574 enqueue happened anyway, and op_data.fid2 will not be initialized.
1576 Severity : enhancement
1578 Description: update patchless client
1579 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1582 Frequency : only with 2.4 kernel
1584 Description: random memory corruption
1585 Details : size of struct ll_inode_info is to big for union inode.u and this
1586 can be cause of random memory corruption.
1591 Description: Memory leak in recovery
1592 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1593 It should also check obd_fail before fsfilt_start, otherwise if
1594 fsfilt_start return -EROFS,(failover mds during mds recovery).
1595 then the req will return with repmsg->transno = 0 and rc = EROFS.
1596 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1597 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1598 be freed no matter whether fsfilt_commit success or not.
1601 Frequency : only with huge count clients
1603 Description: Prevents from taking the superblock lock in llap_from_page for
1605 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1606 for llap_from_page call in ll_removepage prevents from taking the
1607 superblock lock for a soon died page.
1612 Description: Not check open intent error before release open handle
1613 Details : in some rare cases, the open intent error is not checked before
1614 release open handle, which may cause
1615 ASSERTION(open_req->rq_transno != 0), because it tries to release
1616 the failed open handle.
1621 Description: Set cat log bitmap only after create log success.
1622 Details : in some rare cases, the cat log bitmap is set too early. and it
1623 should be set only after create log success.
1627 Description: Accessing a block bevice can re-enable I/O when Lustre is
1628 tearing down a device.
1629 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1634 Description: service threads may hog cpus when there are a lot of requests
1635 Details : Insert cond_resched to give other threads a chance to use some CPU
1640 Description: the cat log was not initialized in recovery
1641 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1642 unlink log on mds will not be initialized until mds post
1643 recovery. And also in mds post recovery, the unlink log will
1644 initialization will be done asynchronausly, so there will be race
1645 between add unlink log and unlink log initialization.
1649 Description: brw_stats were being printed incorrectly
1650 Details : brw_stats were being printed as log2 but all of them were not
1651 recorded as log2. Also remove some code duplication arising from
1652 filter_tally_{read,write}.
1656 Frequency : rare, only in recovery.
1657 Description: ASSERTION(req->rq_type != LI_POISON) failed
1658 Details : imp_lock should be held while iterating over imp_sending_list for
1659 prevent destroy request after get timeout in ptlrpc_queue_wait.
1663 Description: replay-single.sh test 52 fails
1664 Details : A lock's skiplist need to be cleanup when it being unlinked
1665 from its resource list.
1669 Description: Short directio read returns full requested size rather than
1671 Details : Direct I/O operations should return actual amount of bytes
1672 transferred rather than requested size.
1674 Severity : enhancement
1676 Description: metadata RPC reduction (e.g. for rm performance)
1677 Details : decrease the amount of synchronous RPC between clients and servers
1678 by canceling conflicing lock before the operation on the client
1679 and packing thier handles into the main operation RPC to server.
1681 Severity : enhancement
1683 Description: Async OSC create to avoid the blocking unnecessarily.
1684 Details : If a OST has no remain object, system will block on the creating
1685 when need to create a new object on this OST. Now, ways use
1686 pre-created objects when available, instead of blocking on an
1687 empty osc while others are not empty. If we must block, we block
1688 for the shortest possible period of time.
1690 Severity : enhancement
1692 Description: refine locking for avoid write wrong info into lov_objid file
1693 Details : fix possible races with add new target and write/update data in
1696 --------------------------------------------------------------------------------
1698 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1703 Frequency : on some architectures
1705 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1706 Details : Uninitialized flags sometimes cause configuration commands to
1710 Frequency : patchless clients only
1712 Description: missing __iget() symbol export
1713 Details : The __iget() symbol export is missing. To avoid the need for
1714 this on patchless clients the deathrow inode reaper is turned
1715 off, and we depend on the VM to clean up old inodes. This
1716 dependency was during via the fix for bug 12181.
1718 --------------------------------------------------------------------------------
1720 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1722 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1723 INTEROPERATE with older versions automatically. In many cases a
1724 special upgrade step is needed. Please read the
1725 user documentation before upgrading any part of a live system.
1726 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1727 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1728 * WARNING: Lustre configuration and startup changes are required with
1729 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1735 Severity : enhancement
1737 Description: MountConf
1738 Details : Lustre configuration is now managed via mkfs and mount
1739 commands instead of lmc and lconf. New obd types (MGS, MGC)
1740 are added for dynamic configuration management. See
1741 https://mail.clusterfs.com/wikis/lustre/MountConf for
1744 Severity : enhancement
1746 Description: dynamic OST addition
1747 Details : OSTs can now be added to a live filesystem
1749 Severity : enhancement
1751 Description: startup order invariance
1752 Details : MDTs and OSTs can be started in any order. Clients only
1753 require the MDT to complete startup.
1755 Severity : enhancement
1757 Description: parallel, asynchronous orphan cleanup
1758 Details : orphan cleanup is now performed in separate threads for each
1759 OST, allowing parallel non-blocking operation.
1761 Severity : enhancement
1763 Description: optimized stripe assignment
1764 Details : stripe assignments are now made based on ost space available,
1765 ost previous usage, and OSS previous usage, in order to try
1766 to optimize storage space and networking resources.
1768 Severity : enhancement
1770 Description: Permanently set tunables
1771 Details : All writable /proc/fs/lustre tunables can now be permanently
1772 set on a per-server basis, at mkfs time or on a live system.
1774 Severity : enhancement
1776 Description: Lustre message v2
1777 Details : Add lustre message format v2.
1779 Severity : enhancement
1781 Description: client OST exclusion list
1782 Details : Clients can be started with a list of OSTs that should be
1783 declared "inactive" for known non-responsive OSTs.
1787 Description: SPEC SFS validation failure on NFS v2 over lustre.
1788 Details : Changes the blocksize for regular files to be 2x RPC size,
1789 and not depend on stripe size.
1791 Severity : enhancement
1793 Description: Multiple MD RPCs in flight.
1794 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1795 To avoid overly-overloading MDS, introduce a limit on number of
1796 MDS RPCs in flight for a single client and add /proc controls
1797 to adjust this limit.
1799 Severity : enhancement
1801 Description: client read/write statistics
1802 Details : Add client read/write call usage stats for performance
1803 analysis of user processes.
1804 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1805 file access. extents_stats shows chunk size distribution.
1806 extents_stats_per_process show chunk size distribution per
1809 Severity : enhancement
1811 Description: mds statistics
1812 Details : Add detailed mds operations statistics in
1813 /proc/fs/lustre/mds/*/stats.
1817 Description: Failure of copying files with lustre special EAs.
1818 Details : Client side always return success for setxattr call for lustre
1819 special xattr (currently only "trusted.lov").
1824 Description: Refcount LNET uuids
1825 Details : The global LNET uuid list grew linearly with every startup;
1826 refcount repeated list entries instead of always adding to
1829 Severity : enhancement
1831 Description: Dynamic service threads
1832 Details : Within a small range, start extra service threads
1833 automatically when the request queue builds up.
1836 Frequency : mixed-endian client/server environments
1838 Description: mixed-endian crashes
1839 Details : The new msg_v2 system had some failures in mixed-endian
1842 Severity : enhancement
1844 Description: Easy OST removal
1845 Details : OSTs can be permanently deactivated with e.g. 'lctl
1846 conf_param lustre-OST0001.osc.active=0'
1848 Severity : enhancement
1850 Description: MGS proc entries
1851 Details : Added basic proc entries for the MGS showing what filesystems
1854 Severity : enhancement
1856 Description: provide MGS failover
1857 Details : Added config lock reacquisition after MGS server failover.
1859 Severity : enhancement
1861 Description: add Linux 2.4 support
1862 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1866 Description: a large application tries to do I/O to the same resource and dies
1867 in the middle of it.
1868 Details : Check the req->rq_arrival time after the call to
1869 ost_brw_lock_get(), but before we do anything about
1870 processing it & sending the BULK transfer request. This
1871 should help move old stale pending locks off the queue as
1872 quickly as obd_timeout.
1875 Frequency : when an incorrect nid is specified during startup
1877 Description: ptlrpc connect to non-existant node causes kernel crash
1878 Details : LNET can't be re-entered from an event callback, which
1879 happened when we expire a message after the export has been
1880 cleaned up. Instead, hand the zombie cleanup off to another
1883 Severity : enhancement
1885 Description: plain/inodebits lock performance improvement
1886 Details : Grouping plain/inodebits in granted list by their request modes
1887 and bits policy, thus improving the performance of search through
1891 Frequency : only if OST filesystem is corrupted
1893 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1894 Details : for a short time RPCs with bulk IO are in the replay list,
1895 but replay of bulk IOs is unimplemented. If the OST filesystem
1896 is corrupted due to disk cache incoherency and then replay is
1897 started it is possible to trip an assertion. Avoid putting
1898 committed RPCs into the replay list at all to avoid this issue.
1901 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1903 Description: System hang on startup
1904 Details : This bug allowed the liblustre (e.g. catamount) client to
1905 return to the app before handling all startup RPCs. This
1906 could leave the node unresponsive to lustre network traffic
1907 and manifested as a server ptllnd timeout.
1909 Severity : enhancement
1911 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1912 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1913 boolean to control whether to print peer debug info when a
1914 client's RPC times out.
1917 Frequency : only for kernels with patches from Lustre below 1.4.3
1919 Description: Remove old rdonly API
1920 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1923 Frequency : only for devices with external journals
1925 Description: Set external device read-only also
1926 Details : During a commanded failover stop, we set the disk device
1927 read-only while the server shuts down. We now also set any
1928 external journal device read-only at the same time.
1931 Frequency : when upgrading from 1.4 while trying to change parameters
1933 Description: The wrong (new) MDC name was used when setting parameters for
1934 upgraded MDT's. Also allows changing of OSC (and MDC)
1935 parameters if --writeconf is specified at tunefs upgrade time.
1938 Frequency : when setting specific ost indicies
1940 Description: QOS code breaks on skipped indicies
1941 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1942 created with --index need not be sequential.
1944 Severity : enhancement
1946 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1947 Details : The uninit_groups feature works in conjunction with the kernel
1948 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1949 up the pass1 processing of e2fsck. This is a read-only feature
1950 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1951 mount filesystems that have had this feature enabled.
1953 Severity : enhancement
1955 Description: Improve multi-block allocation algorithm to avoid fragmentation
1956 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1957 allocation locality and avoid filesystem fragmentation.
1959 ------------------------------------------------------------------------------
1961 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1963 * Support for kernels:
1964 2.4.21-47.0.1.EL (RHEL 3)
1965 2.6.5-7.283 (SLES 9)
1966 2.6.9-42.0.10.EL (RHEL 4)
1967 2.6.12.6 vanilla (kernel.org)
1968 2.6.16.27-0.9 (SLES 10)
1969 * Recommended e2fsprogs version: 1.39.cfs5
1971 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1975 Frequency : occasional, depends on client load and configuration
1976 Bugzilla : 12181, 12203
1977 Description: data loss for recently-modified files
1979 Details : In some cases it is possible that recently written or created
1980 files may not be written to disk in a timely manner (this should
1981 normally be within 30s unless client IO load is very high).
1982 The problem appears as zero-length files or files that are a
1983 multiple of 1MB in size after a client crash or client eviction
1984 that are missing data at the end of the file.
1986 This problem is more likely to be hit on clients where files are
1987 repeatedly created and unlinked in the same directory, clients
1988 have a large amount of RAM, have many CPUs, the filesystem has
1989 many OSTs, the clients are rebooted frequently, and/or the files
1990 are not accessed by other nodes after being written.
1992 The presence of the problem can be detected by looking at
1993 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1994 smaller than the second (nr_unused) then dirty files will not
1995 be flushed automatically to disk. "sync; sleep 10" should be
1996 run several times on the node before unmounting it to update
1997 Lustre (this is also safe to run on nodes without this problem).
1999 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2000 that can cause this same problem, so customers using that kernel
2001 also need to update the kernel in addition to Lustre. In order
2002 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2004 It is normal that files written just before a client crash (less
2005 than 30s) may not yet have been flushed to disk, even for local
2009 Frequency : frequent on thin XT3 nodes
2011 Description: UUID collision on thin XT3 Linux nodes
2012 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2013 randomly, since we relied on an insufficiently-seeded PRNG.
2018 Description: OSS hangs after "All ost request buffers busy"
2019 Details : A deadlock between quota and journal operations caused OSS
2020 hangs after printing "All ost request buffers busy."
2023 Frequency : always on liblustre builds
2025 Description: Cleanup compiler warnings on liblustre
2028 Frequency : always on liblustre builds on XT3
2030 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2034 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2035 > 32000 subdirectories
2036 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2037 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2038 subdirectory count crosses 32000. This will aid e2fsck to
2039 correctly handle more than 32000 subdirectories.
2042 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2044 Description: System hang on startup
2045 Details : This bug allowed the liblustre (e.g. catamount) client to
2046 return to the app before handling all startup RPCs. This
2047 could leave the node unresponsive to lustre network traffic
2048 and manifested as a server ptllnd timeout.
2050 Severity : enhancement
2052 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2053 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2054 boolean to control whether to print peer debug info when a
2055 client's RPC times out.
2060 Description: make O_SYNC working on 2.6 kernels
2061 Details : 2.6 kernels use different method for mark pages for write,
2062 so need add a code to lustre for O_SYNC work.
2067 Description: Failure to close file and release space on NFS
2068 Details : Put inode details into lock acquired in ll_intent_file_open.
2069 Use mdc_intent_lock in ll_intent_open to properly
2070 detect all kind of errors unhandled by mdc_enqueue.
2075 Description: proc file read during shutdown sometimes raced obd removal,
2077 Details : Add lock to prevent obd access after proc file removal.
2080 Frequency : Only for files larger than 4GB on 32-bit clients.
2082 Description: improperly doing page alignment of locks
2083 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2084 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2089 Description: RPCs being resent when they shouldn't be
2090 Details : Some RPCs that should not be resent are being resent. This
2091 can cause inconsistencies in the RPC state machine. Do not
2092 resend such requests.
2095 Frequency : rare, only with NFS export
2097 Description: Crash on NFS re-export node
2098 Details : under very unusual load conditions an assertion is hit in
2099 ll_intent_file_open()
2102 Frequency : only if OST filesystem is corrupted
2104 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2105 Details : for a short time RPCs with bulk IO are in the replay list,
2106 but replay of bulk IOs is unimplemented. If the OST filesystem
2107 is corrupted due to disk cache incoherency and then replay is
2108 started it is possible to trip an assertion. Avoid putting
2109 committed RPCs into the replay list at all to avoid this issue.
2114 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2115 Details : Large single O_DIRECT read and write calls can fail to allocate
2116 a sufficiently large buffer to process the request. In case of
2117 allocation failure the allocation is retried with a smaller
2118 buffer and broken into smaller requests.
2120 Severity : enhancement
2122 Description: Add -o localflock option to simulate old noflock behaviour.
2123 Details : This will achieve local-only flock/fcntl locks coherentness.
2128 Description: versioning check is incomplete
2129 Details : Checking the version difference of client vs. server, report
2130 error if the gap is too big.
2135 Description: add support PG_writeback bit
2136 Details : add support for PG_writeback bit for Lustre, for more carefull
2137 work with page cache in 2.6 kernel. This also fix some deadlocks
2138 and remove hack for work O_SYNC with 2.6 kernel.
2140 Severity : enhancement
2142 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2143 Details : The uninit_groups feature works in conjunction with the kernel
2144 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2145 up the pass1 processing of e2fsck. This is a read-only feature
2146 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2147 mount filesystems that have had this feature enabled.
2149 Severity : enhancement
2151 Description: Improve multi-block allocation algorithm to avoid fragmentation
2152 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2153 allocation locality and avoid filesystem fragmentation.
2155 ------------------------------------------------------------------------------
2157 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2159 * Support for kernels:
2160 2.6.9-42.0.3.EL (RHEL 4)
2161 2.6.5-7.276 (SLES 9)
2162 2.4.21-47.0.1.EL (RHEL 3)
2163 2.6.12.6 vanilla (kernel.org)
2164 2.6.16.21-0.8 (SLES10)
2165 * Recommended e2fsprogs version: 1.39.cfs2-0
2167 * The backwards-compatible /proc/sys/portals symlink has been removed
2168 in this release. Before upgrading, please ensure that you change
2169 any configuration scripts or /etc/sysctl.conf files that access
2170 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2171 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2172 in advance of the upgrade on any system running Lustre 1.4.6 or
2173 newer, since /proc/sys/lnet was added in that version.
2174 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2178 Frequency : only when quota is used
2180 Description: avoid scanning export list for quota master
2181 Details : Change the algorithms to avoid scanning export list in order
2182 to improve the efficiency.
2185 Frequency : MDS failover only, very rarely
2187 Description: "went back in time" messages on mds failover
2188 Details : The greatest transno may be lost when the current operation
2189 finishes with an error (transno==0) and the client's last_rcvd
2190 record is over-written. Save the greatest transno in the
2191 mds_last_transno for this case.
2194 Frequency : always for specific kernels and striping counts
2196 Description: client may get "Matching packet too big" without ACL support
2197 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2198 when trying to access files in certain configurations. The
2199 clients should in fact be denied when mounting because they do
2200 not understand ACLs.
2203 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2205 Description: many clients connecting with IO in progress causes connect timeouts
2206 Details : Avoid synchronous journal commits to avoid delays caused by many
2207 clients connecting/disconnecting when bulk IO is in progress.
2208 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2209 OST_IO_PORTAL to avoid delays behind potentially many pending
2213 Frequency : occasionally with multiple writers to a single file
2215 Description: shared writes to file may result in wrong size reported by stat()
2216 Details : Allow growing of kms when extent lock is cancelled
2219 Frequency : always with random mmap IO to multi-striped file
2221 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2222 Details : Only if the hole is at the end of OST object so that kms is too
2223 small. Fix is to increase kms accordingly in ll_nopage.
2226 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2228 Description: writes to a missing object would leak memory on the OST
2229 Details : If there is an inconsistency between the MDS and OST filesystems,
2230 such that the MDS references an object that doesn't exist, writes
2231 to that object will leak memory due to incorrect cleanup in the
2232 error handling path, eventually running out of memory on the OST.
2237 Description: Creating too long symlink causes lustre errors
2238 Details : Check symlink and name lengths before sending requests to MDS.
2241 Frequency : only if flock is enabled (not on by default)
2243 Description: posix locks not released on fd closure on 2.6.9+
2244 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2245 kernels, this caused such locks not to be released on fd close and
2246 then assertions on fs unmount about still used locks.
2249 Frequency : MDS failover only, very rarely
2251 Description: clients may get ASSERTION(granted_lock != NULL)
2252 Details : When request was taking a long time, and a client was resending
2253 a getattr by name lock request. The were multiple lock requests
2254 with the same client lock handle and
2255 mds_getattr_name->fixup_handle_for_resent_request found one of the
2256 lock handles but later failed with ASSERTION(granted_lock != NULL).
2261 Description: handle->h_buffer_credits > 0, assertion failure
2262 Details : h_buffer_credits is zero after truncate, causing assertion
2263 failure. This patch extends the transaction or creates a new
2267 Frequency : NFS re-export or patchless client
2268 Bugzilla : 11179, 10796
2269 Description: Crash on NFS re-export node (__d_move)
2270 Details : We do not want to hash the dentry if we don't have a lock.
2271 But if this dentry is later used in d_move, we'd hit uninitialised
2272 list head d_hash, so we just do this to init d_hash field but
2273 leave dentry unhashed.
2276 Frequency : NFS re-export or patchless client
2278 Description: NFS exports has problem with symbolic link
2279 Details : lustre client didn't properly install dentry when re-exported
2280 to NFS or running patchless client.
2283 Frequency : NFS re-export or patchless client
2285 Description: Various nfs/patchless fixes.
2286 Details : fixes reuse disconected alias for lookup process - this fixes
2287 warning "find_exported_dentry: npd != pd",
2288 fix permission error with open files at nfs.
2289 fix apply umask when do revalidate.
2292 Frequency : occasional
2294 Description: Crash on NFS re-export node
2295 Details : calling clear_page() on the wrong pointer triggered oops in
2296 generic_mapping_read().
2299 Frequency : rarely, using O_DIRECT IO
2301 Description: unaligned directio crashes client with LASSERT
2302 Details : check for unaligned buffers before trying any requests.
2305 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2307 Description: stale data returned from RAID cache
2308 Details : If only a small amount of IO is done to the RAID device before
2309 reading it again it is possible to get stale data from the RAID
2310 cache instead of reading it from disk.
2313 Frequency : always for sles10 kernel
2315 Description: sles10 support
2316 Details : ll_follow_link: compile fixes and using of nd_set_link
2317 under newer kernels.
2320 Frequency : depends on arch, kernel and compiler version, always on sles10
2323 Description: recursive or deep enough symlinks cause stack overflow
2324 Details : getting rid of large stack-allocated variable in
2328 Frequency : depends on hardware
2330 Description: lustre write performance loss in the SLES10 kernel
2331 Details : the performance loss is caused by using of write barriers in the
2332 ext3 code. The SLES10 kernel turns barrier support on by
2333 default. The fix is to undo that change for ldiskfs.
2335 ------------------------------------------------------------------------------
2337 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2339 * Support for kernels:
2340 2.6.9-42.0.3EL (RHEL 4)
2341 2.6.5-7.276 (SLES 9)
2342 2.4.21-47.0.1.EL (RHEL 3)
2343 2.6.12.6 vanilla (kernel.org)
2347 Frequency : quota enabled and large files being deleted
2349 Description: releasing more than 4GB of quota at once hangs OST
2350 Details : If a user deletes more than 4GB of files on a single OST it
2351 will cause the OST to spin in an infinite loop. Release
2352 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2357 Description: statfs data retrieved from /proc may be stale or zero
2358 Details : When reading per-device statfs data from /proc, in the
2359 {kbytes,files}_{total,free,avail} files, it may appear
2360 as zero or be out of date.
2363 Frequency : systems with MD RAID1 external journal devices
2365 Description: lconf's call to blkid is confused by RAID1 journal devices
2366 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2367 of returning all block devices that match the journal UUID.
2370 Frequency : always, for aggregate stripe size over 4GB
2372 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2373 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2374 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2377 Frequency : always if "lfs find" used on a local file/directory
2379 Description: "lfs find" segfaults if used on a local file/directory
2380 Details : The case where a directory component was not specified wasn't
2381 handled correctly. Handle this properly.
2384 Frequency : always on ppc64
2386 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2387 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2388 perform endianness conversion.
2391 Frequency : rarely (truncate to non-zero file size after write under load)
2392 Bugzilla : 10730, 10687
2393 Description: Files padded with zeros to next 4K multiple
2394 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2395 it is possible that files that are truncated to a non-zero size
2396 immediately after being written are filled with zero bytes beyond
2397 the truncated size. No file data is lost.
2399 Severity : enhancement
2401 Description: Allow recovery/failover for liblustre clients.
2402 Details : liblustre clients were unaware of failover configurations until
2405 Severity : enhancement
2407 Description: user file locks should fail when not mounting with flock option
2408 Details : Set up an error-returning stub in ll_file_operations.lock field
2409 to prevent incorrect behaviour when client is mounted without
2410 flock option. Also, set up properly f_op->flock field for
2414 Frequency : always on ia64
2416 Description: "lfs df" loops on printing out MDS statfs information
2417 Details : The obd_ioctl_data was not initialized and in some systems
2418 this caused a failure during the ioctl that did not return
2419 an error. Initialize the struct and return an error on failure.
2422 Frequency : SLES 9 only
2424 Description: Error of copying files with lustre special EAs as root
2425 Details : Client side always return success for setxattr call for lustre
2426 special xattr (currently only "trusted.lov").
2429 Frequency : rarely on clusters with both ia64+i386 clients
2431 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2432 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2433 clients are doing concurrent writes to the same file it is
2434 possible that the ia64 clients may LASSERT because the OST
2435 extent locks are not PAGE_SIZE aligned. Ensure that grown
2436 locks are always aligned on the request boundary.
2439 Frequency : specific use, occasional
2441 Description: Overwriting in use executable truncates on-disk binary image
2442 Details : If one node attempts to overwrite an executable in use by
2443 another node, we now correctly return ETXTBSY instead of
2444 truncating the file.
2446 Severity : enhancement
2448 Description: Async OSC create to avoid the blocking unnecessarily.
2449 Details : If a OST has no remain object, system will block on the creating
2450 when need to create a new object on this OST. Now, ways use
2451 pre-created objects when available, instead of blocking on an
2452 empty osc while others are not empty. If we must block, we block
2453 for the shortest possible period of time.
2458 Description: chmod on Lustre root is propagated to other clients
2459 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2460 invalid by the follow_mount time.
2463 Frequency : liblustre clients only
2465 Description: Race in 'instant cancel' lock handling could lead to such locks
2466 never to be granted in case of SMP MDS
2467 Details : Do not destroy not yet granted but cbpending locks in
2471 Frequency : replay/resend of open
2473 Description: non null lock assetion failure in mds_intent_policy
2474 Details : Trying to replay/resend lockless open requests resulted in
2475 mds_open() returning 0 with no lock. Now it sets a flag if
2476 a lock is going to be returned.
2478 Severity : enhancement
2480 Description: Checksum enhancements
2481 Details : New checksum enhancements allow for resending RPCs that failed
2484 Severity : enhancement
2486 Description: Tunables on number of dirty pages in cacche
2487 Details : Allow to set limit on number of dirty pages cached.
2492 Description: client crash on unmount - lock still has references
2493 Details : In some error handling cases it was possible to leak a lock
2494 reference on a client while accessing a file. This was not
2495 harmful to the client during operation, but would cause the
2496 client to crash when the filesystem is unmounted.
2499 Frequency : specific case, rare
2501 Description: ETXTBSY on mds though file not in use by client
2502 Details : ETXTBSY is no longer incorrectly returned when attempting to
2503 chmod or chown a directory that the user previously tried to
2504 execute or a currently-executing binary.
2507 Frequency : extremely rare except on liblustre-based clients
2509 Description: Lustre space not freed when files are deleted
2510 Details : Clean up open-unlinked files after client eviction. Previously
2511 the unlink was skipped and the files remained as orphans.
2516 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2517 Details : In some cases it was possible to send a blocking callback to
2518 a client doing a glimpse, even though that client didn't get
2519 a lock granted. When the glimpse lock is cancelled on the OST
2520 the freed lock is left on the waiting list and corrupted the list.
2523 Frequency : all core dumps
2525 Description: Broke core dumps to lustre
2526 Details : Negative dentry may be unhashed if parent does not have UPDATE
2527 lock, but some callers, e.g. do_coredump, expect dentry to be
2528 hashed after successful create, hash it in ll_create_it.
2530 ------------------------------------------------------------------------------
2532 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2534 * Support for kernels:
2535 2.6.9-42.0.2.EL (RHEL 4)
2536 2.6.5-7.276 (SLES 9)
2537 2.4.21-40.EL (RHEL 3)
2538 2.6.12.6 vanilla (kernel.org)
2542 Frequency : always on RHEL 3
2544 Description: Number of open files grows over time
2545 Details : The number of open files grows over time, whether or not
2546 Lustre is started. This was due to a filp leak introduced
2547 by one of our kernel patches.
2549 ------------------------------------------------------------------------------
2551 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2553 * Support for kernels:
2554 2.6.9-42.EL (RHEL 4)
2555 2.6.5-7.276 (SLES 9)
2556 2.4.21-40.EL (RHEL 3)
2557 2.6.12.6 vanilla (kernel.org)
2562 Bugzilla : 5719, 9635, 9792, 9684
2563 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2564 Details : If a server is under heavy load and cannot reply to new
2565 connection requests before the client resends the (re)connect,
2566 the connection handling code can behave badly if two service
2567 threads are concurrently handing separate (re)connections from
2568 the same client. Add better locking to the connection handling
2569 code, and ensure that only a single connection will be processed
2570 for a given client UUID, even if the lock is dropped.
2572 Severity : enhancement
2574 Description: add TCP zero-copy support to kernel
2575 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2576 sends if the hardware supports scatter-gather and checksumming.
2577 This allows socklnd to do client-write and server-read more
2578 efficiently and reduce CPU utilization from skbuf copying.
2581 Frequency : only if NFS exporting from client
2583 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2584 Details : If setting attributes on a file created under NFS that had
2585 never been opened it would be possible to oops the client
2586 if the file had no objects.
2589 Frequency : always for liblustre
2591 Description: liblustre client does MDS+OSTs setattr RPC for each write
2592 Details : When doing a write from a liblustre client, the client
2593 incorrectly issued an RPC to the MDS and each OST the file was
2594 striped over in order to update the timestamps. When writing
2595 with small chunks and many clients this could overwhelm the MDS
2596 with RPCs. In all cases it would slow down the write because
2597 these RPCs are unnecessary.
2599 Severity : enhancement
2601 Description: allow number of MDS service threads to be changed at module load
2602 Details : It is now possible to change the number of MDS service threads
2603 running. Adding "options mds mds_num_threads={N}" to the MDS's
2604 /etc/modprobe.conf will set the number of threads for the next
2605 time Lustre is restarted (assuming the "mds" module is also
2606 reloaded at that time). The default number of threads will
2607 stay the same, 32 for most systems.
2612 Description: OST crash if filesystem is unformatted or corrupt
2613 Details : If an OST is started on a device that has never been formatted
2614 or if the filesystem is corrupt and cannot even mount then the
2615 error handling cleanup routines would dereference a NULL pointer.
2620 Description: NULL pointer deref in llap_from_page.
2621 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2622 incorrect) mapping if the page was truncated/reclaimed while it was
2623 searched for. Check for this condition and skip such pages when
2624 doing readahead. Introduce extra check to llap_from_page() to
2625 verify page->mapping->host is non-NULL (so page is not anonymous).
2628 Frequency : Sometimes when using sys_sendfile
2630 Description: "page not covered by a lock" warnings from ll_readpage
2631 Details : sendfile called ll_readpage without right page locks present.
2632 Now we introduced ll_file_sendfile that does necessary locking
2633 around call to generic_file_sendfile() much like we do in
2637 Frequency : with certain MDS communication failures at client mount time
2639 Description: NULL pointer deref after failed client mount
2640 Details : a client connection request may delayed by the network layer
2641 and not be sent until after the PTLRPC layer has timed out the
2642 request. If the client fails the mount immediately it will try
2643 to clean up before the network times out the request. Add a
2644 reference from the request import to the obd device and delay
2645 the cleanup until the network drops the request.
2648 Frequency : occasionally during client (re)connect
2650 Description: assertion failure during client (re)connect
2651 Details : processing a client connection request may be delayed by the
2652 client or server longer than the client connect timeout. This
2653 causes the client to resend the connection request. If the
2654 original connection request is replied in this interval, the
2655 client may trip an assertion failure in ptlrpc_connect_interpret()
2656 which thought it would be the only running connect process.
2659 Frequency : only with obd_echo servers and clients that are rebooted
2661 Description: kernel BUG accessing uninitialized data structure
2662 Details : When running an obd_echo server it did not start the ping_evictor
2663 thread, and when a client was evicted an uninitialized data
2664 structure was accessed. Start the ping_evictor in the RPC
2665 service startup instead of the OBD startup.
2667 Severity : enhancement
2668 Bugzilla : 10193 (patchless)
2669 Description: Remove dependency on various unexported kernel interfaces.
2670 Details : No longer need reparent_to_init, exit_mm, exit_files,
2671 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2674 Frequency : rare (only users of deprecated and unsupported LDAP config)
2676 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2677 Details : LDAP apparently contains 'lustreName' attributes instead of
2678 'name'. A simple remapping of the name is sufficient.
2681 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2683 Description: waiting_locks_callback trips kernel BUG if client is evicted
2684 Details : Running with the dump_on_timeout debug flag turned on makes
2685 it possible that the waiting_locks_callback() can try to dump
2686 the Lustre kernel debug logs from an interrupt handler. Defer
2687 this log dumping to the expired_lock_main() thread.
2689 Severity : enhancement
2691 Description: Support NFS exporting on 2.6 kernels.
2692 Details : Implement non-rawops metadata methods for NFS server to use without
2693 changing NFS server code.
2696 Frequency : very rare (synthetic metadata workload only)
2698 Description: two racing renames might cause an MDS thread to deadlock
2699 Details : Running the "racer" program may cause one MDS thread to rename
2700 a file from being the source of a rename to being the target of
2701 a rename at exactly the same time that another thread is doing
2702 so, and the second thread has already enqueued these locks after
2703 doing a lookup of the target and is trying to relock them in
2704 order. Ensure that we don't try to re-lock the same resource.
2707 Frequency : only very large systems with liblustre clients
2709 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2710 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2711 all OSTs in parallel. This allows the eviction of stale liblustre
2712 clients to proceed much faster than if they were done in series,
2713 and also offers similar improvements for other set_info RPCs.
2718 Description: excessive CPU usage during initial read phase on client
2719 Details : During the initial read phase on a client, it would agressively
2720 retry readahead on the file, consuming too much CPU and impacting
2721 performance (since 1.4.5.8). Improve the readahead algorithm
2722 to avoid this, and also improve some other common cases (read
2723 of small files in particular, where "small" is files smaller than
2724 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2729 Description: MDS crash when receiving packet with unknown intent.
2730 Details : Do not LBUG in unknown intent case, just return -EFAULT
2732 Severity : enhancement
2733 Bugzilla : 9293, 9385
2734 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2735 Details : Do not serialize getattr (non-intent version) and statfs.
2738 Frequency : occasional, when OST network is overloaded/intermittent
2740 Description: client evicted by OST after bulk IO timeout
2741 Details : If a client sends a bulk IO request (read or write) the OST
2742 may evict the client if it is unresposive to its data GET/PUT
2743 request. This is incorrect if the network is overloaded (takes
2744 too long to transfer the RPC data) or dropped the OST GET/PUT
2745 request. There is no need to evict the client at all, since
2746 the pinger and/or lock callbacks will handle this, and the
2747 client can restart the bulk request.
2750 Frequency : Always when mmapping file with no objects
2752 Description: client crashes when mmapping file with no objects
2753 Details : Check that we actually have objects in a file before doing any
2754 operations on objects in ll_vm_open, ll_vm_close and
2760 Description: Request leak when working with deleted CWD
2761 Details : Introduce advanced request refcount tracking for requests
2762 referenced from lustre intent.
2764 Severity : Enhancement
2766 Description: Cache open file handles on client.
2767 Details : MDS now will return special lock along with openhandle, if
2768 requested and client is allowed to hold openhandle, even if unused,
2769 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2770 closing files for every read/write openration.
2772 Severity : Enhancement
2774 Description: Cache open negative dentries on client when possible.
2775 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2776 negative dentries on lock revocation.
2781 Description: Remounting a client read-only wasn't possible with a zconf mount
2782 Details : It wasn't possible to remount a client read-only with llmount.
2784 Severity : enhancement
2785 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2786 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2787 MPICH is included to add Lustre-specific ADIO interfaces.
2788 This is based closely on the UFS ADIO layer and only differs
2789 in file creation, in order to allow the OST striping to be set.
2790 This is user-contributed code and not supported by CFS.
2795 Description: extended inode attributes (immutable, append-only) work improperly
2796 when 2.4 and 2.6 kernels are used on client/server or vice versa
2797 Details : Introduce kernel-independent values for these flags.
2799 Severity : enhancement
2802 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2803 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2804 granularity. Now, Fractional MB granularity is be supported,
2805 this is very useful for low memory system.
2807 Severity : enhancement
2809 Description: Getattr by fid
2810 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2811 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2812 have locks and avoids getattr rpc requests for it on every stat.
2815 Frequency : Always, for filesystems larger than 2TB
2817 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2818 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2819 filesystems larger than 2TB to be mounted with mballoc enabled.
2822 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2824 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2825 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2826 filesystems larger than 2TB is mounted then the kernel will
2827 silently corrupt the start of the filesystem. CONFIG_LBD is
2828 enabled for all CFS-supported kernels, but the possibility of
2829 this happening with a modified kernel config exists.
2831 Severity : enhancement
2833 Description: add client O_DIRECT support for 2.6 kernels
2834 Details : It is now possible to do O_DIRECT reads and writes to files
2835 in the Lustre client mountpoint on 2.6 kernel clients.
2837 Severity : enhancement
2839 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2840 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2841 parallel, not waiting for response from every OST before sending
2842 a rpc to the next OST.
2847 Description: setattr vs write race when updating file timestamps
2848 Details : Client processes that update a file timestamp into the past
2849 right after writing to the file (e.g. tar) it is possible that
2850 the updated file modification time can be reset to the current
2851 time due to a race between processing the setattr and write RPC.
2853 Severity : enhancement
2855 Description: Bring 'lfs find' closer in line with regular Linux find.
2856 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2857 -print0 options and obtains all the needed info through the lustre
2860 Severity : enhancement
2862 Description: support up to 1024 configured devices on one node
2863 Details : change obd_dev array from statically allocated to dynamically
2864 allocated structs as they are first used to reduce memory usage
2869 Description: Flush dirty partially truncated pages during truncate
2870 Details : Immediatelly flush partially truncated pages in filter_setattr,
2871 this way we completely avoid having any pages in page cache on OST
2872 and can retire ugly workarounds during writes to flush such pages.
2877 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2878 Details : protect inode from truncation within vfs_unlink() context
2879 just take a reference before calling vfs_unlink() and release it
2880 when parent's i_sem is free.
2885 Description: last_id value checked outside lock on OST caused LASSERT failure
2886 Details : If there were multiple MDS->OST object precreate requests in
2887 flight, it was possible that the OST's last object id was checked
2888 outside a lock and incorrectly tripped an assertion. Move checks
2889 inside locks, and discard old precreate requests.
2892 Frequency : always, if extents are used on OSTs
2894 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2895 Details : OSTs using the extents format would not zero the high 16 bits of
2896 the index physical block number. This is not a problem for any
2897 OST filesystems smaller than 16TB, and no kernels support ext3
2898 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2899 new/modified files) and can be fixed for existing filesystems
2900 with e2fsprogs-1.39-cfs1.
2905 Description: import connection selection may be incorrect if timer wraps
2906 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2907 import connections to be ignored if the 32-bit jiffies counter
2908 wraps. Use a 64-bit jiffies counter.
2911 Frequency : very large clusters immediately after boot
2913 Description: LNET request buffers exhausted under heavy short-term load
2914 Details : If a large number of client requests are generated on a service
2915 that has previously never seen so many requests it is possible
2916 that the request buffer growth cannot keep up with the spike in
2917 demand. Instead of dropping incoming requests, they are held in
2918 the LND until the RPC service can accept more requests.
2921 Frequency : Sometimes during replay
2923 Description: Assertion failure in ll_local_open after replay.
2924 Details : If replay happened on an open request reply before we were able
2925 to set replay handler, reply will become not swabbed tripping the
2926 assertion in ll_local_open. Now we set the handler right after
2927 recognising of open request
2930 Frequency : very rare
2932 Description: kernel reports "badness in vsnprintf"
2933 Details : Reading from the "recovery_status" /proc file in small chunks
2934 may cause a negative length in lprocfs_obd_rd_recovery_status()
2935 call to vsnprintf() (which is otherwise harmless). Exit early
2936 if there is no more space in the output buffer.
2938 Severity : enhancement
2940 Description: clear OBD RPC statistics by writing to them
2941 Details : It is now possible to clear the OBD RPC statistics by writing
2942 to the "stats" file.
2947 Description: Client mtime is not the same on different clients after utimes
2948 Details : In some cases, the client was using the utimes() syscall on
2949 a file cached on another node. The clients now validate the
2950 ctime from the MDS + OSTs to determine which one is right.
2955 Description: Inability to activate failout mode
2956 Details : lconf script incorrectly assumed that in pythong string's numeric
2957 value is used in comparisons.
2960 Frequency : always with multiple stripes per file
2962 Description: Inefficient object allocation for mutli-stripe files
2963 Details : When selecting which OSTs to stripe files over, for files with
2964 a stripe count that divides evenly into the number of OSTs,
2965 the MDS is always picking the same starting OST for each file.
2966 Return the OST selection heuristic to the original design.
2971 Description: mount failures may take full timeout to return an error
2972 Details : Under some heavy load conditions it is possible that a
2973 failed mount can wait for the full obd_timeout interval,
2974 possibly several minutes, before reporting an error.
2975 Instead return an error as soon as the status is known.
2977 Frequency : quota enabled and large files being deleted
2979 Description: releasing more than 4GB of quota at once hangs OST
2980 Details : If a user deletes more than 4GB of files on a single OST it
2981 will cause the OST to spin in an infinite loop. Release
2982 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2987 Description: statfs data retrieved from /proc may be stale or zero
2988 Details : When reading per-device statfs data from /proc, in the
2989 {kbytes,files}_{total,free,avail} files, it may appear
2990 as zero or be out of date.
2993 Frequency : systems with MD RAID1 external journal devices
2995 Description: lconf's call to blkid is confused by RAID1 journal devices
2996 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2997 of returning all block devices that match the journal UUID.
3000 Frequency : always, for aggregate stripe size over 4GB
3002 Description: assertion fails when trying to use 4GB stripe size
3003 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3004 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3007 Frequency : always on ppc64
3009 Description: the first write on an ext3 filesystem with mballoc got stuck
3010 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3011 perform endianness conversion.
3013 ------------------------------------------------------------------------------
3015 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3017 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3018 INTEROPERATE with older versions automatically. Please read the
3019 user documentation before upgrading any part of a live system.
3020 * WARNING: Lustre networking configuration changes are required with
3021 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3024 * Support for kernels:
3025 2.6.9-22.0.2.EL (RHEL 4)
3026 2.6.5-7.244 (SLES 9)
3027 2.6.12.6 vanilla (kernel.org)
3030 Severity : enhancement
3031 Bugzilla : 7981/8208
3032 Description: Introduced Lustre Networking (LNET)
3033 Details : LNET is new networking infrastructure for Lustre, it includes
3034 a reorganized network configuration mode (see the user
3035 documentation for full details) as well as support for routing
3036 between different network fabrics. Lustre Networking Devices
3037 (LNDs) for the supported network fabrics have also been
3038 created for this new infrastructure.
3040 Severity : enhancement
3041 Description: Introduced Access control lists
3042 Details : clients can set ACLs on files and directories in order to have
3043 more fine-grained permissions than the standard Unix UGO+RWX.
3044 The MDS must be started with the "-o acl" mount option.
3046 Severity : enhancement
3047 Description: Introduced filesystem quotas
3048 Details : Administrators may now establish per-user quotas on the
3051 Severity : enhancement
3053 Description: Configuration change for the XT3
3054 The PTLLND is now used to run Lustre over Portals on the XT3
3055 The configure option(s) --with-cray-portals are no longer used.
3056 Rather --with-portals=<path-to-portals-includes> is used to
3057 enable building on the XT3. In addition to enable XT3 specific
3058 features the option --enable-cray-xt3 must be used.
3063 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3064 Details : A race between allocating a new llcd and re-getting the llcd_lock
3065 allowed another thread to grab newly-allocated llcd.
3067 Severity : enhancement
3069 Description: 2.6 OST async journal commit and locking fix to improve performance
3070 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3071 2.6 kernels are now async as they already were in 2.4 kernels so
3072 that they can commit concurrently with the network bulk transfer.
3073 For block-allocated files the filter allocation semaphore is held
3074 to avoid filesystem fragmentation during allocation. BKL lock
3075 removed for 2.6 xattr operations where it is no longer needed.
3080 Description: lconf incorrectly determined whether two IP networks could talk
3081 Details : In some more complicated routing and multiple-network
3082 configurations, lconf will avoid trying to make a network
3083 connection to a disjoint part of the IP space. It was doing the
3084 math incorrectly for one set of cases.
3089 Description: Fix for potential infinite loop processing records in an llog.
3090 Details : If an llog record is corrupted/zeroed, it is possible to loop
3091 forever in llog_process(). Validate the llog record length
3092 and skip the remainder of the block on error.
3095 Frequency : occasional (liblustre only)
3097 Description: liblustre could not open files whose last component is a symlink
3098 Details : sysio_path_walk() would incorrectly pass the open intent to
3099 intermediate path components.
3102 Frequency : rare (liblustre only with non-standard tuning)
3103 Bugzilla : 7201 (7350)
3104 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3105 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3106 after it has been released, only do this for referenced locks
3108 Severity : enhancement
3110 Description: specifying an (invalid) directory default stripe_size of -1
3111 would reset the directory default striping
3112 Details : stripe_size -1 was used internally to signal directory stripe
3113 removal, now use "all default" to signal dir stripe removal
3114 as a directory striping of "all default" is not useful
3117 Frequency : common for large clusters running liblustre clients
3119 Description: doing an ls when liblustre clients are running is slow
3120 Details : sending a glimpse AST to a liblustre client waits for every AST
3121 to time out, as liblustre clients will not respond. Since they
3122 cannot cache data we refresh the OST lock LVB from disk instead.
3124 Severity : enhancement
3126 Description: doing an ls at the same time as file IO can be slow
3127 Details : enqueue and other "small" requests can be blocked behind many
3128 large IO requests. Create a new OST IO portal for non-IO
3129 requests so they can be processed faster.
3132 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3134 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3135 created file with mode 000
3136 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3137 allow an ftruncate() as a non-root user to a file with mode 000.
3138 Lustre now respects this flag to disable mode checks when
3139 truncating a file owned by the user
3142 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3145 Description: Revoking locks from clients that went dead or catatonic might take
3147 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3148 cancellation of such locks instant on servers without waiting for
3149 any reply from clients. Clients drops these locks when cancel
3150 notification from server is received without replying.
3153 Frequency : liblustre-only, when liblustre client dies or becomes busy
3155 Description: Doing ls on Linux clients can take a long time with active
3157 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3158 granting such locks to it in the first place if possible. Locks
3159 are taken by proxy on the OST during the read or write and
3160 dropped immediately afterward. Add connect flags handling, do
3161 not grant locks to liblustre clients for glimpse ASTs.
3163 Severity : enhancement
3165 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3166 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3167 sequential reads and behaves badly with applications doing random
3168 reads. Improve it by reading ahead at least read region, and
3169 avoiding excessive large RPC for small reads.
3171 Severity : enhancement
3173 Description: Creating more than 1000 files for a single job may cause a load
3174 imbalance on the OSTs if there are also a large number of OSTs.
3175 Details : qos_prep_create() uses an OST index reseed value that is an
3176 even multiple of the number of available OSTs so that if the
3177 reseed happens in the middle of the object allocation it will
3178 still utilize the OSTs as uniformly as possible.
3183 Description: OST or MDS may oops in ping_evictor_main()
3184 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3185 but doesn't restart at beginning of obd_exports_timed list
3188 Severity : enhancement
3190 Description: improve by-nid export eviction on the MDS and OST
3191 Details : allow multiple exports with the same NID to be evicted at one
3192 time without re-searching the exports list.
3195 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3197 Description: MDS may oops in groups_free()
3198 Details : in rare race conditions a newly allocated group_info struct is
3199 freed again, and this can be NULL. The 2.4 compatibility code
3200 for groups_free() checked for a NULL pointer, but 2.6 did not.
3203 Frequency : common for liblustre clients doing little filesystem IO
3204 Bugzilla : 9352, 7313
3205 Description: server may evict liblustre clients accessing contended locks
3206 Details : if a client is granted a lock or receives a completion AST
3207 with a blocking AST already set it would not reply to the AST
3208 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3211 Frequency : lfs setstripe, only systems with more than 160 OSTs
3213 Description: unable to set striping with a starting offset beyond OST 160
3214 Details : llapi_create_file() incorrectly limited the starting stripe
3215 index to the maximum single-file stripe count.
3218 Frequency : LDAP users only
3220 Description: lconf did not handle in-kernel recovery with LDAP properly
3221 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3223 Severity : enhancement
3225 Description: bind OST threads to NUMA nodes to improve performance
3226 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3227 node and do their allocations there to localize memory access
3229 Severity : enhancement
3231 Description: llmount can determine client NID directly from Myrinet (GM)
3232 Details : the client NID code from gmnalnid was moved directly into
3233 llmount, removing the need to use this or specifying the
3234 client NID explicitly when mounting GM clients with zeroconf
3237 Frequency : if client is started with down MDS
3239 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3240 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3241 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3242 ptlrpc_import_delayed_req() because we didn't check if the
3243 request was interrupted, and we also didn't break out of the
3244 event loop if there was no timeout
3249 Description: data loss during non-page-aligned writes to a single file from
3250 both multiple nodes and multiple threads on one node at same time
3251 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3252 inconsistency led to false short-reads, that were cached and later
3253 used by ->prepare_write() to fill in partially written page,
3254 leading to data loss.
3257 Frequency : always, if lconf --abort_recovery used
3259 Description: lconf --abort_recovery fails with 'Operation not supported'
3260 Details : lconf was attempting to abort recovery on the MDT device and not
3263 Severity : enhancement
3265 Description: remove cleanup logs
3266 Details : replace lconf-generated cleanup logs with lustre internal
3267 cleanup routines. Eliminates the need for client-cleanup and
3270 Severity : enhancement
3272 Description: add support for EAs (user and system) on lustre filesystems
3273 Details : it is now possible to store extended attributes in the Lustre
3274 client filesystem, and with the user_xattr mount option it
3275 is possible to allow users to store EAs on their files also
3277 Severity : enhancement
3279 Description: Add possibility (config option) to show minimal available OST free
3281 Details : When compiled with --enable-mindf configure option, statfs(2)
3282 (and so, df) will return least minimal free space available from
3283 all OSTs as amount of free space on FS, instead of summary of
3284 free spaces of all OSTs.
3286 Severity : enhancement
3288 Description: do not expand extent locks acquired on OST-side
3289 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3290 by server: they are not cached anyway.
3293 Frequency : when mmap is used/binaries executed from Lustre
3295 Description: Unmmap pages before throwing them away from read cache.
3296 Details : llap_shrink cache now attempts to unmap pages before discarding
3297 them (if unmapping failed - do not discard). SLES9 kernel has
3298 extra checks that trigger if this unmapping is not done first.
3303 Description: lconf didn't resolve symlinks before checking to see whether a
3304 given mountpoint was already in use
3307 Frequency : when migrating failover services
3308 Bugzilla : 6395, 9514
3309 Description: When migrating a subset of services from a node (e.g. failback
3310 from a failover service node) the remaining services would
3311 time out and evict clients.
3312 Details : lconf --force (implied by --failover) sets the global obd_timeout
3313 to 5 seconds in order to quickly disconnect, but this caused
3314 other RPCs to time out too quickly. Do not change the global
3315 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3317 Severity : enhancement
3318 Frequency : if MDS is started with down OST
3319 Bugzilla : 9439,5706
3320 Description: Allow startup/shutdown of an MDS without depending on the
3321 availability of the OSTs.
3322 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3323 Add appropriate locking and lov-osc refcounts for safe
3324 cleaning. Add osc abort_inflight calls in case the
3325 synchronize never started.
3328 Frequency : occasional (Cray XT3 only)
3330 Description: root not authorized to access files in CRAY_PORTALS environment
3331 Details : The client process capabilities were not honoured on the MDS in
3332 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3333 been accessed by an authorized user then root was able to access
3334 the file on the local client also. The root user capabilities
3335 are now allowed on the MDS, as this environment has secure UID.
3338 Frequency : occasional
3340 Description: ldiskfs "too long searching" message happens too often
3341 Details : A debugging message (otherwise harmless) prints too often on
3342 the OST console. This has been reduced to only happen when
3343 there are fragmentation problems on the filesystem.
3348 Description: Division by zero in statfs when all OSCs are inactive
3349 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3350 lov_statfs divides by value returned by lov_get_stripecnt().
3354 Bugzilla : 9489, 3273
3355 Description: First write from each client to each OST was only 4kB in size,
3356 to initialize client writeback cache, which caused sub-optimal
3357 RPCs and poor layout on disk for the first writen file.
3358 Details : Clients now request an initial cache grant at (re)connect time
3359 and so that they can start streaming writes to the cache right
3360 away and always do full-sized RPCs if there is enough data.
3361 If the OST is rebooted the client also re-establishes its grant
3362 so that client cached writes will be honoured under the grant.
3367 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3368 Details : Now I/O RPCs go to different portal number and (presumably) fast
3369 lock requests (and glimses) and other RPCs get their own service
3370 threads pool that should be able to service those RPCs
3373 Severity : enhancement
3375 Description: Ability to exchange lustre version between client and servers and
3376 issue warnings at client side if client is too old. Also for
3377 liblustre clients there is ability to refuse connection of too old
3379 Details : New 'version' field is added to connect data structure that is
3380 filled with version info. That info is later checked by server and
3384 Frequency : rare, liblustre only.
3385 Bugzilla : 9296, 9581
3386 Description: Two simultaneous writes from liblustre at offset within same page
3387 might proceed at the same time overwriting eachother with stale
3389 Details : I/O lock withing llu_file_prwv was released too early, before data
3390 actually was hitting the wire. Extended lock-holding time until
3391 server acknowledges receiving data.
3394 Frequency : extremely rare. Never observed in practice.
3396 Description: avoid generating lustre_handle cookie of 0.
3397 Details : class_handle_hash() generates handle cookies by incrementing
3398 global counter, and can hit 0 occasionaly (this is unlikely, but
3399 not impossible, because initial value of cookie counter is
3400 selected randonly). Value of 0 is used as a sentinel meaning
3401 "unassigned handle" --- avoid it. Also coalesce two critical
3402 sections in this function into one.
3404 Severity : enhancement
3406 Description: allow liblustre clients to delegate truncate locking to OST
3407 Details : To avoid overhead of locking, liblustre client instructs OST to
3408 take extent lock in ost_punch() on client's behalf. New connection
3409 flag is added to handle backward compatibility.
3411 Severity : enhancement
3412 Bugzilla : 4928, 7341, 9758
3413 Description: allow number of OST service threads to be specified
3414 Details : a module parameter allows the number of OST service threads
3415 to be specified via "options ost ost_num_threads={N}" in the
3416 OSS's /etc/modules.conf or /etc/modprobe.conf.
3420 Bugzilla : 6146, 9635, 9895
3421 Description: servers crash with bad pointer in target_handle_connect()
3422 Details : In rare cases when a client is reconnecting it was possible that
3423 the connection request was the last reference for that export.
3424 We would temporarily drop the export reference and get a new
3425 one, but this may have been the last reference and the export
3426 was just destroyed. Get new reference before dropping old one.
3428 Severity : enhancement
3429 Frequency : if client is started with failover MDS
3431 Description: Allow multiple MDS hostnames in the mount command
3432 Details : Try to read the configuration from all specified MDS
3433 hostnames during a client mount in case the "primary"
3436 Severity : enhancement
3438 Description: Stop sending data to evicted clients as soon as possible.
3439 Details : Check if the client we are about to send or are sending data to
3440 was evicted already. (Check is done every second of waiting,
3441 for which l_wait_event interface was extended to allow checking
3442 of exit condition at specified intervals).
3445 Frequency : rare, normally only when NFS exporting is done from client
3447 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3449 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3452 Frequency : always, if srand() is called before liblustre initialization
3454 Description: Liblustre uses system PRNG disturbing its usage by user application
3455 Details : Introduce internal to lustre fast and high-quality PRNG for
3456 lustre usage and make liblustre and some other places in generic
3457 lustre code to use it.
3459 Severity : enhancement
3460 Bugzilla : 9477, 9557, 9870
3461 Description: Verify that the MDS configuration logs are updated when xml is
3462 Details : Check if the .xml configuration logs are newer than the config
3463 logs stored on the MDS and report an error if this is the case.
3464 Request --write-conf, or allow starting with --old_conf.
3466 Severity : enhancement
3468 Description: Handle symlinks in the path when checking if Lustre is mounted.
3469 Details : Resolve intermediate symlinks when checking if a client has
3470 mounted a filesystem to avoid duplicate client mounts.
3475 Description: lconf can hit an error exception but still return success.
3476 Details : The lconf command catches the Command error exception at the top
3477 level script context and will exit with the associated exit
3478 status, but doesn't ensure that this exit status is non-zero.
3483 Description: failure of ptlrpc thread startup can cause oops
3484 Details : Starting a ptlrpc service thread can fail if there are a large
3485 number of threads or the server memory is very fragmented.
3486 Handle this without oopsing.
3489 Frequency : always, only if liblustre and non-default acceptor port was used
3491 Description: liblustre cannot connect to servers with non-default acceptor port
3492 Details : tcpnal_set_default_params() was not called and was therefore
3493 ignoring the environment varaible TCPNAL_PORT, as well as other
3494 TCPNAL_ environment variables
3499 Description: two objects could be created on the same OST for a single file
3500 Details : If an OST is down, in some cases it was possible to create two
3501 objects on a single OST for a single file. No problems other
3502 than potential performance impact and spurious error messages.
3506 Bugzilla : 5681, 9562
3507 Description: Client may oops in ll_unhash_aliases
3508 Details : Client dcache may become inconsistent in race condition.
3509 In some cases "getcwd" can fail if the current directory is
3515 Description: Inode refcounting problems in NFS export code
3516 Details : link_raw functions used to call d_instantiate without obtaining
3517 extra inode reference first.
3521 Bugzilla : 9942, 9903
3522 Description: Referencing freed requests leading to crash, memleaks with NFS.
3523 Details : We used to require that call to ll_revalidate_it was always
3524 followed by ll_lookup_it. Also with revalidate_special() it is
3525 possible to call ll_revalidate_it() twice for the same dentry
3526 even if first occurence returned success. This fix changes semantic
3527 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3528 reference on a request referred from the intent.
3529 ll_intent_release() then releases such a request.
3532 Frequency : rare, normally benchmark loads only
3534 Description: unlinked inodes were kept in memory on the client
3535 Details : If a client is repeatedly creating and unlinking files it
3536 can accumulate a lot of stale inodes in the inode slab cache.
3537 If there is no other client load running this can cause the
3538 client node to run out of memory. Instead flush old inodes
3539 from client cache that have the same inode number as a new inode.
3542 Frequency : SLES9 2.6.5 kernel and long filenames only
3543 Bugzilla : 9969, 10379
3544 Description: utime reports stale NFS file handle
3545 Details : SLES9 uses out-of-dentry names in some cases, which confused
3546 the lustre dentry revalidation. Change it to always use the
3550 Frequency : rare, unless heavy write-truncate concurrency is continuous
3551 Bugzilla : 4180, 6984, 7171, 9963, 9331
3552 Description: OST becomes very slow and/or deadlocked during object unlink
3553 Details : filter_destroy() was holding onto the parent directory lock
3554 while truncating+unlinking objects. For very large objects this
3555 may block other threads for a long time and slow overall OST
3556 responsiveness. It may also be possible to get a lock ordering
3557 deadlock in this case, or run out of journal credits because of
3558 the combined truncate+unlink. Solution is to do object truncate
3559 first in one transaction without parent lock, and then do the
3560 final unlink in a new transaction with the parent lock. This
3561 reduces the lock hold time dramatically.
3564 Frequency : rare, 2.4 kernels only
3566 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3567 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3568 before unmounting their filesystem in order to avoid deadlock.
3569 The kernel_locked() function in 2.4 kernels only checks whether
3570 the kernel lock is held, not whether it is this process that is
3571 holding it as 2.6 kernels do.
3576 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3577 Details : The client ptlrpc code may be trying to reconnect to a down
3578 server before a previous connection attempt has timed out.
3579 Increase the reconnect interval to be longer than the connection
3580 timeout interval to avoid sending duplicate connections to
3584 Frequency : echo_client brw_test command
3586 Description: fix echo_client to work with OST preallocated code
3587 Details : OST preallocation code (5137) didn't take echo_client IO path
3588 into account: echo_client calls filter methods outside of any
3589 OST thread and, hence, there is no per-thread preallocated
3590 pages and buffers to use. Solution: hijack pga pages for IO. As
3591 a byproduct, this avoids unnecessary data copying.
3595 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3596 Description: Client can oops in mdc_commit_close() after open replay
3597 Details : It was possible for the MDS to return an open request with no
3598 transaction number in mds_finish_transno() if the client was
3599 evicted, but without actually returning an error. Clients
3600 would later try to replay that open and may trip an assertion
3601 Simplify the client close codepath, and always return an error
3602 from the MDS in case the open is not successful.
3605 Frequency : rare, 2.6 OSTs only
3607 Description: OST may deadlock under high load on fragmented files
3608 Details : If there was a heavy load and highly-fragmented OST filesystems
3609 it was possible to have all the OST threads deadlock waiting on
3610 allocation of biovecs, because the biovecs were not released
3611 until the entire RPC IO was completed. Instead, release biovecs
3612 as soon as they are complete to ensure forward IO progress.
3614 Severity : enhancement
3616 Description: Support for specifying external journal device at mount
3617 Details : If an OST or MDS device is formatted with an external journal
3618 device, this device major/minor is stored in the ext3 superblock
3619 and may not be valid for failover. Allow detecting and
3620 specifying the external journal at mount time.
3625 Description: Mounting an MDS with pending unlinked files may cause oops
3626 Details : target_finish_recovery() calls mds_postrecov() which returned
3627 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3628 considers this an error and immediately begins cleaning up the
3629 lov, just after starting the mds_lov process
3631 Severity : enhancement
3633 Description: Implement 'lfs df' to report actual free space on per-OST basis
3634 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3635 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3636 usage of objects; '-h' to report in human readable format.
3638 ------------------------------------------------------------------------------
3640 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3647 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3648 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3649 at startup during orphan unlinking. Instead, skip buddy update
3650 before setup, as it will be regenerated after recovery is complete.
3655 Description: If an OST is inactive, its locks might reference stale inodes.
3656 Details : lov_change_cbdata() must iterate over all namespaces, even if
3657 they are inactive to clear inode references from the lock.
3659 Severity : enhancement
3660 Frequency : occasional, if non-standard max_dirty_mb used
3662 Description: Client will block write RPCs if not enough grant
3663 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3664 then the client will block writes while waiting for another RPC
3665 to complete instead of consuming its dirty limit. With change
3666 we get improved performance when max_dirty_mb is small.
3668 Severity : enhancement
3669 Bugzilla : 3389, 6253
3670 Description: Add support for supplementary groups on the MDS.
3671 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3672 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3673 lookups for user supplementary groups into a cache.
3677 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3678 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3679 for liblustre O_EXCL is enforced by MDS.
3683 Description: readdir never returns NULL in liblustre.
3684 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3685 dentry in struct dirent.
3689 Frequency : liblustre only, depends on application IO pattern
3690 Description: liblustre clients evicted if not contacting servers
3691 Details : Don't put liblustre clients into the ping_evictor list, so
3692 they will not be evicted by the pinger ever.
3694 Severity : enhancement
3696 Description: Add ability to evict clients by NID from MDS.
3697 Details : By echoing "nid:$NID" string into
3698 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3699 $NID would be instantly evicted from this MDS and from all active
3700 OSTs connected to it.
3704 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3705 Details : lookup_it_finish() used to query file size from OSTs that was not
3710 Description: service threads change working directory to that of init
3711 Details : Starting lustre service threads may pin the working directory
3712 of the parent thread, making that filesystem busy. Threads
3713 now change to the working directory of init to avoid this.
3717 Frequency : during shutdown only
3718 Description: shutdown with a failed MDS or OST can cause unmount to hang
3719 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3724 Frequency : 2.6 only
3725 Description: chmod/chown may include an extra supplementary group
3726 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3727 supplementary group and if none is specified this is used.
3730 Bugzilla : 5479 (6816)
3732 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3733 Details : If lookup is in progress on a file that is unlinked we might try
3734 to revalidate the inode and fail in revalidate after lookup is
3735 complete and ll_file_open() enqueues the open again but
3736 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3739 Frequency : always, if lconf --abort_recovery used
3741 Description: lconf --abort_recovery fails with 'Operation not supported'
3742 Details : lconf was attempting to abort recovery on the MDT device and not
3745 ------------------------------------------------------------------------------
3747 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3752 Frequency : rare (only unsupported configurations with a node running as an
3754 Bugzilla : 6514, 5137
3755 Description: Mounting a Lustre file system on a node running as an OST could
3757 Details : OSTs now preallocates memory needed to write out data at
3758 startup, instead of when needed, to avoid having to
3759 allocate memory in possibly low memory situations.
3760 Specifically, if the file system is mounted on on OST,
3761 memory pressure could force it to try to write out data,
3762 which it needed to allocate memory to do. Due to the low
3763 memory, it would be unable to do so and the node would
3764 become unresponsive.
3766 Severity : enhancement
3768 Description: Addition of lconf --service command line option
3769 Details : lconf now accepts a '--service <arg>' option, which is
3770 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3772 Severity : enhancement
3774 Description: Failover mode is now the default for OSTs.
3775 Details : By default, OSTs will now run in failover mode. To return to
3776 the old behaviour, add '--failout' to the lmc line for OSTs.
3778 Severity : enhancement
3780 Description: Health checks are now provided for MDS and OSTs
3781 Details : Additional detailed health check information on MSD and OSTs
3782 is now provided through the procfs health_check value.
3785 Frequency : occasional, depends on IO load
3787 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3788 after numerous create/delete cycles
3789 Details : The ext3 inode allocation policy would not allocate new inodes
3790 very well on the OSTs because there are no new directories
3791 being created. Instead we look for groups with free space if
3792 the parent directories are nearly full.
3797 Description: Network or server problems during mount may cause partially
3798 mounted clients instead of returning an error.
3799 Details : The config llog parsing code may overwrite the error return
3800 code during mount error handling, returning success instead
3806 Description: MDS can fail to allocate large reply buffers
3807 Details : After long uptimes the MDS can fail to allocate large reply
3808 buffers (e.g. zconf client mount config records) due to memory
3809 fragmentation or consumption by the buffer cache. Preallocate
3810 some large reply buffers so that these replies can be sent even
3811 under memory pressure.
3815 Frequency : rare (liblustre)
3816 Description: fsx running with liblustre complained that using truncate() to
3817 extend the file doesn't work. This patch corrects that issue.
3818 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3819 ATTR_SIZE and lsm use in llu_setattr_raw.
3823 Frequency : rare, only 2.6 kernels
3824 Description: Unusual file access patterns on the MDS may result in inode
3825 data being lost in very rare circumstances.
3826 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3827 code in the 2.6 kernel caused the inode and/or EA data not to
3828 be read from disk, causing single-file corruption.
3832 Frequency : rare, only 2.6 filesystems using extents
3833 Description: Heavy concurrent write and delete load may cause data corruption.
3834 Details : It was possible under high-load situations to have an extent
3835 metadata block in the block device cache from a just-unlinked
3836 file overwrite a newly-allocated data block. We now unmap any
3837 metadata buffers that alias just-allocated data blocks.
3841 Frequency : filesystems with default stripe_count larger than 77
3842 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3843 Details : lconf specifies an inode size of 4096 bytes when the default
3844 stripe_count is larger than 77. This conflicts with the default
3845 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3846 this case to avoid pinning too much memory for large EAs.
3848 ------------------------------------------------------------------------------
3850 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3855 Frequency : rare (extremely heavy IO load with hundreds of clients)
3857 Description: Client is evicted, gets IO error writing to file
3858 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3859 caused clients to be evicted for AST timeouts. The fixes in
3860 bug 5192 mean we no longer need to have such short AST timeouts
3861 so ldlm_timeout has been increased.
3864 Frequency : occasional during --force or --failover shutdown under load
3865 Bugzilla : 5949, 4834
3866 Description: Server oops/LBUG if stopped with --force or --failover under load
3867 Details : a collection of import/export refcount and cleanup ordering
3868 issues fixed for safer force cleanup
3871 Frequency : only filesystems larger than 120 OSTs
3872 Bugzilla : 5990, 6223
3873 Description: lfs getstripe would oops on a very large filesystem
3874 Details : lov_getconfig used kfree on vmalloc'd memory
3877 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3878 Bugzilla : 6242, 6243
3879 Description: reading from files that had been truncated to a non-zero size
3880 but never opened returned no data
3881 Details : ll_file_read() reads zeros from no-object files to EOF
3886 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3887 Details : There's little harm in aborting MDS/OSS recovery and letting it
3888 try again, so I removed the LASSERT and return an error instead.
3890 Severity : enhancement
3892 Description: New debugging infrastructure for tracking down data corruption
3893 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3894 (b) cover more of the client-side code path, and (c) try to narrow
3895 down where problems occurred
3899 Bugzilla : 3819, 4364, 4397, 6313
3900 Description: Racing close and eviction MDS could cause assertion in mds_close
3901 Details : It was possible to get multiple mfd references during close and
3902 client eviction, leading to one thread referencing a freed mfd.
3904 Severity: : enhancement
3905 Bugzilla : 3262, 6359
3906 Description: Attempts to reconnect to servers are now more aggressive.
3907 Details : This builds on the enhanced upcall-less recovery that was added
3908 in 1.4.2. When trying to reconnect to servers, clients will
3909 now try each server in the failover group every 10 seconds. By
3910 default, clients would previously try one server every 25 seconds.
3915 Description: After recovery, certain operations trigger a failed
3916 assertion on a client.
3917 Details : Failing over an mds, using lconf -d --failover, while a
3918 client was doing a readdir() call would cause the client to
3919 LBUG after recovery completed and the readdir() was resent.
3921 Severity : enhancement
3923 Description: Default groups are now added by lconf
3924 Details : You can now run lconf --group <servicename> without having to
3925 manually add groups with lmc.
3928 Frequency : occasional
3930 Description: Nodes with an elan id of 0 trigger a failed assertion
3933 Frequency : always when accessing e.g. tty/console device nodes
3935 Description: tty and some other devices nodes cannot be used on lustre
3936 Details : file's private_data field is used by device data and lustre
3937 values in there got lost. New field was added to struct file to
3938 store fs-specific private data.
3941 Frequency : when exporting Lustre via NFS
3943 Description: NFSD failed occasionally when looking up a path component
3944 Details : NFSD is looking up ".." which was broken in ext3 directories
3945 that had grown large enough to become hashed.
3948 Frequency : Clusters with multiple interfaces not on the same subnet
3950 Description: Nodes will repeatedly try to reconnect to an interface which it
3951 cannot reach and report an error to the log.
3952 Details : Extra peer list entries will be created by lconf with some peers
3953 unreachable. lconf now validates the peer before adding it.
3956 Frequency : Only if a default stripe is set on the filesystem root.
3958 Description: Setting a default stripe on the filesystem root prevented the
3959 filesystem from being remounted.
3960 Details : The client was sending extra request flags in the root getattr
3961 request and did not allocate a reply buffer for the dir EA.
3964 Frequency : occasional, higher if lots of files are accessed by one client
3965 Bugzilla : 6159, 6097
3966 Description: Client trips assertion regarding lsm mismatch/magic
3967 Details : While revalidating inodes the VFS looks up inodes with ifind()
3968 and in rare cases can find an inode that is being freed.
3969 The ll_test_inode() code will free the lsm during ifind()
3970 when it finds an existing inode and then the VFS later attaches
3971 this free lsm to a new inode.
3975 Bugzilla : 6422, 7030
3976 Description: MDS deadlock between mkdir and client eviction
3977 Details : Creating a new file via mkdir or mknod (starting a transaction
3978 and getting the ns lock) can deadlock with client eviction
3979 (gets ns lock and trying to finish a synchronous transaction).
3982 Frequency : occasional
3983 Description: While starting a server, the fsfilt_ext3 module could not be
3985 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3986 kernels. Previously, lconf would still use the ext3 name
3987 when trying to load modules. Now, it will correctly use
3988 ext3 on 2.4 and ldiskfs on 2.6.
3990 Severity : enhancement
3991 Description: The default stripe count has been changed to 1
3992 Details : The interpretation of the default stripe count (0, to lfs
3993 or lmc) has been changed to mean striping across a single
3994 OST, rather than all available. For general usage we have
3995 found a stripe count of 1 or 2 works best.
3997 Severity : enhancement
3998 Description: Add support for compiling against Cray portals.
3999 Details : Conditional compiling for some areas that are different
4003 Frequency : occasional
4004 Bugzilla : 6409, 6834
4005 Description: Creating files with an explicit stripe count may lead to
4006 a failed assertion on the MDS
4007 Details : If some OSTs are full or unavailable, creating files may
4008 trigger a failed assertion on the MDS. Now, Lustre will
4009 try to use other servers or return an error to the
4013 Frequency : occasional
4015 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4016 caused lock timeout during readahead (since 1.4.2).
4017 Details : Processes doing readahead might match a lock that hasn't been
4018 granted yet if there are overlapping and conflicting lock
4019 requests. The readahead process waits on ungranted lock
4020 (original lock is CBPENDING), while OST waits for that process
4021 to cancel CBPENDING read lock and eventually evicts client.
4023 Severity : enhancement
4025 Description: Initial enabling of flock support for clients
4026 Details : Implements fcntl advisory locking and file status functions.
4027 This feature is provided as an optional mount flag (default
4028 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4029 locking are implemented yet, and those that are are not guaranteed
4030 to be completely correct in production environments.
4031 mount -t lustre -o [flock|noflock] ...
4034 Frequency : occasional
4036 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4037 assertion in the ext3 JBD (journaling) layer.
4038 Details : The b_committed_data struct is protected by the big kernel lock
4039 in 2.4 kernels, serializing journal_commit_transaction() and
4040 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4041 access to this struct. In 2.6 kernels there is finer grained
4042 locking to improve SMP performance of the JBD layer.
4046 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4051 Description: merge and cleanup kernel patches.
4054 -----------------------------------------------------------------------------
4056 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4058 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4059 versions of Lustre. Please update all servers and clients to
4060 version 1.4.2 or later at the same time. You must also run
4061 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4062 to update the configuration logs.
4064 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4065 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4066 - swab llog records of type '0' so we get proper header size/idx (5861)
4067 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4068 - fix rename of one directory over another leaking an inode (5953)
4069 - avoid SetPageDirty on 2.6 (5981)
4070 - don't re-add just-being-destroyed locks to the waiting list (5653)
4071 - when creating new directories, inherit the parent's custom
4072 striping settings if present parent (3048)
4073 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4074 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4075 - don't hold client locks on temporary worklist from l_lru (5666)
4076 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4077 - automatically evict dead clients (5921)
4078 - Update file size properly in create+truncate+fstat case (6196)
4079 - Do not unhash mountpoint dentries, do not allow removal of
4081 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4082 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4083 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4084 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4085 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4086 - Correctly handle reads of files with no objects (6243)
4087 - lctl recover will also mark a device active if deactivate used (5933)
4089 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4090 - allow --write-conf on an MDS with different nettype than client (5619)
4091 - don't write config llogs to MDS for mounts not from that MDS (5617)
4092 - lconf should create multiple TCP connections from a client (5201)
4093 - init scripts are now turned off by default; run chkconfig --on
4094 lustre and chkconfig --on lustrefs to use them
4095 - upcalls are no longer needed for clients to recover to failover
4097 - add --abort-recovery option to lconf to abort recovery on device
4099 - add support for an arbitrary number of OSTs (3026)
4100 - Quota support protocol changes.
4101 - forward compatibility changes to wire structs (6007)
4102 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4103 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4104 - improved "lustre status" script
4105 - initialize blocksize for non-regular files (6062)
4106 - added --disable-server and --disable-client configure options (5782)
4107 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4108 - Vanilla 2.4.29 support
4109 - increase maximum number of obd devices to 520 (6242)
4110 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4111 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4112 SLES 9 2.6.5 kernels
4113 - limit stripes per file to 160 (the maximum EA size) (6093)
4115 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4118 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4119 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4120 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4121 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4122 - small ext3 extents cleanups and fixes (5733)
4123 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4124 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4125 - avoid needless client->OST connect, fix handle mismatch (5317)
4126 - fix DLM error path that led to out-of-sync client, long delays (5779)
4127 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4128 - fix several locking issues related to i_size (5492,5624,5654,5672)
4129 - don't move pending lock onto export if it is already evicted (5683)
4130 - fix kernel oops when creating .foo in unlinked directory (5548)
4131 - fix deadlock in obdfilter statistics vs. object create (5811)
4132 - use time_{before,after} to avoid timer jiffies wrap (5882)
4133 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4134 - Do not leak request if server was not able to process it (5154)
4135 - If mds_open unable to find parent dir, make that negative lookup(5154)
4136 - don't create new directories with extent-mapping (5909, 5936)
4138 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4139 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4140 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4141 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4142 - support for RHEL4 kernel on i686 (5773)
4143 - provide error messages when incompatible logs are encountered (5898)
4145 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4146 * version 1.4.0.10 (1.4.1 release candidate 1)
4148 - don't keep a lock reference when lock is not granted (4238)
4149 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4150 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4151 - drop import inflight refcount on signal_completed_replay error (5255)
4152 - unlock page after async write error during send (3677)
4153 - handle missing objects in filter_preprw_read properly (5265)
4154 - no transno return for symlink open, don't save no-trasno open (3440)
4155 - don't try to complete elan receive that already failed (4012)
4156 - free RPC server reply state on error (5406)
4157 - clean up thread from ptlrpc_start_thread() on error (5160)
4158 - readahead could read extra page into cache that wasn't ejected (5388)
4159 - prevent races in class_attach/setup/cleanup/detach (5260)
4160 - don't dereference de->d_inode after l_dput of de (5458)
4161 - use "int" for stripe value returned from lock_to_stripe (5544)
4162 - mballoc allocation and error-checking fixes in 2.6 (5504)
4163 - block device patches to fix I/O request sizes in 2.6 (5482)
4164 - look up hostnames for IB nals (5602)
4165 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4166 - don't start multiple acceptors for the same port (5277)
4167 - fix incorrect LASSERT in mds_getattr_name (5635)
4168 - export a proc file for general "ping" checking (5628)
4169 - fix "lfs check" to not block when the MDS is down (5628)
4171 - service request history (4965)
4172 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4173 - create an "evict_client" /proc entry on OSTs, like the MDS has
4174 - fix mount usage message, return errors per mount(8) (5168)
4175 - change grep [] to grep "[]" in tests so they work in more UMLs
4176 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4177 - remove ancient LOV_MAGIC_V0 EA support (5047)
4178 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4179 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4180 - allow readahead allocations to fail when low on memory (5383)
4181 - mmap locking landed again, after considerable improvement (2828)
4182 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4184 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4187 - send OST transaction number in read/write reply to free req (4966)
4188 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4189 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4190 - the watchdog thread now runs as interruptible (5246)
4191 - flock/lockf fixes (but it's still disabled, pending 5135)
4192 - don't use EXT3 constants in llite code (5094)
4193 - memory shortage at startup could cause assertion (5176)
4195 - reorganization of lov code
4196 - single portals codebase
4198 - add extents/mballoc support (5025)
4199 - direct I/O reads in the obdfilter (4048)
4200 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4202 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4205 - send OST transaction number in read/write reply to free req (4966)
4206 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4207 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4208 - the watchdog thread now runs as interruptible (5246)
4209 - handle missing objects in filter_preprw_read properly (5265)
4210 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4211 - drop import inflight refcount on signal_completed_replay error (5255)
4212 - unlock page after async write error during send (3677)
4213 - return original error code on reconstructed replies (3761)
4214 - no transno return for symlink open, don't save no-trasno open (3440)
4216 - add pid to ldlm debugging output (4922)
4217 - bump the watchdog timeouts -- we can't handle 30sec yet
4218 - extra debugging for orphan dentry/inode bug (5259)
4220 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4223 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4224 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4225 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4226 - fix NULL dereference in /proc/sys/portals/routes (4827)
4227 - allow failed mdc_close() operations to be interrupted (4561)
4228 - stop precreate on OST before MDS would time out on it (4778)
4229 - don't send partial-page writes before EOF from client (4410)
4230 - discard client grant for sub-page writes on large-page clients (4520)
4231 - don't free dentries not owned by NFS code, check generation (4806)
4232 - fix lsm leak if mds_create_objects() fails (4801)
4233 - limit debug_daemon file size, always print CERROR messages (4789)
4234 - use transno after validating reply (3892)
4235 - process timed out requests if import state changes (3754)
4236 - update mtime on OST during writes, return in glimpse (4829)
4237 - add mkfsoptions to LDAP (4679)
4238 - use ->max_readahead method instead of zapping global ra (5039)
4239 - don't interrupt __l_wait_event() during strace
4241 - add software watchdogs to catch hung threads quickly (4941)
4242 - make lustrefs init script start after nfs is mounted
4243 - fix CWARN/ERROR duplication (4930)
4244 - return async write errors to application if possible (2248)
4245 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4246 - print NAL number in %x format (4645)
4247 - update barely-supported suse-2.4.21-171 series (4842)
4248 - support for sles 9 %post scripts
4249 - support for building 2.6 kernel-source packages
4250 - support for sles km_* packages
4252 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4255 - ignore -ENOENT errors in osc_destroy (3639)
4256 - notify osc create thread that OSC is being cleaned up (4600)
4257 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4258 - reconstruct ost_handle() like mds_handle() (4657)
4259 - create a new thread to do import eviction to avoid deadlock (3969)
4260 - let lconf resolve symlinked-to devices (4629)
4261 - don't unlink "objects" from directory with default EA (4554)
4262 - hold socknal file ref over connect in case target is down (4394)
4263 - allow more than 32000 subdirectories in a single directory (3244)
4264 - fix blocks count for O_DIRECT writes (3751)
4265 - OST returns ENOSPC from object create when no space left (4539)
4266 - don't send truncate RPC if file size isn't changing (4410)
4267 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4268 - bind to privileged port in socknal and tcpnal (3689)
4270 - rate limit CERROR/CWARN console message to avoid overload (4519)
4271 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4272 - basic mmap support (3918)
4273 - kernel patch series update from b1_4 (4711)
4275 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4278 - avoid crash during MDS cleanup with OST shut down (2775)
4279 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4280 - don't use bad inodes on the MDS (3744)
4281 - dynamic object preallocation to improve recovery speed (4236)
4282 - don't hold spinlock over lock dumping or change debug flags (4401)
4283 - don't zero obd_dev when it is force cleaned (3651)
4284 - print grants to console if they go negative (4431)
4285 - "lctl deactivate" will stop automatic recovery attempts (3406)
4286 - look for existing locks in ldlm_handle_enqueue() (3764)
4287 - don't resolve lock handle twice in recovery avoiding race (4401)
4288 - revalidate should check working dir is a directory (4134)
4290 - don't always mark "slow" obdfilter messages as errors (4418)
4292 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4295 - don't close LustreDB during write_conf until it is done (3860)
4296 - fix typo in lconf for_each_profile (3821)
4297 - allow dumping logs from multiple threads at one time (3820)
4298 - don't allow multiple threads in OSC recovery (3812)
4299 - fix debug_size parameters (3864)
4300 - fix mds_postrecov to initialize import for llog ctxt (3121)
4301 - replace config semaphore with spinlock (3306)
4302 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4303 - don't allow enqueue to complete on a destroyed export (3822)
4304 - down write_lock before checking llog header bitmap (3825)
4305 - recover from lock replay timeout (3764)
4306 - up llog sem before sending rpc (3652)
4307 - reduce ns lock hold times when setting kms (3267)
4308 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4309 - fix NULL deref and obd_dev leak on setup error (3312)
4310 - replace some LBUG about llog ops with error handling (3841)
4311 - don't match INVALID dentries from d_lookup and spin (3784)
4312 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4313 - fix invalid assertion in ptlrpc_set_wait (3880)
4315 - add libwrap support for the TCP acceptor (3996)
4316 - add /proc/sys/portals/routes for non-root route listing (3994)
4317 - allow setting MDS UUID in .xml (2580)
4318 - print the stack of a process that LBUGs (4228)
4320 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4323 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4324 - make sure to unset replay flag from failed open requests (3440)
4325 - if default stripe count is 0, use OST count for inode size (3636)
4326 - update parent mtime/ctime on client for create/unlink (2611)
4327 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4328 - free recovery state on server during a forced cleanup (3571)
4329 - unregister_reply for resent reqs (3063)
4330 - loop back devices mounting and status check on 2.6 (3563)
4331 - fix resource-creation race that can provoke i_size == 0 (3513)
4332 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4333 - more debugging for page-accounting assertion (3746)
4334 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4335 - avoid deadlock after precreation failure (3758)
4336 - fix race and lock order deadlock in orphan handling (3450, 3750)
4337 - add validity checks when grabbing inodes from l_ast_data (3599)
4339 - add /proc/.../recovery_status to obdfilter (3428)
4340 - lightweight CDEBUG infrastructure, debug daemon (3668)
4341 - change default OSC RPC parameters to be better on small clusters
4342 - turn off OST read cache for files smaller than 32MB
4343 - install man pages and include them in rpms (3100)
4344 - add new init script for (un)mounting lustre filesystems (2593)
4345 - run chkconfig in %post for init scripts (3701)
4346 - drop scimac NAL (unmaintained)
4348 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4351 - clean kiobufs before and after use (3485)
4352 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4353 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4354 - change init script to fail more clearly if not run as root (1528)
4355 - allow clients to reconnect during replay (1742)
4356 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4357 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4358 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4359 - fix cond_resched (3554)
4360 - only evict unfinished clients after recovery (3515)
4361 - allow bulk resend, prevent data loss (3570)
4362 - dynamic ptlrpc request buffer allocation (2102)
4363 - don't allow unlinking open directory if it isn't empty (2904)
4364 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4365 - remove extraneous obd dereference causing LASSERT failure (3334)
4366 - don't use get_cycles() when creating temp. files on the mds (3156)
4367 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4368 - handle EEXIST for set-stripe, set proper directory name (3336)
4370 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4371 - fix ksocknal_fmb_callback() error messages (2918)
4373 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4376 - don't copy lvb into (possibly NULL) reply on error (2983)
4377 - don't deref dentry after dput, don't free lvb on error (2922)
4378 - use the kms to determine writeback rpc length (2947)
4379 - increment oti_logcookies when osc is inactive (2948)
4380 - update client's i_blocks count via lvb messages (2543)
4381 - handle intent open/close of special files properly (1557)
4382 - mount MDS with errors=remount-ro, like obdfilter (2009)
4383 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4384 - don't use cancelling-locks' kms values (2947)
4385 - use highest lock extent for kms, not last one (2925)
4386 - don't dereference ERR_PTR() dentry in error handling path (3107)
4387 - fix thread race in portals_debug_dumplog() (3122)
4388 - create lprocfs device entries at setup instead of at attach (1519)
4389 - common AST error handler, don't evict client on completion race (3145)
4390 - zero nameidata in detach_mnt in 2.6 (3118)
4391 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4392 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4393 - initialize RPC timeout timer earlier for 2.6 (3219)
4394 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4395 - print nal/nid for unknown nid (3258)
4396 - additional checks for oscc recovery before doing precreate (3284)
4397 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4398 - don't crash in mdc_close for bad permissions on open (3285)
4399 - zero i_rdev for non-device files (3147)
4400 - clear page->private before handing to FS, better assertion (3119)
4401 - tune the read pipeline (3236)
4402 - fix incorrect decref of invalidated dentry (2350)
4403 - provide read-ahead stats and refine rpc in flight stats (3328)
4404 - don't hold journal transaction open across create RPC (3313)
4405 - update atime on MDS at close time (3265)
4406 - close LDAP connection when recovering to avoid server load (3315)
4407 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4408 - don't leak open file on MDS after open resend (3325)
4409 - serialize filter_precreate and filter_destroy_precreated (3329)
4410 - loop device shouldn't call sync_dev() for nul device (3092)
4411 - clear page cache after eviction (2766)
4412 - resynchronize MDS->OST in background (2824)
4413 - refuse to mount the same filesystem twice on same mountpoint (3394)
4414 - allow llmount to create routes for mounting behind routers (3320)
4415 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4416 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4417 - fix rare problem with rename on htree directories (3417)
4419 - allow default OST striping configuration per directory (1414)
4420 - fix compilation for qswnal for 2.6 kernels (3125)
4421 - increase maximum number of MDS request buffers for large systems
4422 - change liblustreapi to be useful for external progs like lfsck (3098)
4423 - increase local configuration timeout for slow disks (3353)
4424 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4426 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4429 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4430 - don't overwrite extent policy data in reply if lock was blocked (2901)
4431 - drop filter export grants atomically with removal from device (2663)
4432 - del obd_self_export from work_list in class_disconnect_exports (2908)
4433 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4434 - swab reply message in mdc_close, other PPC fixes (2464)
4435 - fix destroying of named logs (2325)
4436 - overwrite old logs when running lconf --write_conf (2264)
4437 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4438 - fix race in target_handle_connect (2898)
4439 - mds_reint_create() should take same inode create lock (2926)
4440 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4441 - don't close files for self_export to avoid uninitialized obd (2936)
4442 - allow MDS with the same name as client node (2939)
4443 - hold dentry reference for closed log files for unlink (2325)
4444 - reserve space for all logs during transactions (2059)
4445 - don't evict page beyond end of stripe extent (2925)
4446 - don't oops on a deleted current working directory (2399)
4447 - handle hard links to targets without a parent properly (2517)
4448 - don't dereference NULL lock when racing during eviction (2867)
4449 - don't grow lock extents when lots of conflicting locks (2919)
4451 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4454 - account for cache space usage on clients to avoid data loss (974)
4455 - lfsck support in lustre kernel code (2349)
4456 - reduce journal credits needed for BRW writes (2370)
4457 - orphan handling to avoid losing space on client/server crashes
4458 - ptlrpcd can be blocked, stopping ALL progress (2477)
4459 - use lock value blocks to assist in proper KMS, faster stat (1021)
4460 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4461 - recovery for initial connections (2355)
4462 - fixes for mds_cleanup_orphans (1934)
4463 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4464 - block all file creations until orphan recovery completes (1901)
4465 - client remove rq_connection from request struct (2423)
4466 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4467 - recovery timer race (2670)
4468 - mdc_close recovey bug (2532)
4469 - ptlrpc cleanup bug (2710)
4470 - mds timeout on local locks (2588)
4471 - namespace lock held during RPCs (2431)
4472 - handle interrupted sync write properly (2503)
4473 - don't try to handle a message that hasn't been replied to (2699)
4474 - client assert failure during cleanup after abort recovery (2701)
4475 - leak mdc device after failed mount (2712)
4476 - ptlrpc_check_set allows timedout requests to complete (2714)
4477 - wait for inflight reqs when ptlrpcd finishes (2710)
4478 - make sure unregistered services are removed from the srv_list
4479 - reset bulk XID's when resending them (caught by 1138 test)
4480 - unregister_bulk after timeout
4481 - fix lconf error (2694)
4482 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4483 - readahead locks pages, leaves pending causing memory pressure (2673)
4484 - increase OST request buffers to 4096 on large machines (2729)
4485 - fix up permission of existing directories in simple_mkdir (2661)
4486 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4487 - don't assign transno to errored transactions (2742)
4488 - don't delete objects on OST if given a bogus objid from MDS (2751)
4489 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4490 - if rq_no_resend, then timeout request after recovery (2432)
4491 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4492 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4493 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4494 - MDS recovery shouldn't delete objects that it has given out (2730)
4495 - if enqueue arrives after completion, don't clobber LVB (2819)
4496 - don't unlock pages twice when trigger_group_io returns error (2814)
4497 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4498 - don't write pages to disk if there was an error (1450)
4499 - don't ping imports that have recovery disabled (2676)
4500 - take buffered bytes into account when balancing socknal conn (2817)
4501 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4502 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4503 - remove little-endian swabbing of llog records (1987)
4504 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4505 - retry reposting request buffers if they fail (1191)
4506 - grow extent at grant time to avoid granting a revoked lock (2809)
4507 - lock revoke doesn't evict page if covered by a second lock (2765)
4508 - disable VM readahead to avoid reading outside lock extents (2805)
4510 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4511 - updated kernel patches for hp-2.4.20 kernel (2681)
4513 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4516 - fix truncated write corruption (2366)
4517 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4519 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4520 - ensure len(uuid) < 37 in lmc (1171)
4521 - fix ia64 OOPS in llog_test (2255)
4522 - zero end of page at obdfilter for partial page writes (2648)
4523 - don't leave stale dentries around after renames (bug 2428)
4524 - fix timeouts when evicting a client with a single lock held (2642)
4525 - set deadline for the initial HELLO message to drain (2634)
4526 - print out dotted-quad IP addresses in the socknal (2302)
4528 - additional debugging for MDS client eviction problem (2443)
4529 - fix mkfsoptions support for osts (2603, 2604)
4531 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4534 - add series for the vanilla 2.6.0 kernel
4535 - add series for the vanilla 2.4.24 kernel
4536 - add series for a cray x86/64 UL kernel drop
4537 - fix xattr patches for the vanilla 2.4.19 series
4539 - generate true UUIDs in lmc (1171)
4540 - have portals stack dumping break in UML (2466)
4541 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4542 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4543 - wake callback waiting threads on client eviction (2460)
4544 - Add --ptldebug and --subsystem to lmc (1719)
4545 - update assertion to allow safe interrupt allocation
4546 - set rq_no_resend for cancel requests (2432)
4547 - recalculate ptlrpcd timeout after resend (2494)
4548 - call vfs_rmdir when removing pending directories (2368)
4549 - fix renaming a file to itself (2429)
4550 - lmc creates a default one-stripe lov (2454)
4551 - expand procfs space to handle large clusters (2326)
4552 - increase UML stack to avoid overflow
4553 - update lconf's list of debug and subsystem masks
4554 - fix lfs find --obd (2510)
4555 - /proc tunable for disabling filter read caching (2591)
4556 - stop rpm packages from altering slapd.conf (2301)
4557 - disable nagle in the socknal under 0conf (2578)
4558 - choose mds inode size based on stripe count (2572)
4559 - fix kernel-source rpm problems (2516)
4561 - add --disable-doc to avoid pdf generation (2421)
4562 - update documentation, tests, type-os, comments
4563 - avoid format warnings on ia64
4564 - remove the TOE NAL
4565 - tiny code cleanups by removing unused fields
4567 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4570 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4571 - remove the most blatant lies from BUILDING file (2371)
4572 - change default debug level to reasonable production setting
4573 - reduce client side cache size to reduce cache flush time
4574 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4575 - make TCP zerocopy and pinger support enabled by default (2476)
4576 - sync writes completed after process exits caused crashes (2319)
4577 - maintain correct mount count on the MDS (2356)
4578 - backout 1557, because 2316 wasn't really fixed
4579 - better file I/O statistics gathering in /proc
4580 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4581 - another kernel patch to fix zero-copy TCP function export
4582 - don't take duplicate lock when processing re-sent getattr (2420)
4583 - lctl uses obd_self_export instead of creating new conn (2353)
4584 - MDS/OST recovery case which requires object creation asserted (2425)
4585 - move lfs from /usr/sbin to /usr/bin in packages
4586 - fix race between mds_client_add and mds_client_free (2417)
4587 - use kmalloc instead of slabs in portals (2430)
4588 - don't create duplicate records when a failover MDS is present (2442)
4589 - remove unnecessary mount age check (2332)
4590 - don't remove directory inodes from locks prematurely (2451)
4591 - don't break if MDS service name is the same as hostname (2103)
4592 - fix races in client write RPC generation when cache full (2482)
4594 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4597 - remove now-unused request->rq_obd (278)
4598 - if an allocation fails, print out how much memory we've used (1933)
4599 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4600 - add the "configurable stack size" patch to most series files (1256)
4601 - ability to write large log records, for 100+ OST configs (2306)
4602 - fix NULL deref when filter_prep fails (2314)
4603 - fix operator precedence error in filter_sync
4604 - dynamic allocation of socknal TX descriptors (2315)
4605 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4606 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4607 - ensure that CWARN messages in Portals always get to the syslog
4608 - __init/__exit are not for prototype decls (ldlm_init/exit)
4609 - x86-64 compile warning fixes
4610 - fix gateway LMC keyword conflict (2318)
4611 - fix MDS lock inversions in getattr/reint paths (1844)
4612 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4613 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4614 - DLM race condition prevented some lock evictions (2328)
4615 - ENOMEM detection and retry on socknal sends (2230)
4616 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4617 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4618 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4619 - better error messages when a client is rejected during recovery (1505)
4620 - avoid cancelling locks which were never granted, after failure (2330)
4621 - fix i_sem/journal inversion in mds_client_add (2333)
4622 - fix truncate/getattr lock cycle deadlock (2334)
4623 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4624 - fix two rare exit paths which could leak an l_lock() ref (2321)
4625 - fencepost error in MDS/OST orphan recovery (2226)
4626 - make log record alignment 8 bytes (1988)
4627 - lstripe now fails when requested offset > ost_count (2237)
4628 - ensure that all kernel series have a complete list.h (1607)
4629 - fix crashes in special-file operations (2316)
4630 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4632 - allow configurable automake binary, for testing new versions
4633 - small update to the lfs documentation
4635 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4637 * fix negative export reference count in fsfilt_sync (2312)
4639 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4640 * release candidate 0.9.1
4642 - orphans are moved into the PENDING directory for possible recovery
4643 - replayed opens now open by fid for orphan/rename safety (1042)
4644 - last close of an orphan inode generates a transno (683)
4645 - chdir() and mount() now pin the directory entry (1020)
4646 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4647 - discard very old requests without processing them (1502)
4648 - remove some common, well-understood CERRORs (1505)
4649 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4650 - clear "grant" flags in OST replies until OST grant code lands (1644)
4651 - fix read performance by not clobbering i_blksize on client (1598)
4652 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4653 - make LRU size a /proc tunable, clears locks when reduced (707)
4654 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4655 - server locks take references on exports now (1558)
4656 - build fixes for 2.4.20-rh trees (1663)
4657 - return an error from lov_create if all OSCs are inactive (1751)
4658 - fix import levels when a reconnect happens without a timeout (1597)
4659 - exit early from mds_open if we get a lookup error (1749)
4660 - partial page read at EOF wouldn't wait for disk before sending (1642)
4661 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4662 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4663 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4664 - write/truncate lock inversion (1639)
4665 - Don't auto-load obdclass, portals modules during cleanup (1495)
4666 - fix timestamps from jumping to "now" (1763)
4667 - extra journal assertions (1648)
4668 - add an extra multiunlink test (1771)
4669 - fix read_record/write_record API (1776)
4670 - fix leak of offset_extent, possible incorrect i_size later (1772)
4671 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4672 - Debugging for the kqswnal_get_idle_tx problems (1820)
4673 - Allow recovery to be attempted multiple times (1536)
4674 - Write out MDS last_rcvd file after it is first created (1600)
4675 - Fix tx_descriptor leak in failed transmit situations (1827)
4676 - ext3 journaling fixes for assertion failure after IO error (1871)
4677 - class_export_put() on freed export after completion AST error (1896)
4678 - Fix revalidate looping in VFS (1322)
4679 - Don't access a freed export during MDS_REINT timeout (1521)
4680 - Add open-unlink recovery support on the MDS (1673,1764)
4681 - Return an error if no MDS data was read from last_rcvd (1946)
4682 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4683 - Don't setup a disk device that doesn't match exported UUID (317)
4684 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4685 - avoid committing NULL handle in force close
4686 - local.sh is now a one-stripe LOV configuration
4687 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4688 - fix ext3 htree duplicate directory entry corruption (1516)
4689 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4690 - update to new LOV EA format (2097)
4691 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4693 2003-06-15 Phil Schwan <phil@clusterfs.com>
4696 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4697 - per-import/export recovery handling (958, 931, 959)
4698 - multiple last-rcvd slots, for serving multiple FSes (949)
4699 - connections are again shared between multiple imp/exports (963, 964)
4700 - "umount -f" would hang if any requests needed to be sent (393, 978)
4701 - avoid pinning large req buffer by copying for queued messages (989)
4702 - add "uuid" to "lctl device" command to help upcalls (991)
4703 - "open" RPCs with transnos would confuse recovery counters (1037)
4704 - do proper endian conversion of all wire messages (288, 340, 891)
4705 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4706 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4707 - fix ptlrpc_connection leak in target_handle_connect (1174)
4708 - fix import refcounting bug in OST and MDS cleanup (1134)
4709 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4710 - fix very unlikely obd_types race condition (501)
4711 - remove osc_open hack for echo_client (1187)
4712 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4713 - a failure in read_inode2 leads to deadlock (1139)
4714 - cancel ack-locks as soon as transaction is committed (1072)
4715 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4716 - make sure to commitrw after any preprw to avoid deadlock (1162)
4717 - failing to execute a file in a lustre FS would lock inode (1203)
4718 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4719 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4720 - obd_uuid2tgt would walk off the end of the list (1255)
4721 - on IA64 the osc would give portals incorrect bulk size (1258)
4722 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4723 - fix lock inversion caused by new llite matching code (1282)
4724 - limit the number of dirty pages on a client to 10MB (1286)
4725 - timed out locks were not being corrected cancelled (1289)
4726 - fix O_DIRECT above 4GB on IA-32 (1292)
4727 * major user-visible changes
4728 - fail out/fail over policy now controlled by the upcall (993)
4730 - add OBD_PING to check server availability and failure (954)
4731 - lustre messages are now sent in sending host order (288, 340, 891)
4732 - add eadatalen to MDS getattr reply (340)
4733 - OST read replies may contain second buffer, with per-page status (593)
4735 2003-03-11 Phil Schwan <phil@clusterfs.com>
4738 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4739 - failed open()s could cause deadlock; fixed (867, 869)
4740 - stop cancelling OST locks when files are closed (481)
4741 - overlapping XID spaces caused network corruption (851, 853)
4742 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4743 - setattr_raw added, to do single-RPC, server-side setattrs
4744 - lmc/lconf syntax change for OST UUIDs
4745 - fix crashy race condition between ptlrpc_free_req and osc_close
4746 - don't use request in mdc_enqueue if we hit a timeout (889)
4747 - don't set the inode i_size for regular files from the MDS (896)
4748 - handle out of order completion AST (842)
4749 - don't LBUG if a lock request times out after receiving AST (913)
4750 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4751 - if a bad lock AST arrives, send an error instead of dropping entirely
4752 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4753 - fix leak in bulk IO when only partially completed (899, 900, 926)
4754 - fix O_DIRECT for ia64 (55)
4755 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4756 - C-z after timeout could hang a process forever; fixed (977)
4758 - client-side I/O cache (678, 924, 929, 941, 970)
4760 - READPAGE and SETATTRs which don't take server-side locks get
4763 2003-02-11 Phil Schwan <phil@clusterfs.com>
4766 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4767 - Fix fsfilt_extN_readpage() to read a full page of directory
4768 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4769 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4770 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4771 - Most intent operations were removed, in favour of a new RPC mode
4772 that does a single RPC to the server and bypasses most of the VFS
4773 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4774 - Aggressively cancel local locks on DLM servers
4775 - mds_reint_unlink sends EA to the client if it's the last nlink.
4776 client uses that EA to unlink OST objects.
4777 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4778 - recursive symlinks were fixed (439)
4779 - fixed NULL deref in DEBUG_REQ
4780 - filter_update_lastobjid no longer calls sync, which annoyed extN
4781 - fixed multi-client small-writes to a single file problem (445)
4782 - fixed mtime updates during file writes (607)
4783 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4784 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4785 - fixed Linux OST size reporting problem (444, 656)
4786 - OST now updates object mtime with writes or setattr (607, 619)
4787 - client verifies file size before zeroing page past EOF (445)
4788 - OST now writes last allocated objid to disk with allocation (108)
4789 - LOV on echo now works (409)
4791 - mds_reint_unlink sends a new buffer, with the EA included. this
4792 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4793 set if a regular file was being unlinked, and it was the last link
4794 - use PtlGet from the target for bulk writes (315)
4795 - OST now updates object mtime with writes or setattr (607, 619)
4796 - LDLM now has a grant-time callback to revalidate locked items, if
4798 - Many MDS operations were reorganized to combat race conditions
4800 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4801 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4803 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4806 - Fully reactivate OST imports after reconnection (512, others)
4807 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4808 - More graceful error handling for truncating on dead OST (515)
4809 - Don't error out unless we're actually accessing dead stripes (474)
4810 - Fix garbage sizes when stripes are missing (410)
4811 - LRU counters were broken, causing constant lock purge (433, 432)
4812 - garbage on read from stripes with failed OSTs (441)
4813 - mark OSCs as active before reconnecting during recovery (438)
4814 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4815 - lfind did not preserve OST order in output (443)
4816 - symlinks cause hung clients, incorrect data (439)
4817 - stop dereferencing request after dropping refcount (457)
4818 - don't LASSERT(spin_is_locked) on non-SMP (455)
4819 - fixes for many rename() bugs
4820 - fstat didn't correctly synchronize attributes (399)
4821 - server must handle lock cancellation during blocking AST prep (487)
4822 - bulk descriptors were free()d too soon (511)
4823 - fix paths in lconf, which would load incorrect modules (451, 507)
4824 - fix confusing lconf 'host not found' error message (386)
4825 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4826 - fix race condition in mdc_blocking_ast() for inode access (526)
4827 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4828 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4829 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4830 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4831 - fix lstripe in several ways (kernel side) (527)
4832 - fix request leak in ldlm_cli_enqueue (262)
4833 - incorrect OSC was marked inactive after OST failure
4834 - call mds_fs_cleanup before unmounting filesystem (524)
4835 - fix races between taking ns_lock and ldlm_lock_change_resource
4836 - fix races updating LOV export open file list
4837 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4838 - fix recovery NULL deref in ldlm_cli_cancel_unused
4839 - fix some DLM races by using new hash table for lock handles (419)
4840 - permit the client to specify desired inodes, at replay
4841 - duplicate requests when we queue them for replay reintegration
4842 - fix last_rcvd offset calculation
4843 - sync after each recovered transaction, so we always make progress
4844 - never, not always, ERESTART requests without transnos
4845 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4847 - skip replay if the MDS says that the client is already connected
4848 - don't check for a recovery-enabled export to match lctl's UUID
4849 - don't INC_USE_COUNT for phantom exports
4850 - don't crash when cleaning up phantom exports (567)
4851 - don't double-finish or set replay data for errored mdc_open requests
4852 - abort requests when they time out, so we don't get old replies
4853 - send/receive replies for AST messages again
4854 - if the client says that it doesn't have the lock, cancel it on the
4856 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4857 mark it as destroyed, it will all work out when decref is called
4858 - fix module use counts (22, 581)
4860 - ASTs now expect a reply (server cancels lock on error reply)
4862 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4865 - fix many simultaneous client startup (392)
4866 - fix dentry->d_it clobbering
4867 - credentials weren't being shipped for readdir/getattr operations
4868 - remove invalid assertions triggered during some concurrent MD
4870 - proper Lustre versions added (336, 389)
4871 - fix memory leak for create error case (398)
4872 - fix LOV locking bug that would get cli/srv out of sync
4873 - fix echo client over LOV (409)
4874 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4875 - fix double-O_EXCL intent crash (424)
4876 - avoid sending multiple lock CANCELs (352)
4878 - MDS can do multi-client recovery (modulo bugs in new code)
4880 - many updates, edits, cleanups
4882 2002-11-18 Phil Schwan <phil@clusterfs.com>
4885 - fix null d_it dereference (346)
4886 - fix full OST/dbench hang (333)
4887 - fix permission problem with file removal (286)
4888 - fix removal of OSCs from LOV when they fail
4889 - fix NULL deref during bulk timeout (214)
4890 - fix problems related to multiple filesystems on one MDS (241)
4891 - fixed serious subtle metadata locking bugs
4892 - free locks on clients when inodes are removed due to memory
4894 - fix inode pointer in lock data (285)
4895 - partial support for multiple MDS on a single host (241)
4896 - data locks weren't cancelled at clear_inode time (290, 311)
4897 - intent locks could lead to unbounded lock growth (205)
4898 - added a maximum lock count, an LRU list, and a flusher
4899 - fix multiple rename (365)
4900 - properly abstracted the echo client
4901 - OSC locked 1 byte too many; fixed
4902 - rewrote brw callback code:
4903 - fixed recovery bugs related to LOVs (306)
4904 - fixed too-many-pages-in-one-write crash (191)
4905 - fixed (again) crash in sync_io_timeout (214)
4906 - probably fixed callback-related race (385)
4908 - Add capability to MDS protocol
4909 - LDLM cancellations and callbacks on different portals
4911 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4914 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4915 - defer open object destruction to close time (601981, 138)
4916 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4917 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4918 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4919 - don't call obd_setattr in ll_file_release for destroyed objects
4920 * protocol change to lustre_msg: move |version| and add |flags|
4921 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4922 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4923 * added replay of create, unlink, link and rename operations during
4924 MDS failover; recovery should be much more robust now
4925 * remove failed OSCs from LOVs (only lov_create uses this so far)
4926 * the lustre-HOWTO was brought (more) up to date (582544)
4928 2002-10-23 Phil Schwan <phil@clusterfs.com>
4931 - in-use dentries weren't being reused properly (617851)
4932 - prevent multiple LDLM setup (599178)
4933 - fix LOV size calculations for truncate (617853)
4934 - fix client handling of MDS intent errors (POSIX)
4935 - fix permission bug in lovstripe.c test (624321)
4936 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4937 - truncate past end of file could corrupt data
4938 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4939 - a race in recovery could return ETIMEDOUT to apps (623947)
4940 - building outside the source directory was fixed
4941 * the lustre-HOWTO was brought (more) up to date (582544)
4942 * major progress was made on recovery functionality
4944 2002-10-10 Phil Schwan <phil@clusterfs.com>
4947 - recovery deadlock fix
4948 - rm -rf causes LBUG fix (617817)
4949 - file open by multiple tasks fix (618962)
4950 - directory permissions bugs (602707 and 620007)
4951 - journal_stop fixed with locking (611313)
4952 - O_APPEND failures resolved (618273, perhaps 614459)
4953 - lconf PATH fix (619770)
4954 - IA64 build fix (621450)
4955 - RPC buffer sizes scale with amount of memory
4957 2002-10-01 Phil Schwan <phil@clusterfs.com>
4960 - locks would be cancelled without throwing away data pages,
4961 resulting in inconsistent data (605627)
4962 - inode attributes were not always being refreshed (605627, 612449)
4963 - lconf now continues to cleanup after lctl reports an error
4964 - MDS now enforces user permissions (602707)
4965 - lprocfs cleanup fixed, but not yet enabled (614157)
4966 - fixed infinite server hang, should a client not respond to an AST
4967 - avoid going into recovery if user calls readlink() with a buffer
4968 that's too small (613941)
4969 - AST RPCs no longer require replies (614867) -- this may be changed
4970 - don't crash server if client sends an IOV that's too big (611336)
4971 - fixed lock conversion deadlock (611892)
4972 - fixed the following of symlinks (614622)
4973 * recovery: the server can remove locks from a client that dies, other
4974 clients can make progress
4975 * more extN patch fixes
4976 * compile-time configurable ptlrpc buffer allocations
4978 - collaborative read cache document
4979 - Lustre Lite Performance CDR document-in-progress
4981 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4984 - fix typo in patch-2.4.18
4986 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4989 - clear ptlrpc request each time in handle_incoming_request()
4990 - unlink of files now destroys the object on the OST
4992 2002-09-19 Peter Braam <braam@clusterfs.com>
4994 * add hard link support
4995 * change obdfile creation method
4996 * kernel patch changed
4998 2002-09-19 Peter Braam <braam@clusterfs.com>
5001 - stack overflow bug in extN fixed
5003 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5005 * documentation updates
5006 - add man pages for config tools
5007 - update tests/README to describe testing with new config tools
5008 - finish metadata API descriptions
5009 * bug fixes and cleanups
5010 - statfs workaround for 16TB limit
5011 - LOV stripe allocation improved, can stripe on subset of OSTs
5012 - LOV file size/IO offset was wrong for files > 4GB in size
5013 - object EA data was being dropped, caused files to be unreadable
5014 - memory overflow with non-LOV OST caused memory corruption
5015 - fixed regression tests to work with new config tools, obdfilter
5016 - fixed bug when directory size became larger than 1 block
5017 - fixed bug (for single client case) when PWD was deleted
5018 - invalidate local directory pages when doing intent-based ops
5019 - avoid LDLM oops when lock callback contained bad data
5021 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5023 * documentation updates
5024 * bug fixes and cleanups
5025 - configuration tools
5028 - 64-bit compile warnings
5029 - 64-bit internal statfs data
5031 * test_brw on persistent OST devices
5033 * lprocfs (disabled)
5035 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5037 * documentation updates
5038 * bug fixes and cleanups
5039 * configuration tools
5041 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5044 * many small fixes to 0_5_4
5045 * io/network handling
5046 * thinkos in MDS operations
5048 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5051 * crucial basic fixes to 0.5.3
5052 * IOR, Iozone work over Elan
5055 2002-08-07 Phil Schwan <phil@clusterfs.com>
5056 * version 0_5_3, our first alpha
5057 * we use the new Portals iovs
5058 * documentation updates
5059 * bug fixes and cleanups
5060 * small changes in the DLM wire protocol
5062 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5063 * version 0_5_1 with some initial stability,
5064 * locking on MD and file I/O.
5065 * documentation updates
5066 * several bug fixes since 0.5.0
5067 * small changes in wire protocol
5069 2002-07-18 Phil Schwan <phil@clusterfs.com>
5071 * delivered as Lustre Light Alpha
5072 * fixed a crash after handling invalid MDS requests
5073 * fixed directory pages for architectures with non-4k pages sizes
5075 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5076 * release version v0_4_4
5077 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5079 * Fixes a number of interruption problems with OST operations.
5080 * Update documentation for portals header changes
5081 * Move all wire protocol structs/defines to lustre_idl.h
5082 * Fixes symlink length bug.
5083 * Add tcpdump to repository.
5085 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5086 * release version v0_4_3
5087 * Fixes statfs for inodes on extN.
5088 * Fixes bug in runtests which would delete /etc/hosts.
5089 * Use 64-bit object IDs wherever possible (not into VFS though)
5090 Remove ost_get_info, which is unused by lustre, and out of date.
5092 2002-07-03 Peter Braam <braam@clusterfs.com>
5093 * release version v0_4_2 Fixes a lookup error (type not passed)
5094 * move forward to head of Portals
5095 * move forward to latest Lustre kernel
5097 2002-06-25 Peter Braam <braam@clusterfs.com>
5098 * release version v0_4_1. Hopefully stable on single node use.