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).
17 Description: Remove portals compatibility
18 Details : Remove portals compatibility, not interoperable with releases
23 Description: Resolve device initialization race
24 Details : Prevent proc handler from accessing devices added to the
25 obd_devs array but yet be intialized.
27 Severity : enhancement
29 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
31 Severity : enhancement
33 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
38 Description: Using wrong pointer in osc_brw_prep_request
39 Details : Access to array[-1] can produce panic if kernel compiled with
40 CONFIG_PAGE_ALLOC enabled
44 Description: Client runs out of low memory
45 Details : Consider only lowmem when counting initial number of llap pages
49 Description: *optional* service tags registration
50 Details : if the "service tags" package is installed on a Lustre node
51 When the filesystem is mounted, a local-node service tag will
52 be created. See http://inventory.sun.com/ for more information
53 about the Service Tags asset management system.
57 Description: Kernel BUG tries to release flock
58 Details : Lustre does not destroy flock lock before last reference goes
59 away. So always drop flock locks when client is evicted and
60 perform unlock regardless of successfulness of speaking to MDS.
64 Description: add recount protection for osc callbacks, so avoid panic on shutdown
68 Description: sanity test 65a fails if stripecount of -1 is set
69 Details : handle -1 striping on filesystem in ll_dirstripe_verify
74 Description: ASSERTION(CheckWriteback(page,cmd)) failed
75 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
78 Severity : enhancement
80 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
84 Description: do not process already freed flock
85 Details : flock can possibly be freed by another thread before it reaches
86 to ldlm_flock_completion_ast.
90 Description: LBUG during stress test
91 Details : Need properly lock accesses the flock deadlock detection list.
95 Description: oops in page fault handler
96 Details : kernel page fault handler can return two special 'pages' in error case, don't
97 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
101 Description: timeout with invalidate import.
102 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
103 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
104 ptlrpc_invalidate_import and as result LASSERT.
106 Severity : enhancement
108 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
112 Description: enable MGS and MDT services start separately
113 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
114 a co-located MGS without starting the MGS, which is a complement
115 to 'nosvc' mount option.
119 Frequency : after recovery
120 Description: precreate to many object's after del orphan.
121 Details : del orphan st in oscc last_id == next_id and this triger growing
122 count of precreated objects. Set flag LOW to skip increase count
123 of precreated objects.
127 Frequency : rare, on clear nid stats
128 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
129 Details : when clean nid stats sometimes try destroy live entry,
130 and this produce panic in free.
134 Description: Stack overflow during MDS log replay
135 ease stack pressure by using a thread dealing llog_process.
139 Description: wait until IO finished before start new when do lock cancel.
140 Details : VM protocol want old IO finished before start new, in this case
141 need wait until PG_writeback is cleared until check dirty flag and
142 call writepages in lock cancel callback.
144 Severity : enhancement
146 Description: using special macro for print time and cleanup in includes.
150 Description: mds_mfd_close() ASSERTION(rc == 0)
151 Details : In mds_mfd_close(), we need protect inode's writecount change
152 within its orphan write semaphore to prevent possible races.
156 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
160 Frequency : rare, on shutdown ost
161 Description: don't hit live lock with umount ost.
162 Details : shrink_dcache_parent can be in long loop with destroy dentries,
163 use shrink_dcache_sb instead.
167 Description: don't panic with use echo client
168 Details : echo client pass NULL as client nid pointer and this produce null
173 Description: fix build on ppc32
174 Details : compile code with -m64 flag produce wrong object file for ppc32.
178 Description: add message levels for liblustreapi
182 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
183 Details : Move the 'good_osts' check before the 'total_bavail' check. This
184 will result in an -EAGAIN and in the exit call path we call
185 alloc_rr() which will with increasing aggressiveness attempt to
186 aquire precreated objects on the minimum number of required OSCs.
190 Description: Use old size assignment to avoid deadlock
191 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
192 the scheduling while holding a spinlock. We do not need locking
193 for size in ll_update_inode() because size is only updated from
194 the MDS for directories or files without objects, so there is no
195 other place to do the update, and concurrent access to such inodes
196 are protected by the inode lock.
200 Description: resolve "_IOWR redefined" build error on SLES10
204 Description: dump the memory debugging after all modules are unloaded to
205 suppress false negative in conf_sanity test 39
207 Severity : enhancement
209 Description: build kernel-ib packages for OFED 1.3 in our release cycle
214 Description: fix SLES kernel versioning
215 Details : the kernel version for our SLES 10 kernel did not include a "-"
216 before the "smp" at the end. while this was not a problem in
217 general, it did mean that software trying to use the kernel
218 version to try to detect a vendor specific kernel would fail.
219 this was most evident by the OFED build scripts.
223 Description: Don't update lov_desc members until making sure they are valid
224 Details : When updating lov_desc members via proc fs, need fix their
225 validities before doing the real update.
229 Description: don't put request into delay list while invalidate in flight.
230 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
231 import in flight. this produce timeout for invalidate and sometimes
232 can cause stale data.
236 Frequency : on ppc only
237 Description: not convert ost objects for directory because it's not exist.
238 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
242 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
244 Severity : enhancement
246 Description: Update to support 2.6.22.14 vanilla kernel.
250 Frequency : rare, on recovery
251 Description: read procfs can produce deadlock in some situation
252 Details : Holding lprocfs lock which send rpc can produce block for destroy
253 obd objects and this also block reconnect with -EALREADY. This isn't
254 fix all lprocfs bugs - but make it rare.
256 Severity : enhancement
258 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
261 Frequency : frequent on X2 node
263 Description: mdc_set_open_replay_data LBUG
264 Details : Set replay data for requests that are eligible for replay.
268 Description: lustre_mgs: operation 101 on unconnected MGS
269 Details : When MGC is disconnected from MGS long enough, MGS will evict the
270 MGC, and late on MGC cannot successfully connect to MGS and a lot
271 of the error messages complaining that MGS is not connected.
275 Frequency : on network error
276 Description: panic with double free request if network error
277 Details : mdc_finish_enqueue is finish request if any network error ocuring,
278 but it's true only for synchronus enqueue, for async enqueue
279 (via ptlrpcd) this incorrect and ptlrpcd want finish request
282 Severity : enhancement
284 Description: client-side metadata stat-ahead during readdir(directory readahead)
285 Details : perform client-side metadata stat-ahead when the client detects
286 readdir and sequential stat of dir entries therein
289 Frequency : on start mds
291 Description: Implement get_info(last_id) in obdfilter.
294 Frequency : occasional
296 Description: Correctly check stale fid, not start epoch if ost not support SOM
297 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
298 overwrited with new generated fid, but mds can anwer with one of these
299 two fids and both is not stale. setattr incorectly start epoch and
300 assume will be called done_writeting, but without SOM done_writing
304 Frequency : rare, depends on device drivers and load
306 Description: MDS or OSS nodes crash due to stack overflow
307 Details : Code changes in 1.8.0 increased the stack usage of some functions.
308 In some cases, in conjunction with device drivers that use a lot
309 of stack the MDS (or possibly OSS) service threads could overflow
310 the stack. One change which was identified to consume additional
311 stack has been reworked to avoid the extra stack usage.
314 Frequency : occasional
316 Description: Do not fail import if osc_interpret_create gets -EAGAIN
317 Details : If osc_interpret_create got -EAGAIN it immediately exits and
318 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
319 oscc_has_objects and see OSC has no objests and call
320 oscc_internal_create to resend create request.
322 Severity : enhancement
324 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
326 Severity : enhancement
328 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
331 Frequency : very rare
333 Description: Fix replay if there is an un-replied request and open
334 Details : In some cases, older replay request will revert the
335 mcd->mcd_last_xid on MDS which is used to record the client's
338 Severity : enhancement
340 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
342 Severity : enhancement
344 Description: Add rhel5 support to HEAD.
346 Serverity : enhancement
348 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
353 Description: Don't allow skipping OSTs if index has been specified.
354 Details : Don't allow skipping OSTs if index has been specified, make locking
355 in internal create lots better.
359 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
360 Details : access to bitfield in structure is always rounded to long
361 and this produce problem with not atomic change any bit.
365 Description: Lustre make rpms failed.
366 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
367 builds Lustre rpms from tarball.
370 Frequency : rare on shutdown ost
372 Description: If llog cancel was not send before clean_exports phase, this can
373 produce deadlock in llog code.
374 Details : If llog thread has last reference to obd and call class_import_put
375 this produce deadlock because llog_cleanup_commit_master wait when
376 last llog_commit_thread exited, but this never success because was
377 called from llog_commit_thread.
381 Description: allow userland application know is lost one of stripes.
382 Details : fill lvb_blocks with error code on ost and return it to
383 application if error flag found.
387 Description: NULL lov_tgts causing MDS oops
388 Details : more safe checks for NULL lov_tgts for avoid oops.
390 Severity : enhancement
392 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
396 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
397 Details : If an OST is down the MDS will hang indefinitely in
398 obd_statfs_rqset() waiting for the statfs data. While for
399 MDS QOS usage of statfs, it should not stuck in waiting.
401 Severity : enhancement
403 Description: remote_acl support
404 Details : Support ACL-based permission check for remote user.
405 Support setfacl/getfacl for remote user with the utils
406 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
407 the system "{s,g}etfacl" utils.
409 Severity : enhancement
411 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
413 Severity : enhancement
415 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
419 Description: make mgs_setparam() handle fsname containing dash
420 Details : fsname containing a dash does not work with lctl conf_param
422 Severity : enhancement
424 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
426 Severity : enhancement
428 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
432 Frequency : rare, at shutdown
433 Description: access already free / zero obd_namespace.
434 Details : if client_disconnect_export was called without force flag set,
435 and exist connect request in flight, this can produce access to
436 NULL pointer (or already free pointer) when connect_interpret
437 store ocd flags in obd_namespace.
441 Frequency : only at startup
442 Description: not alloc memory with spinlock held.
443 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
444 if any spinlock held.
446 Severity : enhancement
448 Description: make lustre randomly fail allocating memory
449 Details : Make lustre randomly failed allocating memory for testing purpose.
451 Severity : enhancement
453 Description: lost problems with lov objid file
454 Details : Fixes some scability and access to not inited memory problems
455 in work with lov objdid file.
460 Description: lfs find does not continue on file error
461 Details : Continue other files processing when a file/dir is absent.
465 Description: Inconsistent usage of lustre_pack_reply()
466 Details : Standardize the usage of lustre_pack_reply() such that it
467 always generate a CERROR on failure.
470 Frequency : occasional
472 Description: MDS hang or stay in waiting lock
473 Details : If client receive lock with CBPENDING flag ldlm need send lock
474 cancel as separate rpc, to avoid situation when cancel request
475 can't processed due all i/o threads stay in wait lock.
479 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
483 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
487 Description: Update to SLES9 kernel-2.6.5-7.287.3.
491 Description: Update to RHEL4 latest kernel.
493 Severity : enhancement
495 Description: Build SLES10 patchless client fails
496 Details : The configure was broken by run ./configure with
497 --with-linux-obj=.... argument for patchless client. When the
498 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
499 can't search header adequately. Use absolute path such as
500 -I($LINUX)/include instead.
504 Description: interrupt oig_wait produce painc on resend.
505 Details : brw_redo_request can be used for resend requests from ptlrpcd and
506 private set, and this produce situation when rq_ptlrpcd_data not
507 copyed to new allocated request and triggered LBUG on assert
508 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
509 ptlrpcd set if request is changed and can be safety changed to use
514 Description: LASSERT_{REQ,REP}SWAB macros are buggy
515 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
516 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
517 these macros and replace them with their difinitions instead.
521 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
522 Details : Update which_patch & target file for SLES10 latest kernel.
525 Description: Remove client patches from SLES 10 kernel.
526 Details : This causes SLES 10 clients to behave as patchless clients
527 even on a Lustre-patched (server) kernel.
529 Severity : enhancement
531 Description: self-adjustable client's lru lists
532 Details : use adaptive algorithm for managing client cached locks lru
533 lists according to current server load, other client's work
534 pattern, memory activities, etc. Both, server and client
535 side namespaces provide number of proc tunables for controlling
538 Severity : enhancement
540 Description: light-weight GSS support
541 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
542 and reduce performance overhead.
544 Severity : enhancement
546 Description: Linux keyring support
547 Details : Support using service of Linux keyring for Lustre GSS internal
548 context refresh/cache mechanism.
552 Description: Fix errors in lfs documentation
553 Details : Fixes man pages
557 Description: don't use GFP_* in generic Lustre code.
558 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
562 Description: obdclass is limited by single OBD_ALLOC(idarray)
563 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
567 Description: warnings with build patchless client with vanila 2.6.19 and up
568 Details : change old ctl_table style and replace ctl_table/ctl_table_header
569 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
573 Frequency : Only for SLES
574 Description: sanity_quota fail test_1
575 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
576 file, which makes TSTID[2] inunique.
580 Frequency : Only for SLES9
581 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
582 Details : Update target/ChangeLog/which_patch .
586 Frequency : Only for test_10 in sanity-quota.sh
587 Description: a bug in quota test code
588 Details : checking a test flag in a wrong place
592 Frequency : Only for test_13 in sanity-quota.sh
593 Description: a bug in quota test script
594 Details : shouldn't check overfull allocated quota
598 Description: jbd statistics
599 Details : Port older jbd statistics patch for sles10
601 Severity : enhancement
603 Description: Add printing inode info into message about error in writepage.
608 Description: reply_lock_interpret crash due to race with it and lock cancel.
609 Details : Do not replay locks that are being cancelled. Do not reference
610 locks by their address during replay, just by their handle.
614 Frequency : When flocks are used.
615 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
616 Details : Flock locks might destroy just granted lock if it could be merged
617 with another existing flock, this is done in completion handler,
618 so teach ldlm_cli_enquque_fini that this is a valid case for
624 Description: Oops in read and write path when failing to allocate lock.
625 Details : Check if lock allocation failed and return error back.
629 Description: lstripe command fails for valid OST index
630 Details : The stripe offset is compared to lov->desc.ld_tgt_count
631 instead of lov->desc.ld_active_tgt_count.
635 Description: sanity.sh failed test 103
636 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
641 Description: df doesn't work properly if diskfs blocksize != 4K
642 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
646 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
647 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
649 Severity : enhancement
651 Description: Adaptive timeouts
652 Details : RPC timeouts adapt to changing server load and network
653 conditions to reduce resend attempts and improve recovery time.
657 Description: llapi_file_create() does not allow some changes
658 Details : add llapi_file_open() that allows specifying the mode and
659 open flags, and also returns an open file handle.
663 Description: merge and cleanup kernel patches.
664 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
668 Description: Add journal checksum support.(Kernel part)
669 Details : The journal checksum feature adds two new flags i.e
670 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
671 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
672 indicates that the commit block contains the checksum for
673 the blocks described by the descriptor blocks. Now commit
674 record can be sent to disk without waiting for descriptor
675 blocks to be written to disk. This behavior is controlled
676 using JBD2_FEATURE_ASYNC_COMMIT flag.
680 Description: OSS needs mutliple precreate threads
681 Details : Add ability to start more than one create thread per OSS.
685 Description: Sanity.sh test_65k failed: setstripe should have succeeded
686 Details : execute lfs setstripe on client
690 Description: mds_obd_create error creating tmp object
691 Details : When the user sets quota on root, llog will be affected and can't
692 create files and write files.
695 Frequency : Always on ia64 patchless client, and possibly others.
697 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
698 Details : This allows the patchless client to be loaded on architectures
703 Description: RedHat Update kernel for RHEL5
704 Details : Add the kernel config file for RHEL5.
708 Description: RedHat Update kernel for RHEL5
709 Details : Modify the kernel config file more closer RHEL5.
713 Description: Build failure against Centos5 (RHEL5)
714 Details : Define PAGE_SIZE when it isn't present.
718 Description: test_7 on sanity-quota.sh can't be used on separate machines
719 Details : add facet to handle it
723 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
724 Details : with NFS, the anon dentry's parent was set to itself in
725 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
726 corresponding dentry other than use rec->ur_name.
728 Severity : enhancement
730 Description: lfs setstripe enhancement
731 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
735 Description: enable data checksumming by default
736 Details : enable checksum by default, allow --disable-checksum
737 configure option and "-o nochecksum" mount option. Checksums
738 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
742 Description: lustre support for RHEL5
743 Details : Add support for RHEL5.
747 Description: Client eviction due to failover config
748 Details : after a connection loss, the lustre client should attempt to
749 reconnect to the last active server first before trying the
750 other potential connections.
754 Description: when mds and osts use different quota unit(32bit and 64bit),
755 quota will be released repeatly.
756 Details : void sending multiple quota reqs to mds, which will keep the
757 status between the reqs.
761 Description: osts not allocated evenly to files
762 Details : change the condition to increase offset_idx
765 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
767 Description: Data corruption for OSTs that are formatted larger than 2TB
769 Details : When generating the bio request for lustre file writes the
770 sector number would overflow a temporary variable before being
771 used for the IO. The data reads correctly from Lustre (which
772 will overflow in a similar manner) but other file data or
773 filesystem metadata may be corrupted in some cases.
777 Description: Tune the kernel for good SCSI performance.
778 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
779 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
784 Description: rewrite ext2-derived code in obdclass/uuid.c
785 Details : rewrite inherited code (uuid parsing code from ext2 utils)
786 from scratch preserving functionality.
790 Description: "lfs find -obd UUID" prints directories
791 Details : "lfs find -obd UUID" will return all directory names instead
792 of just file names. It is incorrect because the directories
793 do not reside on the OSTs.
797 Description: use i_size_read and i_size_write in 2.6 port
798 Details : replace inode->i_size access with i_size_read/write()
801 Frequency : when removing large files
803 Description: scheduling issue during removal of large Lustre files
804 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
805 It causes scheduling issues when removing large files (17TB in the
809 Frequency : only with liblustre clients on XT3
811 Description: evictions taking too long
812 Details : allow llrd to evict clients directly on OSTs
815 Frequency : only on ppc
817 Description: /proc/fs/lustre/devices broken on ppc
818 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
819 We should make sure the type of 'index' is loff_t and then cast
820 explicitly as needed below. Do not assign an explicitly cast
824 Frequency : only for rhel5
826 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
827 Details : Modify the target file & which_kernel.
829 Severity : enhancement
831 Description: omit set fsid for export NFS
832 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
833 as FS_REQUIRES_DEV for avoid problems with generate fsid.
837 Frequency : Always, for kernels after 2.6.16
838 Description: Fix warning idr_remove called for id=.. which is not allocated.
839 Details : Last kernels save old s_dev before kill super and not allow
840 to restore from callback - restore it before call kill_anon_super.
844 Description: buffer overruns could theoretically occur
845 Details : llapi_semantic_traverse() modifies the "path" argument by
846 appending values to the end of the origin string, and a
847 overrun may occur. Adding buffer overrun check in liblustreapi.
851 Description: Fix error on 'ls .' at the top of the Lustre mount.
852 Details : Don't revalidate dentry if it is a root dentry.
856 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
857 Details : Modify vm-tunables-rhel4.patch.
861 Description: Kernel config for 2.6.18-vanilla.
862 Details : Modify targets/2.6-vanilla.target.in.
863 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
864 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
865 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
866 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
870 Description: improve handling recoverable errors
871 Details : if request processig with error which can be recoverable on server
872 request should be resend, otherwise page released from cache and
877 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
878 Details : Modify target file & which_patch.
879 A flaw was found in the IA32 system call emulation provided
880 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
881 value could be stored in the %RAX register, which could trigger an
882 out-of-bounds system call table access. An untrusted local user
883 could exploit this flaw to run code in the kernel
884 (ie a root privilege escalation). (CVE-2007-4573).
888 Description: O_DIRECT bypasses client statistics.
889 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
890 read_ahead_stats were updated. copied the stats section from
891 osc_send_oap_rpc() into async_internal().
895 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
899 Description: change order of libsysio includes
900 Details : '#include sysio.h' should always come before '#include xtio.h'
904 Description: handle "serious error: objid * already exists" more gracefully
905 Details : If LAST_ID value on disk is smaller than the objects existing in
906 the O/0/d* directories, it indicates disk corruption and causes an
907 LBUG(). If the object is 0-length, then we should use the existing
908 object. This will help to avoid a full fsck in most cases.
910 Severity : enhancement
912 Description: adapt the lustre_config script to support the upgrade case
913 Details : Add "-u" option for lustre_config script to support upgrading 1.4
914 server targets to 1.6 in parallel.
918 Description: To avoid grant space > avaible space when the disk is almost
919 full. Without this patch you might see the error "grant XXXX >
920 available" or some LBUG about grant, when the disk is almost
922 Details : In filter_check_grant, for non_grant cache write, we should
923 check the left space by if (*left > ungranted + bytes), instead
924 of (*left > ungranted), because only we are sure the left space
925 is enough for another "bytes", then the ungrant space should be
926 increase. In client, we should update cl_avail_grant only there
927 is OBD_MD_FLGRANT in the reply.
931 Description: Update RHEL 4 kernel to fix local root privilege escalation.
932 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
933 described in CVE-2007-4573. This problem could allow untrusted
934 local users to gain root access.
937 Frequency : when using O_DIRECT and quotas
939 Description: Incorrect file ownership on O_DIRECT output files
940 Details : block usage reported by 'lfs quota' does not take into account
941 files that have been written with O_DIRECT.
946 Description: touch file failed when fs is not full
947 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
948 otherwise we can get ENOSP while fs is not full.
952 Description: parallel lock callbacks
953 Details : Instead of sending blocking and completion callbacks as separated
954 requests, adding them to a set and sending in parallel.
957 Frequency : only for Cray XT3
958 Bugzilla : 12829/13455
959 Description: Changing primary group doesn't change the group lustre assigns to
961 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
962 overridden with the primary group provided by the group upcall,
963 whereas the supplied fsgid can be trusted if it is in the list of
964 supplementary groups returned by the group upcall.
966 Severity : enhancement
968 Description: Allow masking D_WARNING, D_ERROR messages from console
969 Details : Console messages can now be disabled via lnet.printk.
973 Description: User code with malformed file open parameter crashes client node
974 Details : Before packing join_file req, all the related reference should be
975 checked carefully in case some malformed flags cause fake
976 join_file req on client.
980 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
981 Details : ldlm_completion_ast() assumes that a lock is granted when the req
982 mode is equal to the granted mode. However, it should also check
983 that LDLM_FL_CP_REQD is not set.
987 Description: Heavy nfs access might result in deadlocks
988 Details : After ELC code landed, it is now improper to enqueue any mds
989 locks under och_sem, because enqueue might want to decide to
990 cancel open locks for same inode we are holding och_sem for.
994 Description: Client eviction while running blogbench
995 Details : A lot of unlink operations with concurrent I/O can lead to a
996 deadlock causing evictions. To address the problem, the number of
997 oustanding OST_DESTROY requests is now throttled to
998 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
999 callbacks are processed in priority.
1003 Description: enable ACLs on MDS by default
1004 Details : ACLs must be enabled on MDS by default.
1007 Frequency : PPC/PPC64 only
1009 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1010 Details : fix duplicated definitions between asm-ppc64/types.h and
1011 lustre_types.h on PPC.
1014 Frequency : PPC/PPC64 only
1016 Description: asm-ppc/segment.h does not exist
1017 Details : fix compile issue on PPC.
1021 Description: better handle error messages in extents code
1024 Frequency : RHEL4 only
1026 Description: mkfs is very slow on IA64/RHEL4
1027 Details : A performance regression has been discovered in the MPT Fusion
1028 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1029 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1030 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1032 Severity : enhancement
1034 Description: SNMP support enhancement
1035 Details : Adding total number of sampled request for an MDS node in snmp
1038 Severity : enhancement
1040 Description: Optimize ldlm waiting list processing for PR extent locks
1041 Details : When processing waiting list for read extent lock and meeting read
1042 lock that is same or wider to it that is not contended, skip
1043 processing rest of the list and immediatelly return current
1044 status of conflictness, since we are guaranteed there are no
1045 conflicting locks in the rest of the list.
1049 Description: Time out and refuse to reconnect
1050 Details : When the failover node is the primary node, it is possible
1051 to have two identical connections in imp_conn_list. We must
1052 compare not conn's pointers but NIDs, otherwise we can defeat
1053 connection throttling.
1057 Description: port llog fixes from b1_6 into HEAD
1058 Details : Port llog reference couting and some llog cleanups from b1_6
1059 (bug 10800) into HEAD, for protect from panic and access to already
1060 free llog structures.
1064 Description: Detect stride IO mode in read-ahead
1065 Details : When a client does stride read, read-ahead should detect that and
1066 read-ahead pages according to the detected stride pattern.
1070 Description: data checksumming impacts single node performance
1071 Details : add support for several checksum algorithm. Currently, only CRC32
1072 and Adler-32 are supported. The checksum type can be changed on
1073 the fly via /proc/fs/lustre/osc/*/checksum_type.
1077 Description: use adler32 for page checksums
1078 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1083 Description: build for x2 fails
1084 Details : fix compile issue on Cray systems.
1088 Description: Properly match for duplicate locks
1089 Details : Due to different lock order from skiplists code, we need to
1090 traverse entire list for now
1093 Frequency : only on PPC/SLES10
1095 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1096 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1097 be defined. Add a hack in mkfs_lustre.c to work around this compile
1102 Description: LASSERT on MDS when client holding flock lock dies
1103 Details : ldlm pool logic depends on number of granted locks equal to
1104 number of released locks which is not true for flock locks, so
1105 just exclude such locks from consideration.
1109 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1110 Details : Use fsfilt_sync() for both the whole filesystem sync and
1111 individual file sync to eliminate dangerous inode locking
1112 with I_LOCK that can lead to a deadlock.
1116 Description: performance in 1.6.3
1117 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1118 sure that 1MB requests are not fragmented by the block layer.
1120 Severity : enhancement
1122 Description: organize the server-side client stats on per-nid basis
1123 Details : Change the structure of stats under obdfilter and mds to
1131 The "uuid"s file would list the uuids of _active_ exports.
1132 And the clear entry is to clear all stats and stale nids.
1134 Severity : enhancement
1136 Description: eliminate client locks in face of contention
1137 Details : file contention detection and lockless i/o implementation
1138 for contended files.
1142 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1145 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1147 Description: MDS may lose file striping (and hence file data) in some cases
1148 Details : If there are additional extended attributes stored on the MDS,
1149 in particular ACLs, SELinux, or user attributes (if user_xattr
1150 is specified for the client mount options) then there is a risk
1151 of attribute loss. Additionally, the Lustre file striping
1152 needs to be larger than default (e.g. striped over all OSTs),
1153 and an additional attribute must be stored initially in the
1154 inode and then increase in size enough to be moved to the
1155 external attribute block (e.g. ACL growing in size) for file
1160 Description: skiplist implementation simplification
1161 Details : skiplists are used to group compatible locks on granted list
1162 that was implemented as tracking first and last lock of each lock group
1163 the patch changes that to using doubly linked lists
1167 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1168 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1169 because LASSERT on both the data supplied by a client, and the data
1170 on disk is dangerous and incorrect.
1172 Severity : enhancement
1174 Description: Slow truncate/writes to huge files at high offsets.
1175 Details : Directly associate cached pages to lock that protect those pages,
1176 this allows us to quickly find what pages to write and remove
1177 once lock callback is received.
1181 Description: more ldlm soft lockups
1182 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1183 starve other threads from the resource lock for a long time in
1184 case of long waiting queue, so change the debug level from
1185 D_OTHER to the less frequently used D_INFO.
1187 Severity : enhancement
1189 Description: add -gid, -group, -uid, -user options to lfs find
1193 Description: Hung threads in invalidate_inode_pages2_range
1194 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1195 one is completed. As a result, some RPCs are stuck in the queue
1200 Description: filter threads hungs on waiting journal commit
1201 Details : Cleanup filter group llog code, then only filter group llog will
1202 be only created in the MDS/OST syncing process.
1206 Description: Procfs and llog threads access destoryed import sometimes.
1207 Details : Sync the import destoryed process with procfs and llog threads by
1208 the import refcount and semaphore.
1210 Severity : enhancement
1212 Description: openlock cache of b1_6 port to HEAD
1217 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1218 Details : If insertion of an extent fails, then discard the inode
1219 preallocation and free data blocks else it can lead to duplicate
1224 Description: don't always update ctime in ext3_xattr_set_handle()
1225 Details : Current xattr code updates the inode ctime in ext3_xattr_set_handle.
1226 In some cases the ctime should not be updated, for example for
1227 2.0->1.8 compatibility it is necessary to delete an xattr and it
1228 should not update the ctime.
1232 Bugzilla : 15713/16362
1233 Description: Assertion in iopen_connect_dentry in 1.6.3
1234 Details : looking up an inode via iopen with the wrong generation number can
1235 populate the dcache with a disconneced dentry while the inode
1236 number is in the process of being reallocated. This causes an
1237 assertion failure in iopen since the inode's dentry list contains
1238 both a connected and disconnected dentry.
1242 Description: assertion failure in ldlm_handle2lock()
1243 Details : fix a race between class_handle_unhash() and class_handle2object()
1244 introduced in lustre 1.6.5 by bug 13622.
1249 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1250 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1255 Description: oops in sd_iostats_seq_show()
1256 Details : unloading/reloading the scsi low level driver triggers a kernel
1257 bug when trying to access the sd iostat file.
1262 Description: Kernel panics during QLogic driver reload
1263 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1264 patch, causing memory corruption.
1269 Description: journal_dev option does not work in b1_6
1270 Details : pass mount option during pre-mount.
1272 --------------------------------------------------------------------------------
1274 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1276 * Support for kernels:
1277 2.6.5-7.283 (SLES 9),
1278 2.6.9-55.EL (RHEL 4),
1279 2.6.16.46-0.14 (SLES 10),
1280 2.6.18.8 vanilla (kernel.org)
1281 * Client support for unpatched kernels:
1282 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1283 2.6.16 - 2.6.22 vanilla (kernel.org)
1284 * Due to recently discovered recovery problems, we do not recommend
1285 using patchless RHEL 4 clients with this or any earlier release.
1286 * Recommended e2fsprogs version: 1.39.cfs8
1287 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1288 * Starting with this release, the ldiskfs backing filesystem required
1289 by Lustre is now in its own package, lustre-ldiskfs. This package
1290 should be installed. It is versioned separately from Lustre and
1291 may be released separately in future.
1295 Description: block reactivating mgc import until all deactivates complete
1296 Details : Fix race when failing back MDT/MGS to itself (testing)
1299 Frequency : at statup only
1301 Description: mds_lov_synchronize race leads to various problems
1302 Details : simultaneous MDT->OST connections at startup can cause the
1303 sync to abort, leaving the OSC in a bad state.
1305 Severity : enhancement
1307 Description: add optional extra BUILD_VERSION info
1308 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1309 to override the lustre version.
1312 Frequency : 2.6.18 servers only
1314 Description: ll_kern_mount() doesn't release the module reference
1315 Details : The ldiskfs module reference count never drops down to 0
1316 because ll_kern_mount() doesn't release the module reference.
1321 Description: server LBUG when using old ost_num_threads parameter
1322 Details : Accept the old ost_num_threads parameter but warn that it
1323 is deprecated, and fix an off-by-one error that caused an LBUG.
1328 Description: Transient SCSI error results in persistent IO issue
1329 Details : iobuf->dr_error is not reinitialized to 0 between two
1333 Frequency : sometimes when underlying device returns I/O errors
1335 Description: OSTs not going read-only during write failures
1336 Details : OSTs are not remounted read-only when the journal commit threads
1337 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1338 instead of the ext3 wrappers.
1342 Description: poor connect scaling with increasing client count
1343 Details : Don't run filter_grant_sanity_check for more than 100 exports
1344 to improve scaling for large numbers of clients.
1347 Frequency : SLES10 only
1349 Description: sanity-quota.sh quotacheck failed: rc = -22
1350 Details : Quotas cannot be enabled on SLES10.
1353 Frequency : liblustre clients only
1355 Description: getdirentries does not give error when run on compute nodes
1356 Details : getdirentries does not fail when the size specified as an argument
1357 is too small to contain at least one entry
1359 Severity : enhancement
1361 Description: Add LNET router traceability for debug purposes
1362 Details : If a checksum failure occurs with a router as part of the
1363 IO path, the NID of the last router that forwarded the bulk data
1364 is printed so it can be identified.
1369 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1370 Details : Due to a race condition, liblustre clients were occasionally
1371 evicted incorrectly.
1373 Severity : enhancement
1375 Description: lfs setstripe use optional parameters instead of postional
1378 Severity : enhancement
1380 Description: Nanosecond timestamp support for ldiskfs
1381 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1382 resolution timestamps. There is not yet support for this at
1383 the Lustre filesystem level.
1386 Frequency : during server recovery
1388 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1389 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1393 Frequency : nfs export on patchless client
1395 Description: connectathon hang when test nfs export over patchless client
1396 Details : Disconnected dentry cannot be found with lookup, so we do not need
1397 to unhash it or make it invalid
1400 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1401 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1402 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1407 Description: open req refcounting wrong on reconnect
1408 Details : If reconnect happened between getting open reply from server and
1409 call to mdc_set_replay_data in ll_file_open, we will schedule
1410 replay for unreferenced request that we are about to free.
1411 Subsequent close will crash in variety of ways.
1412 Check that request is still eligible for replay in
1413 mdc_set_replay_data().
1418 Description: disable writes to filesystem when reading health_check file
1419 Details : the default for reading the health_check proc file has changed
1420 to NOT do a journal transaction and write to disk, because this
1421 can cause reads of the /proc file to hang and block HA state
1422 checking on a healthy but otherwise heavily loaded system. It
1423 is possible to return to the previous behaviour during configure
1424 with --enable-health-write.
1426 Severity : enhancement
1428 Description: 64-bit inode version
1429 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1430 to the inode. This will be required for version-based recovery.
1435 Description: MDS fails to start if a duplicate client export is detected
1436 Details : in some rare cases it was possible for a client to connect to
1437 an MDS multiple times. Upon recovery the MDS would detect this
1438 and fail during startup. Handle this more gracefully.
1440 Severity : enhancement
1442 Description: Add -o localflock option to simulate old noflock
1444 Details : This will achieve local-only flock/fcntl locks
1450 Description: log_commit_thread vs filter_destroy race leads to crash
1451 Details : Take import reference before releasing llog record semaphore
1456 Description: Wrong request locking in request set processing
1457 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1458 imp_delayed_list, in this place should be used imp_lock.
1461 Frequency : when reconnection
1463 Description: Grant Leak when osc reconnect to OST
1464 Details : When osc reconnect ost, OST(filter) should check whether it
1465 should grant more space to client by comparing fed_grant and
1466 cl_avail_grant, and return the granted space to client instead
1467 of "new granted" space, because client will call osc_init_grant
1468 to update the client grant space info.
1471 Frequency : when client reconnect to OST
1473 Description: Grant Leak when osc do resend and replay bulk write
1474 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1475 bulk write request, because the grant info while be sync between
1476 OSC and OST when reconnect, and we should ignore the grant info
1477 these of resend/replay write req.
1482 Description: Grant space more than avaiable left space sometimes.
1483 Details : When then OST is about to be full, if two bulk writing from
1484 different clients came to OST. Accord the avaliable space of the
1485 OST, the first req should be permitted, and the second one
1486 should be denied by ENOSPC. But if the seconde arrived before
1487 the first one is commited. The OST might wrongly permit second
1488 writing, which will cause grant space > avaiable space.
1491 Frequency : when client is evicted
1493 Description: Grant might be wrongly erased when osc is evicted by OST
1494 Details : when the import is evicted by server, it will fork another
1495 thread ptlrpc_invalidate_import_thread to invalidate the
1496 import, where the grant will be set to 0. While the original
1497 thread will update the grant it got when connecting. So if
1498 the former happened latter, the grant will be wrongly errased
1499 because of this race.
1504 Description: Checking Stale with correct fid
1505 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1506 to check whether it is stale, because sometimes, we want the
1507 enqueue happened anyway, and op_data.fid2 will not be initialized.
1509 Severity : enhancement
1511 Description: update patchless client
1512 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1515 Frequency : only with 2.4 kernel
1517 Description: random memory corruption
1518 Details : size of struct ll_inode_info is to big for union inode.u and this
1519 can be cause of random memory corruption.
1524 Description: Memory leak in recovery
1525 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1526 It should also check obd_fail before fsfilt_start, otherwise if
1527 fsfilt_start return -EROFS,(failover mds during mds recovery).
1528 then the req will return with repmsg->transno = 0 and rc = EROFS.
1529 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1530 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1531 be freed no matter whether fsfilt_commit success or not.
1534 Frequency : only with huge count clients
1536 Description: Prevents from taking the superblock lock in llap_from_page for
1538 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1539 for llap_from_page call in ll_removepage prevents from taking the
1540 superblock lock for a soon died page.
1545 Description: Not check open intent error before release open handle
1546 Details : in some rare cases, the open intent error is not checked before
1547 release open handle, which may cause
1548 ASSERTION(open_req->rq_transno != 0), because it tries to release
1549 the failed open handle.
1554 Description: Set cat log bitmap only after create log success.
1555 Details : in some rare cases, the cat log bitmap is set too early. and it
1556 should be set only after create log success.
1560 Description: Accessing a block bevice can re-enable I/O when Lustre is
1561 tearing down a device.
1562 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1567 Description: service threads may hog cpus when there are a lot of requests
1568 Details : Insert cond_resched to give other threads a chance to use some CPU
1573 Description: the cat log was not initialized in recovery
1574 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1575 unlink log on mds will not be initialized until mds post
1576 recovery. And also in mds post recovery, the unlink log will
1577 initialization will be done asynchronausly, so there will be race
1578 between add unlink log and unlink log initialization.
1582 Description: brw_stats were being printed incorrectly
1583 Details : brw_stats were being printed as log2 but all of them were not
1584 recorded as log2. Also remove some code duplication arising from
1585 filter_tally_{read,write}.
1589 Frequency : rare, only in recovery.
1590 Description: ASSERTION(req->rq_type != LI_POISON) failed
1591 Details : imp_lock should be held while iterating over imp_sending_list for
1592 prevent destroy request after get timeout in ptlrpc_queue_wait.
1596 Description: replay-single.sh test 52 fails
1597 Details : A lock's skiplist need to be cleanup when it being unlinked
1598 from its resource list.
1602 Description: Short directio read returns full requested size rather than
1604 Details : Direct I/O operations should return actual amount of bytes
1605 transferred rather than requested size.
1607 Severity : enhancement
1609 Description: metadata RPC reduction (e.g. for rm performance)
1610 Details : decrease the amount of synchronous RPC between clients and servers
1611 by canceling conflicing lock before the operation on the client
1612 and packing thier handles into the main operation RPC to server.
1614 Severity : enhancement
1616 Description: Async OSC create to avoid the blocking unnecessarily.
1617 Details : If a OST has no remain object, system will block on the creating
1618 when need to create a new object on this OST. Now, ways use
1619 pre-created objects when available, instead of blocking on an
1620 empty osc while others are not empty. If we must block, we block
1621 for the shortest possible period of time.
1623 Severity : enhancement
1625 Description: refine locking for avoid write wrong info into lov_objid file
1626 Details : fix possible races with add new target and write/update data in
1629 --------------------------------------------------------------------------------
1631 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1636 Frequency : on some architectures
1638 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1639 Details : Uninitialized flags sometimes cause configuration commands to
1643 Frequency : patchless clients only
1645 Description: missing __iget() symbol export
1646 Details : The __iget() symbol export is missing. To avoid the need for
1647 this on patchless clients the deathrow inode reaper is turned
1648 off, and we depend on the VM to clean up old inodes. This
1649 dependency was during via the fix for bug 12181.
1651 --------------------------------------------------------------------------------
1653 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1655 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1656 INTEROPERATE with older versions automatically. In many cases a
1657 special upgrade step is needed. Please read the
1658 user documentation before upgrading any part of a live system.
1659 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1660 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1661 * WARNING: Lustre configuration and startup changes are required with
1662 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1668 Severity : enhancement
1670 Description: MountConf
1671 Details : Lustre configuration is now managed via mkfs and mount
1672 commands instead of lmc and lconf. New obd types (MGS, MGC)
1673 are added for dynamic configuration management. See
1674 https://mail.clusterfs.com/wikis/lustre/MountConf for
1677 Severity : enhancement
1679 Description: dynamic OST addition
1680 Details : OSTs can now be added to a live filesystem
1682 Severity : enhancement
1684 Description: startup order invariance
1685 Details : MDTs and OSTs can be started in any order. Clients only
1686 require the MDT to complete startup.
1688 Severity : enhancement
1690 Description: parallel, asynchronous orphan cleanup
1691 Details : orphan cleanup is now performed in separate threads for each
1692 OST, allowing parallel non-blocking operation.
1694 Severity : enhancement
1696 Description: optimized stripe assignment
1697 Details : stripe assignments are now made based on ost space available,
1698 ost previous usage, and OSS previous usage, in order to try
1699 to optimize storage space and networking resources.
1701 Severity : enhancement
1703 Description: Permanently set tunables
1704 Details : All writable /proc/fs/lustre tunables can now be permanently
1705 set on a per-server basis, at mkfs time or on a live system.
1707 Severity : enhancement
1709 Description: Lustre message v2
1710 Details : Add lustre message format v2.
1712 Severity : enhancement
1714 Description: client OST exclusion list
1715 Details : Clients can be started with a list of OSTs that should be
1716 declared "inactive" for known non-responsive OSTs.
1720 Description: SPEC SFS validation failure on NFS v2 over lustre.
1721 Details : Changes the blocksize for regular files to be 2x RPC size,
1722 and not depend on stripe size.
1724 Severity : enhancement
1726 Description: Multiple MD RPCs in flight.
1727 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1728 To avoid overly-overloading MDS, introduce a limit on number of
1729 MDS RPCs in flight for a single client and add /proc controls
1730 to adjust this limit.
1732 Severity : enhancement
1734 Description: client read/write statistics
1735 Details : Add client read/write call usage stats for performance
1736 analysis of user processes.
1737 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1738 file access. extents_stats shows chunk size distribution.
1739 extents_stats_per_process show chunk size distribution per
1742 Severity : enhancement
1744 Description: mds statistics
1745 Details : Add detailed mds operations statistics in
1746 /proc/fs/lustre/mds/*/stats.
1750 Description: Failure of copying files with lustre special EAs.
1751 Details : Client side always return success for setxattr call for lustre
1752 special xattr (currently only "trusted.lov").
1757 Description: Refcount LNET uuids
1758 Details : The global LNET uuid list grew linearly with every startup;
1759 refcount repeated list entries instead of always adding to
1762 Severity : enhancement
1764 Description: Dynamic service threads
1765 Details : Within a small range, start extra service threads
1766 automatically when the request queue builds up.
1769 Frequency : mixed-endian client/server environments
1771 Description: mixed-endian crashes
1772 Details : The new msg_v2 system had some failures in mixed-endian
1775 Severity : enhancement
1777 Description: Easy OST removal
1778 Details : OSTs can be permanently deactivated with e.g. 'lctl
1779 conf_param lustre-OST0001.osc.active=0'
1781 Severity : enhancement
1783 Description: MGS proc entries
1784 Details : Added basic proc entries for the MGS showing what filesystems
1787 Severity : enhancement
1789 Description: provide MGS failover
1790 Details : Added config lock reacquisition after MGS server failover.
1792 Severity : enhancement
1794 Description: add Linux 2.4 support
1795 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1799 Description: a large application tries to do I/O to the same resource and dies
1800 in the middle of it.
1801 Details : Check the req->rq_arrival time after the call to
1802 ost_brw_lock_get(), but before we do anything about
1803 processing it & sending the BULK transfer request. This
1804 should help move old stale pending locks off the queue as
1805 quickly as obd_timeout.
1808 Frequency : when an incorrect nid is specified during startup
1810 Description: ptlrpc connect to non-existant node causes kernel crash
1811 Details : LNET can't be re-entered from an event callback, which
1812 happened when we expire a message after the export has been
1813 cleaned up. Instead, hand the zombie cleanup off to another
1816 Severity : enhancement
1818 Description: plain/inodebits lock performance improvement
1819 Details : Grouping plain/inodebits in granted list by their request modes
1820 and bits policy, thus improving the performance of search through
1824 Frequency : only if OST filesystem is corrupted
1826 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1827 Details : for a short time RPCs with bulk IO are in the replay list,
1828 but replay of bulk IOs is unimplemented. If the OST filesystem
1829 is corrupted due to disk cache incoherency and then replay is
1830 started it is possible to trip an assertion. Avoid putting
1831 committed RPCs into the replay list at all to avoid this issue.
1834 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1836 Description: System hang on startup
1837 Details : This bug allowed the liblustre (e.g. catamount) client to
1838 return to the app before handling all startup RPCs. This
1839 could leave the node unresponsive to lustre network traffic
1840 and manifested as a server ptllnd timeout.
1842 Severity : enhancement
1844 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1845 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1846 boolean to control whether to print peer debug info when a
1847 client's RPC times out.
1850 Frequency : only for kernels with patches from Lustre below 1.4.3
1852 Description: Remove old rdonly API
1853 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1856 Frequency : only for devices with external journals
1858 Description: Set external device read-only also
1859 Details : During a commanded failover stop, we set the disk device
1860 read-only while the server shuts down. We now also set any
1861 external journal device read-only at the same time.
1864 Frequency : when upgrading from 1.4 while trying to change parameters
1866 Description: The wrong (new) MDC name was used when setting parameters for
1867 upgraded MDT's. Also allows changing of OSC (and MDC)
1868 parameters if --writeconf is specified at tunefs upgrade time.
1871 Frequency : when setting specific ost indicies
1873 Description: QOS code breaks on skipped indicies
1874 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1875 created with --index need not be sequential.
1877 Severity : enhancement
1879 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1880 Details : The uninit_groups feature works in conjunction with the kernel
1881 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1882 up the pass1 processing of e2fsck. This is a read-only feature
1883 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1884 mount filesystems that have had this feature enabled.
1886 Severity : enhancement
1888 Description: Improve multi-block allocation algorithm to avoid fragmentation
1889 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1890 allocation locality and avoid filesystem fragmentation.
1892 ------------------------------------------------------------------------------
1894 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1896 * Support for kernels:
1897 2.4.21-47.0.1.EL (RHEL 3)
1898 2.6.5-7.283 (SLES 9)
1899 2.6.9-42.0.10.EL (RHEL 4)
1900 2.6.12.6 vanilla (kernel.org)
1901 2.6.16.27-0.9 (SLES 10)
1902 * Recommended e2fsprogs version: 1.39.cfs5
1904 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1908 Frequency : occasional, depends on client load and configuration
1909 Bugzilla : 12181, 12203
1910 Description: data loss for recently-modified files
1912 Details : In some cases it is possible that recently written or created
1913 files may not be written to disk in a timely manner (this should
1914 normally be within 30s unless client IO load is very high).
1915 The problem appears as zero-length files or files that are a
1916 multiple of 1MB in size after a client crash or client eviction
1917 that are missing data at the end of the file.
1919 This problem is more likely to be hit on clients where files are
1920 repeatedly created and unlinked in the same directory, clients
1921 have a large amount of RAM, have many CPUs, the filesystem has
1922 many OSTs, the clients are rebooted frequently, and/or the files
1923 are not accessed by other nodes after being written.
1925 The presence of the problem can be detected by looking at
1926 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1927 smaller than the second (nr_unused) then dirty files will not
1928 be flushed automatically to disk. "sync; sleep 10" should be
1929 run several times on the node before unmounting it to update
1930 Lustre (this is also safe to run on nodes without this problem).
1932 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1933 that can cause this same problem, so customers using that kernel
1934 also need to update the kernel in addition to Lustre. In order
1935 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1937 It is normal that files written just before a client crash (less
1938 than 30s) may not yet have been flushed to disk, even for local
1942 Frequency : frequent on thin XT3 nodes
1944 Description: UUID collision on thin XT3 Linux nodes
1945 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1946 randomly, since we relied on an insufficiently-seeded PRNG.
1951 Description: OSS hangs after "All ost request buffers busy"
1952 Details : A deadlock between quota and journal operations caused OSS
1953 hangs after printing "All ost request buffers busy."
1956 Frequency : always on liblustre builds
1958 Description: Cleanup compiler warnings on liblustre
1961 Frequency : always on liblustre builds on XT3
1963 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1967 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1968 > 32000 subdirectories
1969 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1970 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1971 subdirectory count crosses 32000. This will aid e2fsck to
1972 correctly handle more than 32000 subdirectories.
1975 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1977 Description: System hang on startup
1978 Details : This bug allowed the liblustre (e.g. catamount) client to
1979 return to the app before handling all startup RPCs. This
1980 could leave the node unresponsive to lustre network traffic
1981 and manifested as a server ptllnd timeout.
1983 Severity : enhancement
1985 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1986 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1987 boolean to control whether to print peer debug info when a
1988 client's RPC times out.
1993 Description: make O_SYNC working on 2.6 kernels
1994 Details : 2.6 kernels use different method for mark pages for write,
1995 so need add a code to lustre for O_SYNC work.
2000 Description: Failure to close file and release space on NFS
2001 Details : Put inode details into lock acquired in ll_intent_file_open.
2002 Use mdc_intent_lock in ll_intent_open to properly
2003 detect all kind of errors unhandled by mdc_enqueue.
2008 Description: proc file read during shutdown sometimes raced obd removal,
2010 Details : Add lock to prevent obd access after proc file removal.
2013 Frequency : Only for files larger than 4GB on 32-bit clients.
2015 Description: improperly doing page alignment of locks
2016 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2017 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2022 Description: RPCs being resent when they shouldn't be
2023 Details : Some RPCs that should not be resent are being resent. This
2024 can cause inconsistencies in the RPC state machine. Do not
2025 resend such requests.
2028 Frequency : rare, only with NFS export
2030 Description: Crash on NFS re-export node
2031 Details : under very unusual load conditions an assertion is hit in
2032 ll_intent_file_open()
2035 Frequency : only if OST filesystem is corrupted
2037 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2038 Details : for a short time RPCs with bulk IO are in the replay list,
2039 but replay of bulk IOs is unimplemented. If the OST filesystem
2040 is corrupted due to disk cache incoherency and then replay is
2041 started it is possible to trip an assertion. Avoid putting
2042 committed RPCs into the replay list at all to avoid this issue.
2047 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2048 Details : Large single O_DIRECT read and write calls can fail to allocate
2049 a sufficiently large buffer to process the request. In case of
2050 allocation failure the allocation is retried with a smaller
2051 buffer and broken into smaller requests.
2053 Severity : enhancement
2055 Description: Add -o localflock option to simulate old noflock behaviour.
2056 Details : This will achieve local-only flock/fcntl locks coherentness.
2061 Description: versioning check is incomplete
2062 Details : Checking the version difference of client vs. server, report
2063 error if the gap is too big.
2068 Description: add support PG_writeback bit
2069 Details : add support for PG_writeback bit for Lustre, for more carefull
2070 work with page cache in 2.6 kernel. This also fix some deadlocks
2071 and remove hack for work O_SYNC with 2.6 kernel.
2073 Severity : enhancement
2075 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2076 Details : The uninit_groups feature works in conjunction with the kernel
2077 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2078 up the pass1 processing of e2fsck. This is a read-only feature
2079 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2080 mount filesystems that have had this feature enabled.
2082 Severity : enhancement
2084 Description: Improve multi-block allocation algorithm to avoid fragmentation
2085 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2086 allocation locality and avoid filesystem fragmentation.
2088 ------------------------------------------------------------------------------
2090 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2092 * Support for kernels:
2093 2.6.9-42.0.3.EL (RHEL 4)
2094 2.6.5-7.276 (SLES 9)
2095 2.4.21-47.0.1.EL (RHEL 3)
2096 2.6.12.6 vanilla (kernel.org)
2097 2.6.16.21-0.8 (SLES10)
2098 * Recommended e2fsprogs version: 1.39.cfs2-0
2100 * The backwards-compatible /proc/sys/portals symlink has been removed
2101 in this release. Before upgrading, please ensure that you change
2102 any configuration scripts or /etc/sysctl.conf files that access
2103 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2104 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2105 in advance of the upgrade on any system running Lustre 1.4.6 or
2106 newer, since /proc/sys/lnet was added in that version.
2107 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2111 Frequency : only when quota is used
2113 Description: avoid scanning export list for quota master
2114 Details : Change the algorithms to avoid scanning export list in order
2115 to improve the efficiency.
2118 Frequency : MDS failover only, very rarely
2120 Description: "went back in time" messages on mds failover
2121 Details : The greatest transno may be lost when the current operation
2122 finishes with an error (transno==0) and the client's last_rcvd
2123 record is over-written. Save the greatest transno in the
2124 mds_last_transno for this case.
2127 Frequency : always for specific kernels and striping counts
2129 Description: client may get "Matching packet too big" without ACL support
2130 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2131 when trying to access files in certain configurations. The
2132 clients should in fact be denied when mounting because they do
2133 not understand ACLs.
2136 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2138 Description: many clients connecting with IO in progress causes connect timeouts
2139 Details : Avoid synchronous journal commits to avoid delays caused by many
2140 clients connecting/disconnecting when bulk IO is in progress.
2141 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2142 OST_IO_PORTAL to avoid delays behind potentially many pending
2146 Frequency : occasionally with multiple writers to a single file
2148 Description: shared writes to file may result in wrong size reported by stat()
2149 Details : Allow growing of kms when extent lock is cancelled
2152 Frequency : always with random mmap IO to multi-striped file
2154 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2155 Details : Only if the hole is at the end of OST object so that kms is too
2156 small. Fix is to increase kms accordingly in ll_nopage.
2159 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2161 Description: writes to a missing object would leak memory on the OST
2162 Details : If there is an inconsistency between the MDS and OST filesystems,
2163 such that the MDS references an object that doesn't exist, writes
2164 to that object will leak memory due to incorrect cleanup in the
2165 error handling path, eventually running out of memory on the OST.
2170 Description: Creating too long symlink causes lustre errors
2171 Details : Check symlink and name lengths before sending requests to MDS.
2174 Frequency : only if flock is enabled (not on by default)
2176 Description: posix locks not released on fd closure on 2.6.9+
2177 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2178 kernels, this caused such locks not to be released on fd close and
2179 then assertions on fs unmount about still used locks.
2182 Frequency : MDS failover only, very rarely
2184 Description: clients may get ASSERTION(granted_lock != NULL)
2185 Details : When request was taking a long time, and a client was resending
2186 a getattr by name lock request. The were multiple lock requests
2187 with the same client lock handle and
2188 mds_getattr_name->fixup_handle_for_resent_request found one of the
2189 lock handles but later failed with ASSERTION(granted_lock != NULL).
2194 Description: handle->h_buffer_credits > 0, assertion failure
2195 Details : h_buffer_credits is zero after truncate, causing assertion
2196 failure. This patch extends the transaction or creates a new
2200 Frequency : NFS re-export or patchless client
2201 Bugzilla : 11179, 10796
2202 Description: Crash on NFS re-export node (__d_move)
2203 Details : We do not want to hash the dentry if we don't have a lock.
2204 But if this dentry is later used in d_move, we'd hit uninitialised
2205 list head d_hash, so we just do this to init d_hash field but
2206 leave dentry unhashed.
2209 Frequency : NFS re-export or patchless client
2211 Description: NFS exports has problem with symbolic link
2212 Details : lustre client didn't properly install dentry when re-exported
2213 to NFS or running patchless client.
2216 Frequency : NFS re-export or patchless client
2218 Description: Various nfs/patchless fixes.
2219 Details : fixes reuse disconected alias for lookup process - this fixes
2220 warning "find_exported_dentry: npd != pd",
2221 fix permission error with open files at nfs.
2222 fix apply umask when do revalidate.
2225 Frequency : occasional
2227 Description: Crash on NFS re-export node
2228 Details : calling clear_page() on the wrong pointer triggered oops in
2229 generic_mapping_read().
2232 Frequency : rarely, using O_DIRECT IO
2234 Description: unaligned directio crashes client with LASSERT
2235 Details : check for unaligned buffers before trying any requests.
2238 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2240 Description: stale data returned from RAID cache
2241 Details : If only a small amount of IO is done to the RAID device before
2242 reading it again it is possible to get stale data from the RAID
2243 cache instead of reading it from disk.
2246 Frequency : always for sles10 kernel
2248 Description: sles10 support
2249 Details : ll_follow_link: compile fixes and using of nd_set_link
2250 under newer kernels.
2253 Frequency : depends on arch, kernel and compiler version, always on sles10
2256 Description: recursive or deep enough symlinks cause stack overflow
2257 Details : getting rid of large stack-allocated variable in
2261 Frequency : depends on hardware
2263 Description: lustre write performance loss in the SLES10 kernel
2264 Details : the performance loss is caused by using of write barriers in the
2265 ext3 code. The SLES10 kernel turns barrier support on by
2266 default. The fix is to undo that change for ldiskfs.
2268 ------------------------------------------------------------------------------
2270 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2272 * Support for kernels:
2273 2.6.9-42.0.3EL (RHEL 4)
2274 2.6.5-7.276 (SLES 9)
2275 2.4.21-47.0.1.EL (RHEL 3)
2276 2.6.12.6 vanilla (kernel.org)
2280 Frequency : quota enabled and large files being deleted
2282 Description: releasing more than 4GB of quota at once hangs OST
2283 Details : If a user deletes more than 4GB of files on a single OST it
2284 will cause the OST to spin in an infinite loop. Release
2285 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2290 Description: statfs data retrieved from /proc may be stale or zero
2291 Details : When reading per-device statfs data from /proc, in the
2292 {kbytes,files}_{total,free,avail} files, it may appear
2293 as zero or be out of date.
2296 Frequency : systems with MD RAID1 external journal devices
2298 Description: lconf's call to blkid is confused by RAID1 journal devices
2299 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2300 of returning all block devices that match the journal UUID.
2303 Frequency : always, for aggregate stripe size over 4GB
2305 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2306 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2307 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2310 Frequency : always if "lfs find" used on a local file/directory
2312 Description: "lfs find" segfaults if used on a local file/directory
2313 Details : The case where a directory component was not specified wasn't
2314 handled correctly. Handle this properly.
2317 Frequency : always on ppc64
2319 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2320 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2321 perform endianness conversion.
2324 Frequency : rarely (truncate to non-zero file size after write under load)
2325 Bugzilla : 10730, 10687
2326 Description: Files padded with zeros to next 4K multiple
2327 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2328 it is possible that files that are truncated to a non-zero size
2329 immediately after being written are filled with zero bytes beyond
2330 the truncated size. No file data is lost.
2332 Severity : enhancement
2334 Description: Allow recovery/failover for liblustre clients.
2335 Details : liblustre clients were unaware of failover configurations until
2338 Severity : enhancement
2340 Description: user file locks should fail when not mounting with flock option
2341 Details : Set up an error-returning stub in ll_file_operations.lock field
2342 to prevent incorrect behaviour when client is mounted without
2343 flock option. Also, set up properly f_op->flock field for
2347 Frequency : always on ia64
2349 Description: "lfs df" loops on printing out MDS statfs information
2350 Details : The obd_ioctl_data was not initialized and in some systems
2351 this caused a failure during the ioctl that did not return
2352 an error. Initialize the struct and return an error on failure.
2355 Frequency : SLES 9 only
2357 Description: Error of copying files with lustre special EAs as root
2358 Details : Client side always return success for setxattr call for lustre
2359 special xattr (currently only "trusted.lov").
2362 Frequency : rarely on clusters with both ia64+i386 clients
2364 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2365 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2366 clients are doing concurrent writes to the same file it is
2367 possible that the ia64 clients may LASSERT because the OST
2368 extent locks are not PAGE_SIZE aligned. Ensure that grown
2369 locks are always aligned on the request boundary.
2372 Frequency : specific use, occasional
2374 Description: Overwriting in use executable truncates on-disk binary image
2375 Details : If one node attempts to overwrite an executable in use by
2376 another node, we now correctly return ETXTBSY instead of
2377 truncating the file.
2379 Severity : enhancement
2381 Description: Async OSC create to avoid the blocking unnecessarily.
2382 Details : If a OST has no remain object, system will block on the creating
2383 when need to create a new object on this OST. Now, ways use
2384 pre-created objects when available, instead of blocking on an
2385 empty osc while others are not empty. If we must block, we block
2386 for the shortest possible period of time.
2391 Description: chmod on Lustre root is propagated to other clients
2392 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2393 invalid by the follow_mount time.
2396 Frequency : liblustre clients only
2398 Description: Race in 'instant cancel' lock handling could lead to such locks
2399 never to be granted in case of SMP MDS
2400 Details : Do not destroy not yet granted but cbpending locks in
2404 Frequency : replay/resend of open
2406 Description: non null lock assetion failure in mds_intent_policy
2407 Details : Trying to replay/resend lockless open requests resulted in
2408 mds_open() returning 0 with no lock. Now it sets a flag if
2409 a lock is going to be returned.
2411 Severity : enhancement
2413 Description: Checksum enhancements
2414 Details : New checksum enhancements allow for resending RPCs that failed
2417 Severity : enhancement
2419 Description: Tunables on number of dirty pages in cacche
2420 Details : Allow to set limit on number of dirty pages cached.
2425 Description: client crash on unmount - lock still has references
2426 Details : In some error handling cases it was possible to leak a lock
2427 reference on a client while accessing a file. This was not
2428 harmful to the client during operation, but would cause the
2429 client to crash when the filesystem is unmounted.
2432 Frequency : specific case, rare
2434 Description: ETXTBSY on mds though file not in use by client
2435 Details : ETXTBSY is no longer incorrectly returned when attempting to
2436 chmod or chown a directory that the user previously tried to
2437 execute or a currently-executing binary.
2440 Frequency : extremely rare except on liblustre-based clients
2442 Description: Lustre space not freed when files are deleted
2443 Details : Clean up open-unlinked files after client eviction. Previously
2444 the unlink was skipped and the files remained as orphans.
2449 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2450 Details : In some cases it was possible to send a blocking callback to
2451 a client doing a glimpse, even though that client didn't get
2452 a lock granted. When the glimpse lock is cancelled on the OST
2453 the freed lock is left on the waiting list and corrupted the list.
2456 Frequency : all core dumps
2458 Description: Broke core dumps to lustre
2459 Details : Negative dentry may be unhashed if parent does not have UPDATE
2460 lock, but some callers, e.g. do_coredump, expect dentry to be
2461 hashed after successful create, hash it in ll_create_it.
2463 ------------------------------------------------------------------------------
2465 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2467 * Support for kernels:
2468 2.6.9-42.0.2.EL (RHEL 4)
2469 2.6.5-7.276 (SLES 9)
2470 2.4.21-40.EL (RHEL 3)
2471 2.6.12.6 vanilla (kernel.org)
2475 Frequency : always on RHEL 3
2477 Description: Number of open files grows over time
2478 Details : The number of open files grows over time, whether or not
2479 Lustre is started. This was due to a filp leak introduced
2480 by one of our kernel patches.
2482 ------------------------------------------------------------------------------
2484 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2486 * Support for kernels:
2487 2.6.9-42.EL (RHEL 4)
2488 2.6.5-7.276 (SLES 9)
2489 2.4.21-40.EL (RHEL 3)
2490 2.6.12.6 vanilla (kernel.org)
2495 Bugzilla : 5719, 9635, 9792, 9684
2496 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2497 Details : If a server is under heavy load and cannot reply to new
2498 connection requests before the client resends the (re)connect,
2499 the connection handling code can behave badly if two service
2500 threads are concurrently handing separate (re)connections from
2501 the same client. Add better locking to the connection handling
2502 code, and ensure that only a single connection will be processed
2503 for a given client UUID, even if the lock is dropped.
2505 Severity : enhancement
2507 Description: add TCP zero-copy support to kernel
2508 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2509 sends if the hardware supports scatter-gather and checksumming.
2510 This allows socklnd to do client-write and server-read more
2511 efficiently and reduce CPU utilization from skbuf copying.
2514 Frequency : only if NFS exporting from client
2516 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2517 Details : If setting attributes on a file created under NFS that had
2518 never been opened it would be possible to oops the client
2519 if the file had no objects.
2522 Frequency : always for liblustre
2524 Description: liblustre client does MDS+OSTs setattr RPC for each write
2525 Details : When doing a write from a liblustre client, the client
2526 incorrectly issued an RPC to the MDS and each OST the file was
2527 striped over in order to update the timestamps. When writing
2528 with small chunks and many clients this could overwhelm the MDS
2529 with RPCs. In all cases it would slow down the write because
2530 these RPCs are unnecessary.
2532 Severity : enhancement
2534 Description: allow number of MDS service threads to be changed at module load
2535 Details : It is now possible to change the number of MDS service threads
2536 running. Adding "options mds mds_num_threads={N}" to the MDS's
2537 /etc/modprobe.conf will set the number of threads for the next
2538 time Lustre is restarted (assuming the "mds" module is also
2539 reloaded at that time). The default number of threads will
2540 stay the same, 32 for most systems.
2545 Description: OST crash if filesystem is unformatted or corrupt
2546 Details : If an OST is started on a device that has never been formatted
2547 or if the filesystem is corrupt and cannot even mount then the
2548 error handling cleanup routines would dereference a NULL pointer.
2553 Description: NULL pointer deref in llap_from_page.
2554 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2555 incorrect) mapping if the page was truncated/reclaimed while it was
2556 searched for. Check for this condition and skip such pages when
2557 doing readahead. Introduce extra check to llap_from_page() to
2558 verify page->mapping->host is non-NULL (so page is not anonymous).
2561 Frequency : Sometimes when using sys_sendfile
2563 Description: "page not covered by a lock" warnings from ll_readpage
2564 Details : sendfile called ll_readpage without right page locks present.
2565 Now we introduced ll_file_sendfile that does necessary locking
2566 around call to generic_file_sendfile() much like we do in
2570 Frequency : with certain MDS communication failures at client mount time
2572 Description: NULL pointer deref after failed client mount
2573 Details : a client connection request may delayed by the network layer
2574 and not be sent until after the PTLRPC layer has timed out the
2575 request. If the client fails the mount immediately it will try
2576 to clean up before the network times out the request. Add a
2577 reference from the request import to the obd device and delay
2578 the cleanup until the network drops the request.
2581 Frequency : occasionally during client (re)connect
2583 Description: assertion failure during client (re)connect
2584 Details : processing a client connection request may be delayed by the
2585 client or server longer than the client connect timeout. This
2586 causes the client to resend the connection request. If the
2587 original connection request is replied in this interval, the
2588 client may trip an assertion failure in ptlrpc_connect_interpret()
2589 which thought it would be the only running connect process.
2592 Frequency : only with obd_echo servers and clients that are rebooted
2594 Description: kernel BUG accessing uninitialized data structure
2595 Details : When running an obd_echo server it did not start the ping_evictor
2596 thread, and when a client was evicted an uninitialized data
2597 structure was accessed. Start the ping_evictor in the RPC
2598 service startup instead of the OBD startup.
2600 Severity : enhancement
2601 Bugzilla : 10193 (patchless)
2602 Description: Remove dependency on various unexported kernel interfaces.
2603 Details : No longer need reparent_to_init, exit_mm, exit_files,
2604 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2607 Frequency : rare (only users of deprecated and unsupported LDAP config)
2609 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2610 Details : LDAP apparently contains 'lustreName' attributes instead of
2611 'name'. A simple remapping of the name is sufficient.
2614 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2616 Description: waiting_locks_callback trips kernel BUG if client is evicted
2617 Details : Running with the dump_on_timeout debug flag turned on makes
2618 it possible that the waiting_locks_callback() can try to dump
2619 the Lustre kernel debug logs from an interrupt handler. Defer
2620 this log dumping to the expired_lock_main() thread.
2622 Severity : enhancement
2624 Description: Support NFS exporting on 2.6 kernels.
2625 Details : Implement non-rawops metadata methods for NFS server to use without
2626 changing NFS server code.
2629 Frequency : very rare (synthetic metadata workload only)
2631 Description: two racing renames might cause an MDS thread to deadlock
2632 Details : Running the "racer" program may cause one MDS thread to rename
2633 a file from being the source of a rename to being the target of
2634 a rename at exactly the same time that another thread is doing
2635 so, and the second thread has already enqueued these locks after
2636 doing a lookup of the target and is trying to relock them in
2637 order. Ensure that we don't try to re-lock the same resource.
2640 Frequency : only very large systems with liblustre clients
2642 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2643 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2644 all OSTs in parallel. This allows the eviction of stale liblustre
2645 clients to proceed much faster than if they were done in series,
2646 and also offers similar improvements for other set_info RPCs.
2651 Description: excessive CPU usage during initial read phase on client
2652 Details : During the initial read phase on a client, it would agressively
2653 retry readahead on the file, consuming too much CPU and impacting
2654 performance (since 1.4.5.8). Improve the readahead algorithm
2655 to avoid this, and also improve some other common cases (read
2656 of small files in particular, where "small" is files smaller than
2657 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2662 Description: MDS crash when receiving packet with unknown intent.
2663 Details : Do not LBUG in unknown intent case, just return -EFAULT
2665 Severity : enhancement
2666 Bugzilla : 9293, 9385
2667 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2668 Details : Do not serialize getattr (non-intent version) and statfs.
2671 Frequency : occasional, when OST network is overloaded/intermittent
2673 Description: client evicted by OST after bulk IO timeout
2674 Details : If a client sends a bulk IO request (read or write) the OST
2675 may evict the client if it is unresposive to its data GET/PUT
2676 request. This is incorrect if the network is overloaded (takes
2677 too long to transfer the RPC data) or dropped the OST GET/PUT
2678 request. There is no need to evict the client at all, since
2679 the pinger and/or lock callbacks will handle this, and the
2680 client can restart the bulk request.
2683 Frequency : Always when mmapping file with no objects
2685 Description: client crashes when mmapping file with no objects
2686 Details : Check that we actually have objects in a file before doing any
2687 operations on objects in ll_vm_open, ll_vm_close and
2693 Description: Request leak when working with deleted CWD
2694 Details : Introduce advanced request refcount tracking for requests
2695 referenced from lustre intent.
2697 Severity : Enhancement
2699 Description: Cache open file handles on client.
2700 Details : MDS now will return special lock along with openhandle, if
2701 requested and client is allowed to hold openhandle, even if unused,
2702 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2703 closing files for every read/write openration.
2705 Severity : Enhancement
2707 Description: Cache open negative dentries on client when possible.
2708 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2709 negative dentries on lock revocation.
2714 Description: Remounting a client read-only wasn't possible with a zconf mount
2715 Details : It wasn't possible to remount a client read-only with llmount.
2717 Severity : enhancement
2718 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2719 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2720 MPICH is included to add Lustre-specific ADIO interfaces.
2721 This is based closely on the UFS ADIO layer and only differs
2722 in file creation, in order to allow the OST striping to be set.
2723 This is user-contributed code and not supported by CFS.
2728 Description: extended inode attributes (immutable, append-only) work improperly
2729 when 2.4 and 2.6 kernels are used on client/server or vice versa
2730 Details : Introduce kernel-independent values for these flags.
2732 Severity : enhancement
2735 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2736 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2737 granularity. Now, Fractional MB granularity is be supported,
2738 this is very useful for low memory system.
2740 Severity : enhancement
2742 Description: Getattr by fid
2743 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2744 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2745 have locks and avoids getattr rpc requests for it on every stat.
2748 Frequency : Always, for filesystems larger than 2TB
2750 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2751 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2752 filesystems larger than 2TB to be mounted with mballoc enabled.
2755 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2757 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2758 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2759 filesystems larger than 2TB is mounted then the kernel will
2760 silently corrupt the start of the filesystem. CONFIG_LBD is
2761 enabled for all CFS-supported kernels, but the possibility of
2762 this happening with a modified kernel config exists.
2764 Severity : enhancement
2766 Description: add client O_DIRECT support for 2.6 kernels
2767 Details : It is now possible to do O_DIRECT reads and writes to files
2768 in the Lustre client mountpoint on 2.6 kernel clients.
2770 Severity : enhancement
2772 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2773 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2774 parallel, not waiting for response from every OST before sending
2775 a rpc to the next OST.
2780 Description: setattr vs write race when updating file timestamps
2781 Details : Client processes that update a file timestamp into the past
2782 right after writing to the file (e.g. tar) it is possible that
2783 the updated file modification time can be reset to the current
2784 time due to a race between processing the setattr and write RPC.
2786 Severity : enhancement
2788 Description: Bring 'lfs find' closer in line with regular Linux find.
2789 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2790 -print0 options and obtains all the needed info through the lustre
2793 Severity : enhancement
2795 Description: support up to 1024 configured devices on one node
2796 Details : change obd_dev array from statically allocated to dynamically
2797 allocated structs as they are first used to reduce memory usage
2802 Description: Flush dirty partially truncated pages during truncate
2803 Details : Immediatelly flush partially truncated pages in filter_setattr,
2804 this way we completely avoid having any pages in page cache on OST
2805 and can retire ugly workarounds during writes to flush such pages.
2810 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2811 Details : protect inode from truncation within vfs_unlink() context
2812 just take a reference before calling vfs_unlink() and release it
2813 when parent's i_sem is free.
2818 Description: last_id value checked outside lock on OST caused LASSERT failure
2819 Details : If there were multiple MDS->OST object precreate requests in
2820 flight, it was possible that the OST's last object id was checked
2821 outside a lock and incorrectly tripped an assertion. Move checks
2822 inside locks, and discard old precreate requests.
2825 Frequency : always, if extents are used on OSTs
2827 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2828 Details : OSTs using the extents format would not zero the high 16 bits of
2829 the index physical block number. This is not a problem for any
2830 OST filesystems smaller than 16TB, and no kernels support ext3
2831 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2832 new/modified files) and can be fixed for existing filesystems
2833 with e2fsprogs-1.39-cfs1.
2838 Description: import connection selection may be incorrect if timer wraps
2839 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2840 import connections to be ignored if the 32-bit jiffies counter
2841 wraps. Use a 64-bit jiffies counter.
2844 Frequency : very large clusters immediately after boot
2846 Description: LNET request buffers exhausted under heavy short-term load
2847 Details : If a large number of client requests are generated on a service
2848 that has previously never seen so many requests it is possible
2849 that the request buffer growth cannot keep up with the spike in
2850 demand. Instead of dropping incoming requests, they are held in
2851 the LND until the RPC service can accept more requests.
2854 Frequency : Sometimes during replay
2856 Description: Assertion failure in ll_local_open after replay.
2857 Details : If replay happened on an open request reply before we were able
2858 to set replay handler, reply will become not swabbed tripping the
2859 assertion in ll_local_open. Now we set the handler right after
2860 recognising of open request
2863 Frequency : very rare
2865 Description: kernel reports "badness in vsnprintf"
2866 Details : Reading from the "recovery_status" /proc file in small chunks
2867 may cause a negative length in lprocfs_obd_rd_recovery_status()
2868 call to vsnprintf() (which is otherwise harmless). Exit early
2869 if there is no more space in the output buffer.
2871 Severity : enhancement
2873 Description: clear OBD RPC statistics by writing to them
2874 Details : It is now possible to clear the OBD RPC statistics by writing
2875 to the "stats" file.
2880 Description: Client mtime is not the same on different clients after utimes
2881 Details : In some cases, the client was using the utimes() syscall on
2882 a file cached on another node. The clients now validate the
2883 ctime from the MDS + OSTs to determine which one is right.
2888 Description: Inability to activate failout mode
2889 Details : lconf script incorrectly assumed that in pythong string's numeric
2890 value is used in comparisons.
2893 Frequency : always with multiple stripes per file
2895 Description: Inefficient object allocation for mutli-stripe files
2896 Details : When selecting which OSTs to stripe files over, for files with
2897 a stripe count that divides evenly into the number of OSTs,
2898 the MDS is always picking the same starting OST for each file.
2899 Return the OST selection heuristic to the original design.
2904 Description: mount failures may take full timeout to return an error
2905 Details : Under some heavy load conditions it is possible that a
2906 failed mount can wait for the full obd_timeout interval,
2907 possibly several minutes, before reporting an error.
2908 Instead return an error as soon as the status is known.
2910 Frequency : quota enabled and large files being deleted
2912 Description: releasing more than 4GB of quota at once hangs OST
2913 Details : If a user deletes more than 4GB of files on a single OST it
2914 will cause the OST to spin in an infinite loop. Release
2915 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2920 Description: statfs data retrieved from /proc may be stale or zero
2921 Details : When reading per-device statfs data from /proc, in the
2922 {kbytes,files}_{total,free,avail} files, it may appear
2923 as zero or be out of date.
2926 Frequency : systems with MD RAID1 external journal devices
2928 Description: lconf's call to blkid is confused by RAID1 journal devices
2929 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2930 of returning all block devices that match the journal UUID.
2933 Frequency : always, for aggregate stripe size over 4GB
2935 Description: assertion fails when trying to use 4GB stripe size
2936 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2937 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2940 Frequency : always on ppc64
2942 Description: the first write on an ext3 filesystem with mballoc got stuck
2943 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2944 perform endianness conversion.
2946 ------------------------------------------------------------------------------
2948 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2950 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2951 INTEROPERATE with older versions automatically. Please read the
2952 user documentation before upgrading any part of a live system.
2953 * WARNING: Lustre networking configuration changes are required with
2954 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2957 * Support for kernels:
2958 2.6.9-22.0.2.EL (RHEL 4)
2959 2.6.5-7.244 (SLES 9)
2960 2.6.12.6 vanilla (kernel.org)
2963 Severity : enhancement
2964 Bugzilla : 7981/8208
2965 Description: Introduced Lustre Networking (LNET)
2966 Details : LNET is new networking infrastructure for Lustre, it includes
2967 a reorganized network configuration mode (see the user
2968 documentation for full details) as well as support for routing
2969 between different network fabrics. Lustre Networking Devices
2970 (LNDs) for the supported network fabrics have also been
2971 created for this new infrastructure.
2973 Severity : enhancement
2974 Description: Introduced Access control lists
2975 Details : clients can set ACLs on files and directories in order to have
2976 more fine-grained permissions than the standard Unix UGO+RWX.
2977 The MDS must be started with the "-o acl" mount option.
2979 Severity : enhancement
2980 Description: Introduced filesystem quotas
2981 Details : Administrators may now establish per-user quotas on the
2984 Severity : enhancement
2986 Description: Configuration change for the XT3
2987 The PTLLND is now used to run Lustre over Portals on the XT3
2988 The configure option(s) --with-cray-portals are no longer used.
2989 Rather --with-portals=<path-to-portals-includes> is used to
2990 enable building on the XT3. In addition to enable XT3 specific
2991 features the option --enable-cray-xt3 must be used.
2996 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2997 Details : A race between allocating a new llcd and re-getting the llcd_lock
2998 allowed another thread to grab newly-allocated llcd.
3000 Severity : enhancement
3002 Description: 2.6 OST async journal commit and locking fix to improve performance
3003 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3004 2.6 kernels are now async as they already were in 2.4 kernels so
3005 that they can commit concurrently with the network bulk transfer.
3006 For block-allocated files the filter allocation semaphore is held
3007 to avoid filesystem fragmentation during allocation. BKL lock
3008 removed for 2.6 xattr operations where it is no longer needed.
3013 Description: lconf incorrectly determined whether two IP networks could talk
3014 Details : In some more complicated routing and multiple-network
3015 configurations, lconf will avoid trying to make a network
3016 connection to a disjoint part of the IP space. It was doing the
3017 math incorrectly for one set of cases.
3022 Description: Fix for potential infinite loop processing records in an llog.
3023 Details : If an llog record is corrupted/zeroed, it is possible to loop
3024 forever in llog_process(). Validate the llog record length
3025 and skip the remainder of the block on error.
3028 Frequency : occasional (liblustre only)
3030 Description: liblustre could not open files whose last component is a symlink
3031 Details : sysio_path_walk() would incorrectly pass the open intent to
3032 intermediate path components.
3035 Frequency : rare (liblustre only with non-standard tuning)
3036 Bugzilla : 7201 (7350)
3037 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3038 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3039 after it has been released, only do this for referenced locks
3041 Severity : enhancement
3043 Description: specifying an (invalid) directory default stripe_size of -1
3044 would reset the directory default striping
3045 Details : stripe_size -1 was used internally to signal directory stripe
3046 removal, now use "all default" to signal dir stripe removal
3047 as a directory striping of "all default" is not useful
3050 Frequency : common for large clusters running liblustre clients
3052 Description: doing an ls when liblustre clients are running is slow
3053 Details : sending a glimpse AST to a liblustre client waits for every AST
3054 to time out, as liblustre clients will not respond. Since they
3055 cannot cache data we refresh the OST lock LVB from disk instead.
3057 Severity : enhancement
3059 Description: doing an ls at the same time as file IO can be slow
3060 Details : enqueue and other "small" requests can be blocked behind many
3061 large IO requests. Create a new OST IO portal for non-IO
3062 requests so they can be processed faster.
3065 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3067 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3068 created file with mode 000
3069 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3070 allow an ftruncate() as a non-root user to a file with mode 000.
3071 Lustre now respects this flag to disable mode checks when
3072 truncating a file owned by the user
3075 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3078 Description: Revoking locks from clients that went dead or catatonic might take
3080 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3081 cancellation of such locks instant on servers without waiting for
3082 any reply from clients. Clients drops these locks when cancel
3083 notification from server is received without replying.
3086 Frequency : liblustre-only, when liblustre client dies or becomes busy
3088 Description: Doing ls on Linux clients can take a long time with active
3090 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3091 granting such locks to it in the first place if possible. Locks
3092 are taken by proxy on the OST during the read or write and
3093 dropped immediately afterward. Add connect flags handling, do
3094 not grant locks to liblustre clients for glimpse ASTs.
3096 Severity : enhancement
3098 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3099 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3100 sequential reads and behaves badly with applications doing random
3101 reads. Improve it by reading ahead at least read region, and
3102 avoiding excessive large RPC for small reads.
3104 Severity : enhancement
3106 Description: Creating more than 1000 files for a single job may cause a load
3107 imbalance on the OSTs if there are also a large number of OSTs.
3108 Details : qos_prep_create() uses an OST index reseed value that is an
3109 even multiple of the number of available OSTs so that if the
3110 reseed happens in the middle of the object allocation it will
3111 still utilize the OSTs as uniformly as possible.
3116 Description: OST or MDS may oops in ping_evictor_main()
3117 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3118 but doesn't restart at beginning of obd_exports_timed list
3121 Severity : enhancement
3123 Description: improve by-nid export eviction on the MDS and OST
3124 Details : allow multiple exports with the same NID to be evicted at one
3125 time without re-searching the exports list.
3128 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3130 Description: MDS may oops in groups_free()
3131 Details : in rare race conditions a newly allocated group_info struct is
3132 freed again, and this can be NULL. The 2.4 compatibility code
3133 for groups_free() checked for a NULL pointer, but 2.6 did not.
3136 Frequency : common for liblustre clients doing little filesystem IO
3137 Bugzilla : 9352, 7313
3138 Description: server may evict liblustre clients accessing contended locks
3139 Details : if a client is granted a lock or receives a completion AST
3140 with a blocking AST already set it would not reply to the AST
3141 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3144 Frequency : lfs setstripe, only systems with more than 160 OSTs
3146 Description: unable to set striping with a starting offset beyond OST 160
3147 Details : llapi_create_file() incorrectly limited the starting stripe
3148 index to the maximum single-file stripe count.
3151 Frequency : LDAP users only
3153 Description: lconf did not handle in-kernel recovery with LDAP properly
3154 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3156 Severity : enhancement
3158 Description: bind OST threads to NUMA nodes to improve performance
3159 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3160 node and do their allocations there to localize memory access
3162 Severity : enhancement
3164 Description: llmount can determine client NID directly from Myrinet (GM)
3165 Details : the client NID code from gmnalnid was moved directly into
3166 llmount, removing the need to use this or specifying the
3167 client NID explicitly when mounting GM clients with zeroconf
3170 Frequency : if client is started with down MDS
3172 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3173 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3174 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3175 ptlrpc_import_delayed_req() because we didn't check if the
3176 request was interrupted, and we also didn't break out of the
3177 event loop if there was no timeout
3182 Description: data loss during non-page-aligned writes to a single file from
3183 both multiple nodes and multiple threads on one node at same time
3184 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3185 inconsistency led to false short-reads, that were cached and later
3186 used by ->prepare_write() to fill in partially written page,
3187 leading to data loss.
3190 Frequency : always, if lconf --abort_recovery used
3192 Description: lconf --abort_recovery fails with 'Operation not supported'
3193 Details : lconf was attempting to abort recovery on the MDT device and not
3196 Severity : enhancement
3198 Description: remove cleanup logs
3199 Details : replace lconf-generated cleanup logs with lustre internal
3200 cleanup routines. Eliminates the need for client-cleanup and
3203 Severity : enhancement
3205 Description: add support for EAs (user and system) on lustre filesystems
3206 Details : it is now possible to store extended attributes in the Lustre
3207 client filesystem, and with the user_xattr mount option it
3208 is possible to allow users to store EAs on their files also
3210 Severity : enhancement
3212 Description: Add possibility (config option) to show minimal available OST free
3214 Details : When compiled with --enable-mindf configure option, statfs(2)
3215 (and so, df) will return least minimal free space available from
3216 all OSTs as amount of free space on FS, instead of summary of
3217 free spaces of all OSTs.
3219 Severity : enhancement
3221 Description: do not expand extent locks acquired on OST-side
3222 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3223 by server: they are not cached anyway.
3226 Frequency : when mmap is used/binaries executed from Lustre
3228 Description: Unmmap pages before throwing them away from read cache.
3229 Details : llap_shrink cache now attempts to unmap pages before discarding
3230 them (if unmapping failed - do not discard). SLES9 kernel has
3231 extra checks that trigger if this unmapping is not done first.
3236 Description: lconf didn't resolve symlinks before checking to see whether a
3237 given mountpoint was already in use
3240 Frequency : when migrating failover services
3241 Bugzilla : 6395, 9514
3242 Description: When migrating a subset of services from a node (e.g. failback
3243 from a failover service node) the remaining services would
3244 time out and evict clients.
3245 Details : lconf --force (implied by --failover) sets the global obd_timeout
3246 to 5 seconds in order to quickly disconnect, but this caused
3247 other RPCs to time out too quickly. Do not change the global
3248 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3250 Severity : enhancement
3251 Frequency : if MDS is started with down OST
3252 Bugzilla : 9439,5706
3253 Description: Allow startup/shutdown of an MDS without depending on the
3254 availability of the OSTs.
3255 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3256 Add appropriate locking and lov-osc refcounts for safe
3257 cleaning. Add osc abort_inflight calls in case the
3258 synchronize never started.
3261 Frequency : occasional (Cray XT3 only)
3263 Description: root not authorized to access files in CRAY_PORTALS environment
3264 Details : The client process capabilities were not honoured on the MDS in
3265 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3266 been accessed by an authorized user then root was able to access
3267 the file on the local client also. The root user capabilities
3268 are now allowed on the MDS, as this environment has secure UID.
3271 Frequency : occasional
3273 Description: ldiskfs "too long searching" message happens too often
3274 Details : A debugging message (otherwise harmless) prints too often on
3275 the OST console. This has been reduced to only happen when
3276 there are fragmentation problems on the filesystem.
3281 Description: Division by zero in statfs when all OSCs are inactive
3282 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3283 lov_statfs divides by value returned by lov_get_stripecnt().
3287 Bugzilla : 9489, 3273
3288 Description: First write from each client to each OST was only 4kB in size,
3289 to initialize client writeback cache, which caused sub-optimal
3290 RPCs and poor layout on disk for the first writen file.
3291 Details : Clients now request an initial cache grant at (re)connect time
3292 and so that they can start streaming writes to the cache right
3293 away and always do full-sized RPCs if there is enough data.
3294 If the OST is rebooted the client also re-establishes its grant
3295 so that client cached writes will be honoured under the grant.
3300 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3301 Details : Now I/O RPCs go to different portal number and (presumably) fast
3302 lock requests (and glimses) and other RPCs get their own service
3303 threads pool that should be able to service those RPCs
3306 Severity : enhancement
3308 Description: Ability to exchange lustre version between client and servers and
3309 issue warnings at client side if client is too old. Also for
3310 liblustre clients there is ability to refuse connection of too old
3312 Details : New 'version' field is added to connect data structure that is
3313 filled with version info. That info is later checked by server and
3317 Frequency : rare, liblustre only.
3318 Bugzilla : 9296, 9581
3319 Description: Two simultaneous writes from liblustre at offset within same page
3320 might proceed at the same time overwriting eachother with stale
3322 Details : I/O lock withing llu_file_prwv was released too early, before data
3323 actually was hitting the wire. Extended lock-holding time until
3324 server acknowledges receiving data.
3327 Frequency : extremely rare. Never observed in practice.
3329 Description: avoid generating lustre_handle cookie of 0.
3330 Details : class_handle_hash() generates handle cookies by incrementing
3331 global counter, and can hit 0 occasionaly (this is unlikely, but
3332 not impossible, because initial value of cookie counter is
3333 selected randonly). Value of 0 is used as a sentinel meaning
3334 "unassigned handle" --- avoid it. Also coalesce two critical
3335 sections in this function into one.
3337 Severity : enhancement
3339 Description: allow liblustre clients to delegate truncate locking to OST
3340 Details : To avoid overhead of locking, liblustre client instructs OST to
3341 take extent lock in ost_punch() on client's behalf. New connection
3342 flag is added to handle backward compatibility.
3344 Severity : enhancement
3345 Bugzilla : 4928, 7341, 9758
3346 Description: allow number of OST service threads to be specified
3347 Details : a module parameter allows the number of OST service threads
3348 to be specified via "options ost ost_num_threads={N}" in the
3349 OSS's /etc/modules.conf or /etc/modprobe.conf.
3353 Bugzilla : 6146, 9635, 9895
3354 Description: servers crash with bad pointer in target_handle_connect()
3355 Details : In rare cases when a client is reconnecting it was possible that
3356 the connection request was the last reference for that export.
3357 We would temporarily drop the export reference and get a new
3358 one, but this may have been the last reference and the export
3359 was just destroyed. Get new reference before dropping old one.
3361 Severity : enhancement
3362 Frequency : if client is started with failover MDS
3364 Description: Allow multiple MDS hostnames in the mount command
3365 Details : Try to read the configuration from all specified MDS
3366 hostnames during a client mount in case the "primary"
3369 Severity : enhancement
3371 Description: Stop sending data to evicted clients as soon as possible.
3372 Details : Check if the client we are about to send or are sending data to
3373 was evicted already. (Check is done every second of waiting,
3374 for which l_wait_event interface was extended to allow checking
3375 of exit condition at specified intervals).
3378 Frequency : rare, normally only when NFS exporting is done from client
3380 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3382 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3385 Frequency : always, if srand() is called before liblustre initialization
3387 Description: Liblustre uses system PRNG disturbing its usage by user application
3388 Details : Introduce internal to lustre fast and high-quality PRNG for
3389 lustre usage and make liblustre and some other places in generic
3390 lustre code to use it.
3392 Severity : enhancement
3393 Bugzilla : 9477, 9557, 9870
3394 Description: Verify that the MDS configuration logs are updated when xml is
3395 Details : Check if the .xml configuration logs are newer than the config
3396 logs stored on the MDS and report an error if this is the case.
3397 Request --write-conf, or allow starting with --old_conf.
3399 Severity : enhancement
3401 Description: Handle symlinks in the path when checking if Lustre is mounted.
3402 Details : Resolve intermediate symlinks when checking if a client has
3403 mounted a filesystem to avoid duplicate client mounts.
3408 Description: lconf can hit an error exception but still return success.
3409 Details : The lconf command catches the Command error exception at the top
3410 level script context and will exit with the associated exit
3411 status, but doesn't ensure that this exit status is non-zero.
3416 Description: failure of ptlrpc thread startup can cause oops
3417 Details : Starting a ptlrpc service thread can fail if there are a large
3418 number of threads or the server memory is very fragmented.
3419 Handle this without oopsing.
3422 Frequency : always, only if liblustre and non-default acceptor port was used
3424 Description: liblustre cannot connect to servers with non-default acceptor port
3425 Details : tcpnal_set_default_params() was not called and was therefore
3426 ignoring the environment varaible TCPNAL_PORT, as well as other
3427 TCPNAL_ environment variables
3432 Description: two objects could be created on the same OST for a single file
3433 Details : If an OST is down, in some cases it was possible to create two
3434 objects on a single OST for a single file. No problems other
3435 than potential performance impact and spurious error messages.
3439 Bugzilla : 5681, 9562
3440 Description: Client may oops in ll_unhash_aliases
3441 Details : Client dcache may become inconsistent in race condition.
3442 In some cases "getcwd" can fail if the current directory is
3448 Description: Inode refcounting problems in NFS export code
3449 Details : link_raw functions used to call d_instantiate without obtaining
3450 extra inode reference first.
3454 Bugzilla : 9942, 9903
3455 Description: Referencing freed requests leading to crash, memleaks with NFS.
3456 Details : We used to require that call to ll_revalidate_it was always
3457 followed by ll_lookup_it. Also with revalidate_special() it is
3458 possible to call ll_revalidate_it() twice for the same dentry
3459 even if first occurence returned success. This fix changes semantic
3460 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3461 reference on a request referred from the intent.
3462 ll_intent_release() then releases such a request.
3465 Frequency : rare, normally benchmark loads only
3467 Description: unlinked inodes were kept in memory on the client
3468 Details : If a client is repeatedly creating and unlinking files it
3469 can accumulate a lot of stale inodes in the inode slab cache.
3470 If there is no other client load running this can cause the
3471 client node to run out of memory. Instead flush old inodes
3472 from client cache that have the same inode number as a new inode.
3475 Frequency : SLES9 2.6.5 kernel and long filenames only
3476 Bugzilla : 9969, 10379
3477 Description: utime reports stale NFS file handle
3478 Details : SLES9 uses out-of-dentry names in some cases, which confused
3479 the lustre dentry revalidation. Change it to always use the
3483 Frequency : rare, unless heavy write-truncate concurrency is continuous
3484 Bugzilla : 4180, 6984, 7171, 9963, 9331
3485 Description: OST becomes very slow and/or deadlocked during object unlink
3486 Details : filter_destroy() was holding onto the parent directory lock
3487 while truncating+unlinking objects. For very large objects this
3488 may block other threads for a long time and slow overall OST
3489 responsiveness. It may also be possible to get a lock ordering
3490 deadlock in this case, or run out of journal credits because of
3491 the combined truncate+unlink. Solution is to do object truncate
3492 first in one transaction without parent lock, and then do the
3493 final unlink in a new transaction with the parent lock. This
3494 reduces the lock hold time dramatically.
3497 Frequency : rare, 2.4 kernels only
3499 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3500 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3501 before unmounting their filesystem in order to avoid deadlock.
3502 The kernel_locked() function in 2.4 kernels only checks whether
3503 the kernel lock is held, not whether it is this process that is
3504 holding it as 2.6 kernels do.
3509 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3510 Details : The client ptlrpc code may be trying to reconnect to a down
3511 server before a previous connection attempt has timed out.
3512 Increase the reconnect interval to be longer than the connection
3513 timeout interval to avoid sending duplicate connections to
3517 Frequency : echo_client brw_test command
3519 Description: fix echo_client to work with OST preallocated code
3520 Details : OST preallocation code (5137) didn't take echo_client IO path
3521 into account: echo_client calls filter methods outside of any
3522 OST thread and, hence, there is no per-thread preallocated
3523 pages and buffers to use. Solution: hijack pga pages for IO. As
3524 a byproduct, this avoids unnecessary data copying.
3528 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3529 Description: Client can oops in mdc_commit_close() after open replay
3530 Details : It was possible for the MDS to return an open request with no
3531 transaction number in mds_finish_transno() if the client was
3532 evicted, but without actually returning an error. Clients
3533 would later try to replay that open and may trip an assertion
3534 Simplify the client close codepath, and always return an error
3535 from the MDS in case the open is not successful.
3538 Frequency : rare, 2.6 OSTs only
3540 Description: OST may deadlock under high load on fragmented files
3541 Details : If there was a heavy load and highly-fragmented OST filesystems
3542 it was possible to have all the OST threads deadlock waiting on
3543 allocation of biovecs, because the biovecs were not released
3544 until the entire RPC IO was completed. Instead, release biovecs
3545 as soon as they are complete to ensure forward IO progress.
3547 Severity : enhancement
3549 Description: Support for specifying external journal device at mount
3550 Details : If an OST or MDS device is formatted with an external journal
3551 device, this device major/minor is stored in the ext3 superblock
3552 and may not be valid for failover. Allow detecting and
3553 specifying the external journal at mount time.
3558 Description: Mounting an MDS with pending unlinked files may cause oops
3559 Details : target_finish_recovery() calls mds_postrecov() which returned
3560 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3561 considers this an error and immediately begins cleaning up the
3562 lov, just after starting the mds_lov process
3564 Severity : enhancement
3566 Description: Implement 'lfs df' to report actual free space on per-OST basis
3567 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3568 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3569 usage of objects; '-h' to report in human readable format.
3571 ------------------------------------------------------------------------------
3573 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3580 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3581 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3582 at startup during orphan unlinking. Instead, skip buddy update
3583 before setup, as it will be regenerated after recovery is complete.
3588 Description: If an OST is inactive, its locks might reference stale inodes.
3589 Details : lov_change_cbdata() must iterate over all namespaces, even if
3590 they are inactive to clear inode references from the lock.
3592 Severity : enhancement
3593 Frequency : occasional, if non-standard max_dirty_mb used
3595 Description: Client will block write RPCs if not enough grant
3596 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3597 then the client will block writes while waiting for another RPC
3598 to complete instead of consuming its dirty limit. With change
3599 we get improved performance when max_dirty_mb is small.
3601 Severity : enhancement
3602 Bugzilla : 3389, 6253
3603 Description: Add support for supplementary groups on the MDS.
3604 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3605 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3606 lookups for user supplementary groups into a cache.
3610 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3611 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3612 for liblustre O_EXCL is enforced by MDS.
3616 Description: readdir never returns NULL in liblustre.
3617 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3618 dentry in struct dirent.
3622 Frequency : liblustre only, depends on application IO pattern
3623 Description: liblustre clients evicted if not contacting servers
3624 Details : Don't put liblustre clients into the ping_evictor list, so
3625 they will not be evicted by the pinger ever.
3627 Severity : enhancement
3629 Description: Add ability to evict clients by NID from MDS.
3630 Details : By echoing "nid:$NID" string into
3631 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3632 $NID would be instantly evicted from this MDS and from all active
3633 OSTs connected to it.
3637 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3638 Details : lookup_it_finish() used to query file size from OSTs that was not
3643 Description: service threads change working directory to that of init
3644 Details : Starting lustre service threads may pin the working directory
3645 of the parent thread, making that filesystem busy. Threads
3646 now change to the working directory of init to avoid this.
3650 Frequency : during shutdown only
3651 Description: shutdown with a failed MDS or OST can cause unmount to hang
3652 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3657 Frequency : 2.6 only
3658 Description: chmod/chown may include an extra supplementary group
3659 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3660 supplementary group and if none is specified this is used.
3663 Bugzilla : 5479 (6816)
3665 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3666 Details : If lookup is in progress on a file that is unlinked we might try
3667 to revalidate the inode and fail in revalidate after lookup is
3668 complete and ll_file_open() enqueues the open again but
3669 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3672 Frequency : always, if lconf --abort_recovery used
3674 Description: lconf --abort_recovery fails with 'Operation not supported'
3675 Details : lconf was attempting to abort recovery on the MDT device and not
3678 ------------------------------------------------------------------------------
3680 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3685 Frequency : rare (only unsupported configurations with a node running as an
3687 Bugzilla : 6514, 5137
3688 Description: Mounting a Lustre file system on a node running as an OST could
3690 Details : OSTs now preallocates memory needed to write out data at
3691 startup, instead of when needed, to avoid having to
3692 allocate memory in possibly low memory situations.
3693 Specifically, if the file system is mounted on on OST,
3694 memory pressure could force it to try to write out data,
3695 which it needed to allocate memory to do. Due to the low
3696 memory, it would be unable to do so and the node would
3697 become unresponsive.
3699 Severity : enhancement
3701 Description: Addition of lconf --service command line option
3702 Details : lconf now accepts a '--service <arg>' option, which is
3703 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3705 Severity : enhancement
3707 Description: Failover mode is now the default for OSTs.
3708 Details : By default, OSTs will now run in failover mode. To return to
3709 the old behaviour, add '--failout' to the lmc line for OSTs.
3711 Severity : enhancement
3713 Description: Health checks are now provided for MDS and OSTs
3714 Details : Additional detailed health check information on MSD and OSTs
3715 is now provided through the procfs health_check value.
3718 Frequency : occasional, depends on IO load
3720 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3721 after numerous create/delete cycles
3722 Details : The ext3 inode allocation policy would not allocate new inodes
3723 very well on the OSTs because there are no new directories
3724 being created. Instead we look for groups with free space if
3725 the parent directories are nearly full.
3730 Description: Network or server problems during mount may cause partially
3731 mounted clients instead of returning an error.
3732 Details : The config llog parsing code may overwrite the error return
3733 code during mount error handling, returning success instead
3739 Description: MDS can fail to allocate large reply buffers
3740 Details : After long uptimes the MDS can fail to allocate large reply
3741 buffers (e.g. zconf client mount config records) due to memory
3742 fragmentation or consumption by the buffer cache. Preallocate
3743 some large reply buffers so that these replies can be sent even
3744 under memory pressure.
3748 Frequency : rare (liblustre)
3749 Description: fsx running with liblustre complained that using truncate() to
3750 extend the file doesn't work. This patch corrects that issue.
3751 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3752 ATTR_SIZE and lsm use in llu_setattr_raw.
3756 Frequency : rare, only 2.6 kernels
3757 Description: Unusual file access patterns on the MDS may result in inode
3758 data being lost in very rare circumstances.
3759 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3760 code in the 2.6 kernel caused the inode and/or EA data not to
3761 be read from disk, causing single-file corruption.
3765 Frequency : rare, only 2.6 filesystems using extents
3766 Description: Heavy concurrent write and delete load may cause data corruption.
3767 Details : It was possible under high-load situations to have an extent
3768 metadata block in the block device cache from a just-unlinked
3769 file overwrite a newly-allocated data block. We now unmap any
3770 metadata buffers that alias just-allocated data blocks.
3774 Frequency : filesystems with default stripe_count larger than 77
3775 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3776 Details : lconf specifies an inode size of 4096 bytes when the default
3777 stripe_count is larger than 77. This conflicts with the default
3778 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3779 this case to avoid pinning too much memory for large EAs.
3781 ------------------------------------------------------------------------------
3783 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3788 Frequency : rare (extremely heavy IO load with hundreds of clients)
3790 Description: Client is evicted, gets IO error writing to file
3791 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3792 caused clients to be evicted for AST timeouts. The fixes in
3793 bug 5192 mean we no longer need to have such short AST timeouts
3794 so ldlm_timeout has been increased.
3797 Frequency : occasional during --force or --failover shutdown under load
3798 Bugzilla : 5949, 4834
3799 Description: Server oops/LBUG if stopped with --force or --failover under load
3800 Details : a collection of import/export refcount and cleanup ordering
3801 issues fixed for safer force cleanup
3804 Frequency : only filesystems larger than 120 OSTs
3805 Bugzilla : 5990, 6223
3806 Description: lfs getstripe would oops on a very large filesystem
3807 Details : lov_getconfig used kfree on vmalloc'd memory
3810 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3811 Bugzilla : 6242, 6243
3812 Description: reading from files that had been truncated to a non-zero size
3813 but never opened returned no data
3814 Details : ll_file_read() reads zeros from no-object files to EOF
3819 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3820 Details : There's little harm in aborting MDS/OSS recovery and letting it
3821 try again, so I removed the LASSERT and return an error instead.
3823 Severity : enhancement
3825 Description: New debugging infrastructure for tracking down data corruption
3826 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3827 (b) cover more of the client-side code path, and (c) try to narrow
3828 down where problems occurred
3832 Bugzilla : 3819, 4364, 4397, 6313
3833 Description: Racing close and eviction MDS could cause assertion in mds_close
3834 Details : It was possible to get multiple mfd references during close and
3835 client eviction, leading to one thread referencing a freed mfd.
3837 Severity: : enhancement
3838 Bugzilla : 3262, 6359
3839 Description: Attempts to reconnect to servers are now more aggressive.
3840 Details : This builds on the enhanced upcall-less recovery that was added
3841 in 1.4.2. When trying to reconnect to servers, clients will
3842 now try each server in the failover group every 10 seconds. By
3843 default, clients would previously try one server every 25 seconds.
3848 Description: After recovery, certain operations trigger a failed
3849 assertion on a client.
3850 Details : Failing over an mds, using lconf -d --failover, while a
3851 client was doing a readdir() call would cause the client to
3852 LBUG after recovery completed and the readdir() was resent.
3854 Severity : enhancement
3856 Description: Default groups are now added by lconf
3857 Details : You can now run lconf --group <servicename> without having to
3858 manually add groups with lmc.
3861 Frequency : occasional
3863 Description: Nodes with an elan id of 0 trigger a failed assertion
3866 Frequency : always when accessing e.g. tty/console device nodes
3868 Description: tty and some other devices nodes cannot be used on lustre
3869 Details : file's private_data field is used by device data and lustre
3870 values in there got lost. New field was added to struct file to
3871 store fs-specific private data.
3874 Frequency : when exporting Lustre via NFS
3876 Description: NFSD failed occasionally when looking up a path component
3877 Details : NFSD is looking up ".." which was broken in ext3 directories
3878 that had grown large enough to become hashed.
3881 Frequency : Clusters with multiple interfaces not on the same subnet
3883 Description: Nodes will repeatedly try to reconnect to an interface which it
3884 cannot reach and report an error to the log.
3885 Details : Extra peer list entries will be created by lconf with some peers
3886 unreachable. lconf now validates the peer before adding it.
3889 Frequency : Only if a default stripe is set on the filesystem root.
3891 Description: Setting a default stripe on the filesystem root prevented the
3892 filesystem from being remounted.
3893 Details : The client was sending extra request flags in the root getattr
3894 request and did not allocate a reply buffer for the dir EA.
3897 Frequency : occasional, higher if lots of files are accessed by one client
3898 Bugzilla : 6159, 6097
3899 Description: Client trips assertion regarding lsm mismatch/magic
3900 Details : While revalidating inodes the VFS looks up inodes with ifind()
3901 and in rare cases can find an inode that is being freed.
3902 The ll_test_inode() code will free the lsm during ifind()
3903 when it finds an existing inode and then the VFS later attaches
3904 this free lsm to a new inode.
3908 Bugzilla : 6422, 7030
3909 Description: MDS deadlock between mkdir and client eviction
3910 Details : Creating a new file via mkdir or mknod (starting a transaction
3911 and getting the ns lock) can deadlock with client eviction
3912 (gets ns lock and trying to finish a synchronous transaction).
3915 Frequency : occasional
3916 Description: While starting a server, the fsfilt_ext3 module could not be
3918 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3919 kernels. Previously, lconf would still use the ext3 name
3920 when trying to load modules. Now, it will correctly use
3921 ext3 on 2.4 and ldiskfs on 2.6.
3923 Severity : enhancement
3924 Description: The default stripe count has been changed to 1
3925 Details : The interpretation of the default stripe count (0, to lfs
3926 or lmc) has been changed to mean striping across a single
3927 OST, rather than all available. For general usage we have
3928 found a stripe count of 1 or 2 works best.
3930 Severity : enhancement
3931 Description: Add support for compiling against Cray portals.
3932 Details : Conditional compiling for some areas that are different
3936 Frequency : occasional
3937 Bugzilla : 6409, 6834
3938 Description: Creating files with an explicit stripe count may lead to
3939 a failed assertion on the MDS
3940 Details : If some OSTs are full or unavailable, creating files may
3941 trigger a failed assertion on the MDS. Now, Lustre will
3942 try to use other servers or return an error to the
3946 Frequency : occasional
3948 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3949 caused lock timeout during readahead (since 1.4.2).
3950 Details : Processes doing readahead might match a lock that hasn't been
3951 granted yet if there are overlapping and conflicting lock
3952 requests. The readahead process waits on ungranted lock
3953 (original lock is CBPENDING), while OST waits for that process
3954 to cancel CBPENDING read lock and eventually evicts client.
3956 Severity : enhancement
3958 Description: Initial enabling of flock support for clients
3959 Details : Implements fcntl advisory locking and file status functions.
3960 This feature is provided as an optional mount flag (default
3961 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3962 locking are implemented yet, and those that are are not guaranteed
3963 to be completely correct in production environments.
3964 mount -t lustre -o [flock|noflock] ...
3967 Frequency : occasional
3969 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3970 assertion in the ext3 JBD (journaling) layer.
3971 Details : The b_committed_data struct is protected by the big kernel lock
3972 in 2.4 kernels, serializing journal_commit_transaction() and
3973 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3974 access to this struct. In 2.6 kernels there is finer grained
3975 locking to improve SMP performance of the JBD layer.
3979 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3984 Description: merge and cleanup kernel patches.
3987 -----------------------------------------------------------------------------
3989 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3991 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3992 versions of Lustre. Please update all servers and clients to
3993 version 1.4.2 or later at the same time. You must also run
3994 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3995 to update the configuration logs.
3997 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3998 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3999 - swab llog records of type '0' so we get proper header size/idx (5861)
4000 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4001 - fix rename of one directory over another leaking an inode (5953)
4002 - avoid SetPageDirty on 2.6 (5981)
4003 - don't re-add just-being-destroyed locks to the waiting list (5653)
4004 - when creating new directories, inherit the parent's custom
4005 striping settings if present parent (3048)
4006 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4007 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4008 - don't hold client locks on temporary worklist from l_lru (5666)
4009 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4010 - automatically evict dead clients (5921)
4011 - Update file size properly in create+truncate+fstat case (6196)
4012 - Do not unhash mountpoint dentries, do not allow removal of
4014 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4015 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4016 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4017 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4018 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4019 - Correctly handle reads of files with no objects (6243)
4020 - lctl recover will also mark a device active if deactivate used (5933)
4022 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4023 - allow --write-conf on an MDS with different nettype than client (5619)
4024 - don't write config llogs to MDS for mounts not from that MDS (5617)
4025 - lconf should create multiple TCP connections from a client (5201)
4026 - init scripts are now turned off by default; run chkconfig --on
4027 lustre and chkconfig --on lustrefs to use them
4028 - upcalls are no longer needed for clients to recover to failover
4030 - add --abort-recovery option to lconf to abort recovery on device
4032 - add support for an arbitrary number of OSTs (3026)
4033 - Quota support protocol changes.
4034 - forward compatibility changes to wire structs (6007)
4035 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4036 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4037 - improved "lustre status" script
4038 - initialize blocksize for non-regular files (6062)
4039 - added --disable-server and --disable-client configure options (5782)
4040 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4041 - Vanilla 2.4.29 support
4042 - increase maximum number of obd devices to 520 (6242)
4043 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4044 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4045 SLES 9 2.6.5 kernels
4046 - limit stripes per file to 160 (the maximum EA size) (6093)
4048 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4051 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4052 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4053 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4054 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4055 - small ext3 extents cleanups and fixes (5733)
4056 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4057 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4058 - avoid needless client->OST connect, fix handle mismatch (5317)
4059 - fix DLM error path that led to out-of-sync client, long delays (5779)
4060 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4061 - fix several locking issues related to i_size (5492,5624,5654,5672)
4062 - don't move pending lock onto export if it is already evicted (5683)
4063 - fix kernel oops when creating .foo in unlinked directory (5548)
4064 - fix deadlock in obdfilter statistics vs. object create (5811)
4065 - use time_{before,after} to avoid timer jiffies wrap (5882)
4066 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4067 - Do not leak request if server was not able to process it (5154)
4068 - If mds_open unable to find parent dir, make that negative lookup(5154)
4069 - don't create new directories with extent-mapping (5909, 5936)
4071 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4072 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4073 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4074 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4075 - support for RHEL4 kernel on i686 (5773)
4076 - provide error messages when incompatible logs are encountered (5898)
4078 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4079 * version 1.4.0.10 (1.4.1 release candidate 1)
4081 - don't keep a lock reference when lock is not granted (4238)
4082 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4083 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4084 - drop import inflight refcount on signal_completed_replay error (5255)
4085 - unlock page after async write error during send (3677)
4086 - handle missing objects in filter_preprw_read properly (5265)
4087 - no transno return for symlink open, don't save no-trasno open (3440)
4088 - don't try to complete elan receive that already failed (4012)
4089 - free RPC server reply state on error (5406)
4090 - clean up thread from ptlrpc_start_thread() on error (5160)
4091 - readahead could read extra page into cache that wasn't ejected (5388)
4092 - prevent races in class_attach/setup/cleanup/detach (5260)
4093 - don't dereference de->d_inode after l_dput of de (5458)
4094 - use "int" for stripe value returned from lock_to_stripe (5544)
4095 - mballoc allocation and error-checking fixes in 2.6 (5504)
4096 - block device patches to fix I/O request sizes in 2.6 (5482)
4097 - look up hostnames for IB nals (5602)
4098 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4099 - don't start multiple acceptors for the same port (5277)
4100 - fix incorrect LASSERT in mds_getattr_name (5635)
4101 - export a proc file for general "ping" checking (5628)
4102 - fix "lfs check" to not block when the MDS is down (5628)
4104 - service request history (4965)
4105 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4106 - create an "evict_client" /proc entry on OSTs, like the MDS has
4107 - fix mount usage message, return errors per mount(8) (5168)
4108 - change grep [] to grep "[]" in tests so they work in more UMLs
4109 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4110 - remove ancient LOV_MAGIC_V0 EA support (5047)
4111 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4112 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4113 - allow readahead allocations to fail when low on memory (5383)
4114 - mmap locking landed again, after considerable improvement (2828)
4115 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4117 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4120 - send OST transaction number in read/write reply to free req (4966)
4121 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4122 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4123 - the watchdog thread now runs as interruptible (5246)
4124 - flock/lockf fixes (but it's still disabled, pending 5135)
4125 - don't use EXT3 constants in llite code (5094)
4126 - memory shortage at startup could cause assertion (5176)
4128 - reorganization of lov code
4129 - single portals codebase
4131 - add extents/mballoc support (5025)
4132 - direct I/O reads in the obdfilter (4048)
4133 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4135 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4138 - send OST transaction number in read/write reply to free req (4966)
4139 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4140 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4141 - the watchdog thread now runs as interruptible (5246)
4142 - handle missing objects in filter_preprw_read properly (5265)
4143 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4144 - drop import inflight refcount on signal_completed_replay error (5255)
4145 - unlock page after async write error during send (3677)
4146 - return original error code on reconstructed replies (3761)
4147 - no transno return for symlink open, don't save no-trasno open (3440)
4149 - add pid to ldlm debugging output (4922)
4150 - bump the watchdog timeouts -- we can't handle 30sec yet
4151 - extra debugging for orphan dentry/inode bug (5259)
4153 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4156 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4157 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4158 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4159 - fix NULL dereference in /proc/sys/portals/routes (4827)
4160 - allow failed mdc_close() operations to be interrupted (4561)
4161 - stop precreate on OST before MDS would time out on it (4778)
4162 - don't send partial-page writes before EOF from client (4410)
4163 - discard client grant for sub-page writes on large-page clients (4520)
4164 - don't free dentries not owned by NFS code, check generation (4806)
4165 - fix lsm leak if mds_create_objects() fails (4801)
4166 - limit debug_daemon file size, always print CERROR messages (4789)
4167 - use transno after validating reply (3892)
4168 - process timed out requests if import state changes (3754)
4169 - update mtime on OST during writes, return in glimpse (4829)
4170 - add mkfsoptions to LDAP (4679)
4171 - use ->max_readahead method instead of zapping global ra (5039)
4172 - don't interrupt __l_wait_event() during strace
4174 - add software watchdogs to catch hung threads quickly (4941)
4175 - make lustrefs init script start after nfs is mounted
4176 - fix CWARN/ERROR duplication (4930)
4177 - return async write errors to application if possible (2248)
4178 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4179 - print NAL number in %x format (4645)
4180 - update barely-supported suse-2.4.21-171 series (4842)
4181 - support for sles 9 %post scripts
4182 - support for building 2.6 kernel-source packages
4183 - support for sles km_* packages
4185 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4188 - ignore -ENOENT errors in osc_destroy (3639)
4189 - notify osc create thread that OSC is being cleaned up (4600)
4190 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4191 - reconstruct ost_handle() like mds_handle() (4657)
4192 - create a new thread to do import eviction to avoid deadlock (3969)
4193 - let lconf resolve symlinked-to devices (4629)
4194 - don't unlink "objects" from directory with default EA (4554)
4195 - hold socknal file ref over connect in case target is down (4394)
4196 - allow more than 32000 subdirectories in a single directory (3244)
4197 - fix blocks count for O_DIRECT writes (3751)
4198 - OST returns ENOSPC from object create when no space left (4539)
4199 - don't send truncate RPC if file size isn't changing (4410)
4200 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4201 - bind to privileged port in socknal and tcpnal (3689)
4203 - rate limit CERROR/CWARN console message to avoid overload (4519)
4204 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4205 - basic mmap support (3918)
4206 - kernel patch series update from b1_4 (4711)
4208 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4211 - avoid crash during MDS cleanup with OST shut down (2775)
4212 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4213 - don't use bad inodes on the MDS (3744)
4214 - dynamic object preallocation to improve recovery speed (4236)
4215 - don't hold spinlock over lock dumping or change debug flags (4401)
4216 - don't zero obd_dev when it is force cleaned (3651)
4217 - print grants to console if they go negative (4431)
4218 - "lctl deactivate" will stop automatic recovery attempts (3406)
4219 - look for existing locks in ldlm_handle_enqueue() (3764)
4220 - don't resolve lock handle twice in recovery avoiding race (4401)
4221 - revalidate should check working dir is a directory (4134)
4223 - don't always mark "slow" obdfilter messages as errors (4418)
4225 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4228 - don't close LustreDB during write_conf until it is done (3860)
4229 - fix typo in lconf for_each_profile (3821)
4230 - allow dumping logs from multiple threads at one time (3820)
4231 - don't allow multiple threads in OSC recovery (3812)
4232 - fix debug_size parameters (3864)
4233 - fix mds_postrecov to initialize import for llog ctxt (3121)
4234 - replace config semaphore with spinlock (3306)
4235 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4236 - don't allow enqueue to complete on a destroyed export (3822)
4237 - down write_lock before checking llog header bitmap (3825)
4238 - recover from lock replay timeout (3764)
4239 - up llog sem before sending rpc (3652)
4240 - reduce ns lock hold times when setting kms (3267)
4241 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4242 - fix NULL deref and obd_dev leak on setup error (3312)
4243 - replace some LBUG about llog ops with error handling (3841)
4244 - don't match INVALID dentries from d_lookup and spin (3784)
4245 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4246 - fix invalid assertion in ptlrpc_set_wait (3880)
4248 - add libwrap support for the TCP acceptor (3996)
4249 - add /proc/sys/portals/routes for non-root route listing (3994)
4250 - allow setting MDS UUID in .xml (2580)
4251 - print the stack of a process that LBUGs (4228)
4253 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4256 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4257 - make sure to unset replay flag from failed open requests (3440)
4258 - if default stripe count is 0, use OST count for inode size (3636)
4259 - update parent mtime/ctime on client for create/unlink (2611)
4260 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4261 - free recovery state on server during a forced cleanup (3571)
4262 - unregister_reply for resent reqs (3063)
4263 - loop back devices mounting and status check on 2.6 (3563)
4264 - fix resource-creation race that can provoke i_size == 0 (3513)
4265 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4266 - more debugging for page-accounting assertion (3746)
4267 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4268 - avoid deadlock after precreation failure (3758)
4269 - fix race and lock order deadlock in orphan handling (3450, 3750)
4270 - add validity checks when grabbing inodes from l_ast_data (3599)
4272 - add /proc/.../recovery_status to obdfilter (3428)
4273 - lightweight CDEBUG infrastructure, debug daemon (3668)
4274 - change default OSC RPC parameters to be better on small clusters
4275 - turn off OST read cache for files smaller than 32MB
4276 - install man pages and include them in rpms (3100)
4277 - add new init script for (un)mounting lustre filesystems (2593)
4278 - run chkconfig in %post for init scripts (3701)
4279 - drop scimac NAL (unmaintained)
4281 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4284 - clean kiobufs before and after use (3485)
4285 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4286 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4287 - change init script to fail more clearly if not run as root (1528)
4288 - allow clients to reconnect during replay (1742)
4289 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4290 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4291 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4292 - fix cond_resched (3554)
4293 - only evict unfinished clients after recovery (3515)
4294 - allow bulk resend, prevent data loss (3570)
4295 - dynamic ptlrpc request buffer allocation (2102)
4296 - don't allow unlinking open directory if it isn't empty (2904)
4297 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4298 - remove extraneous obd dereference causing LASSERT failure (3334)
4299 - don't use get_cycles() when creating temp. files on the mds (3156)
4300 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4301 - handle EEXIST for set-stripe, set proper directory name (3336)
4303 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4304 - fix ksocknal_fmb_callback() error messages (2918)
4306 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4309 - don't copy lvb into (possibly NULL) reply on error (2983)
4310 - don't deref dentry after dput, don't free lvb on error (2922)
4311 - use the kms to determine writeback rpc length (2947)
4312 - increment oti_logcookies when osc is inactive (2948)
4313 - update client's i_blocks count via lvb messages (2543)
4314 - handle intent open/close of special files properly (1557)
4315 - mount MDS with errors=remount-ro, like obdfilter (2009)
4316 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4317 - don't use cancelling-locks' kms values (2947)
4318 - use highest lock extent for kms, not last one (2925)
4319 - don't dereference ERR_PTR() dentry in error handling path (3107)
4320 - fix thread race in portals_debug_dumplog() (3122)
4321 - create lprocfs device entries at setup instead of at attach (1519)
4322 - common AST error handler, don't evict client on completion race (3145)
4323 - zero nameidata in detach_mnt in 2.6 (3118)
4324 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4325 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4326 - initialize RPC timeout timer earlier for 2.6 (3219)
4327 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4328 - print nal/nid for unknown nid (3258)
4329 - additional checks for oscc recovery before doing precreate (3284)
4330 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4331 - don't crash in mdc_close for bad permissions on open (3285)
4332 - zero i_rdev for non-device files (3147)
4333 - clear page->private before handing to FS, better assertion (3119)
4334 - tune the read pipeline (3236)
4335 - fix incorrect decref of invalidated dentry (2350)
4336 - provide read-ahead stats and refine rpc in flight stats (3328)
4337 - don't hold journal transaction open across create RPC (3313)
4338 - update atime on MDS at close time (3265)
4339 - close LDAP connection when recovering to avoid server load (3315)
4340 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4341 - don't leak open file on MDS after open resend (3325)
4342 - serialize filter_precreate and filter_destroy_precreated (3329)
4343 - loop device shouldn't call sync_dev() for nul device (3092)
4344 - clear page cache after eviction (2766)
4345 - resynchronize MDS->OST in background (2824)
4346 - refuse to mount the same filesystem twice on same mountpoint (3394)
4347 - allow llmount to create routes for mounting behind routers (3320)
4348 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4349 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4350 - fix rare problem with rename on htree directories (3417)
4352 - allow default OST striping configuration per directory (1414)
4353 - fix compilation for qswnal for 2.6 kernels (3125)
4354 - increase maximum number of MDS request buffers for large systems
4355 - change liblustreapi to be useful for external progs like lfsck (3098)
4356 - increase local configuration timeout for slow disks (3353)
4357 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4359 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4362 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4363 - don't overwrite extent policy data in reply if lock was blocked (2901)
4364 - drop filter export grants atomically with removal from device (2663)
4365 - del obd_self_export from work_list in class_disconnect_exports (2908)
4366 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4367 - swab reply message in mdc_close, other PPC fixes (2464)
4368 - fix destroying of named logs (2325)
4369 - overwrite old logs when running lconf --write_conf (2264)
4370 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4371 - fix race in target_handle_connect (2898)
4372 - mds_reint_create() should take same inode create lock (2926)
4373 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4374 - don't close files for self_export to avoid uninitialized obd (2936)
4375 - allow MDS with the same name as client node (2939)
4376 - hold dentry reference for closed log files for unlink (2325)
4377 - reserve space for all logs during transactions (2059)
4378 - don't evict page beyond end of stripe extent (2925)
4379 - don't oops on a deleted current working directory (2399)
4380 - handle hard links to targets without a parent properly (2517)
4381 - don't dereference NULL lock when racing during eviction (2867)
4382 - don't grow lock extents when lots of conflicting locks (2919)
4384 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4387 - account for cache space usage on clients to avoid data loss (974)
4388 - lfsck support in lustre kernel code (2349)
4389 - reduce journal credits needed for BRW writes (2370)
4390 - orphan handling to avoid losing space on client/server crashes
4391 - ptlrpcd can be blocked, stopping ALL progress (2477)
4392 - use lock value blocks to assist in proper KMS, faster stat (1021)
4393 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4394 - recovery for initial connections (2355)
4395 - fixes for mds_cleanup_orphans (1934)
4396 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4397 - block all file creations until orphan recovery completes (1901)
4398 - client remove rq_connection from request struct (2423)
4399 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4400 - recovery timer race (2670)
4401 - mdc_close recovey bug (2532)
4402 - ptlrpc cleanup bug (2710)
4403 - mds timeout on local locks (2588)
4404 - namespace lock held during RPCs (2431)
4405 - handle interrupted sync write properly (2503)
4406 - don't try to handle a message that hasn't been replied to (2699)
4407 - client assert failure during cleanup after abort recovery (2701)
4408 - leak mdc device after failed mount (2712)
4409 - ptlrpc_check_set allows timedout requests to complete (2714)
4410 - wait for inflight reqs when ptlrpcd finishes (2710)
4411 - make sure unregistered services are removed from the srv_list
4412 - reset bulk XID's when resending them (caught by 1138 test)
4413 - unregister_bulk after timeout
4414 - fix lconf error (2694)
4415 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4416 - readahead locks pages, leaves pending causing memory pressure (2673)
4417 - increase OST request buffers to 4096 on large machines (2729)
4418 - fix up permission of existing directories in simple_mkdir (2661)
4419 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4420 - don't assign transno to errored transactions (2742)
4421 - don't delete objects on OST if given a bogus objid from MDS (2751)
4422 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4423 - if rq_no_resend, then timeout request after recovery (2432)
4424 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4425 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4426 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4427 - MDS recovery shouldn't delete objects that it has given out (2730)
4428 - if enqueue arrives after completion, don't clobber LVB (2819)
4429 - don't unlock pages twice when trigger_group_io returns error (2814)
4430 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4431 - don't write pages to disk if there was an error (1450)
4432 - don't ping imports that have recovery disabled (2676)
4433 - take buffered bytes into account when balancing socknal conn (2817)
4434 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4435 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4436 - remove little-endian swabbing of llog records (1987)
4437 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4438 - retry reposting request buffers if they fail (1191)
4439 - grow extent at grant time to avoid granting a revoked lock (2809)
4440 - lock revoke doesn't evict page if covered by a second lock (2765)
4441 - disable VM readahead to avoid reading outside lock extents (2805)
4443 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4444 - updated kernel patches for hp-2.4.20 kernel (2681)
4446 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4449 - fix truncated write corruption (2366)
4450 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4452 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4453 - ensure len(uuid) < 37 in lmc (1171)
4454 - fix ia64 OOPS in llog_test (2255)
4455 - zero end of page at obdfilter for partial page writes (2648)
4456 - don't leave stale dentries around after renames (bug 2428)
4457 - fix timeouts when evicting a client with a single lock held (2642)
4458 - set deadline for the initial HELLO message to drain (2634)
4459 - print out dotted-quad IP addresses in the socknal (2302)
4461 - additional debugging for MDS client eviction problem (2443)
4462 - fix mkfsoptions support for osts (2603, 2604)
4464 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4467 - add series for the vanilla 2.6.0 kernel
4468 - add series for the vanilla 2.4.24 kernel
4469 - add series for a cray x86/64 UL kernel drop
4470 - fix xattr patches for the vanilla 2.4.19 series
4472 - generate true UUIDs in lmc (1171)
4473 - have portals stack dumping break in UML (2466)
4474 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4475 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4476 - wake callback waiting threads on client eviction (2460)
4477 - Add --ptldebug and --subsystem to lmc (1719)
4478 - update assertion to allow safe interrupt allocation
4479 - set rq_no_resend for cancel requests (2432)
4480 - recalculate ptlrpcd timeout after resend (2494)
4481 - call vfs_rmdir when removing pending directories (2368)
4482 - fix renaming a file to itself (2429)
4483 - lmc creates a default one-stripe lov (2454)
4484 - expand procfs space to handle large clusters (2326)
4485 - increase UML stack to avoid overflow
4486 - update lconf's list of debug and subsystem masks
4487 - fix lfs find --obd (2510)
4488 - /proc tunable for disabling filter read caching (2591)
4489 - stop rpm packages from altering slapd.conf (2301)
4490 - disable nagle in the socknal under 0conf (2578)
4491 - choose mds inode size based on stripe count (2572)
4492 - fix kernel-source rpm problems (2516)
4494 - add --disable-doc to avoid pdf generation (2421)
4495 - update documentation, tests, type-os, comments
4496 - avoid format warnings on ia64
4497 - remove the TOE NAL
4498 - tiny code cleanups by removing unused fields
4500 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4503 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4504 - remove the most blatant lies from BUILDING file (2371)
4505 - change default debug level to reasonable production setting
4506 - reduce client side cache size to reduce cache flush time
4507 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4508 - make TCP zerocopy and pinger support enabled by default (2476)
4509 - sync writes completed after process exits caused crashes (2319)
4510 - maintain correct mount count on the MDS (2356)
4511 - backout 1557, because 2316 wasn't really fixed
4512 - better file I/O statistics gathering in /proc
4513 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4514 - another kernel patch to fix zero-copy TCP function export
4515 - don't take duplicate lock when processing re-sent getattr (2420)
4516 - lctl uses obd_self_export instead of creating new conn (2353)
4517 - MDS/OST recovery case which requires object creation asserted (2425)
4518 - move lfs from /usr/sbin to /usr/bin in packages
4519 - fix race between mds_client_add and mds_client_free (2417)
4520 - use kmalloc instead of slabs in portals (2430)
4521 - don't create duplicate records when a failover MDS is present (2442)
4522 - remove unnecessary mount age check (2332)
4523 - don't remove directory inodes from locks prematurely (2451)
4524 - don't break if MDS service name is the same as hostname (2103)
4525 - fix races in client write RPC generation when cache full (2482)
4527 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4530 - remove now-unused request->rq_obd (278)
4531 - if an allocation fails, print out how much memory we've used (1933)
4532 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4533 - add the "configurable stack size" patch to most series files (1256)
4534 - ability to write large log records, for 100+ OST configs (2306)
4535 - fix NULL deref when filter_prep fails (2314)
4536 - fix operator precedence error in filter_sync
4537 - dynamic allocation of socknal TX descriptors (2315)
4538 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4539 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4540 - ensure that CWARN messages in Portals always get to the syslog
4541 - __init/__exit are not for prototype decls (ldlm_init/exit)
4542 - x86-64 compile warning fixes
4543 - fix gateway LMC keyword conflict (2318)
4544 - fix MDS lock inversions in getattr/reint paths (1844)
4545 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4546 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4547 - DLM race condition prevented some lock evictions (2328)
4548 - ENOMEM detection and retry on socknal sends (2230)
4549 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4550 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4551 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4552 - better error messages when a client is rejected during recovery (1505)
4553 - avoid cancelling locks which were never granted, after failure (2330)
4554 - fix i_sem/journal inversion in mds_client_add (2333)
4555 - fix truncate/getattr lock cycle deadlock (2334)
4556 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4557 - fix two rare exit paths which could leak an l_lock() ref (2321)
4558 - fencepost error in MDS/OST orphan recovery (2226)
4559 - make log record alignment 8 bytes (1988)
4560 - lstripe now fails when requested offset > ost_count (2237)
4561 - ensure that all kernel series have a complete list.h (1607)
4562 - fix crashes in special-file operations (2316)
4563 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4565 - allow configurable automake binary, for testing new versions
4566 - small update to the lfs documentation
4568 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4570 * fix negative export reference count in fsfilt_sync (2312)
4572 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4573 * release candidate 0.9.1
4575 - orphans are moved into the PENDING directory for possible recovery
4576 - replayed opens now open by fid for orphan/rename safety (1042)
4577 - last close of an orphan inode generates a transno (683)
4578 - chdir() and mount() now pin the directory entry (1020)
4579 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4580 - discard very old requests without processing them (1502)
4581 - remove some common, well-understood CERRORs (1505)
4582 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4583 - clear "grant" flags in OST replies until OST grant code lands (1644)
4584 - fix read performance by not clobbering i_blksize on client (1598)
4585 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4586 - make LRU size a /proc tunable, clears locks when reduced (707)
4587 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4588 - server locks take references on exports now (1558)
4589 - build fixes for 2.4.20-rh trees (1663)
4590 - return an error from lov_create if all OSCs are inactive (1751)
4591 - fix import levels when a reconnect happens without a timeout (1597)
4592 - exit early from mds_open if we get a lookup error (1749)
4593 - partial page read at EOF wouldn't wait for disk before sending (1642)
4594 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4595 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4596 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4597 - write/truncate lock inversion (1639)
4598 - Don't auto-load obdclass, portals modules during cleanup (1495)
4599 - fix timestamps from jumping to "now" (1763)
4600 - extra journal assertions (1648)
4601 - add an extra multiunlink test (1771)
4602 - fix read_record/write_record API (1776)
4603 - fix leak of offset_extent, possible incorrect i_size later (1772)
4604 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4605 - Debugging for the kqswnal_get_idle_tx problems (1820)
4606 - Allow recovery to be attempted multiple times (1536)
4607 - Write out MDS last_rcvd file after it is first created (1600)
4608 - Fix tx_descriptor leak in failed transmit situations (1827)
4609 - ext3 journaling fixes for assertion failure after IO error (1871)
4610 - class_export_put() on freed export after completion AST error (1896)
4611 - Fix revalidate looping in VFS (1322)
4612 - Don't access a freed export during MDS_REINT timeout (1521)
4613 - Add open-unlink recovery support on the MDS (1673,1764)
4614 - Return an error if no MDS data was read from last_rcvd (1946)
4615 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4616 - Don't setup a disk device that doesn't match exported UUID (317)
4617 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4618 - avoid committing NULL handle in force close
4619 - local.sh is now a one-stripe LOV configuration
4620 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4621 - fix ext3 htree duplicate directory entry corruption (1516)
4622 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4623 - update to new LOV EA format (2097)
4624 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4626 2003-06-15 Phil Schwan <phil@clusterfs.com>
4629 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4630 - per-import/export recovery handling (958, 931, 959)
4631 - multiple last-rcvd slots, for serving multiple FSes (949)
4632 - connections are again shared between multiple imp/exports (963, 964)
4633 - "umount -f" would hang if any requests needed to be sent (393, 978)
4634 - avoid pinning large req buffer by copying for queued messages (989)
4635 - add "uuid" to "lctl device" command to help upcalls (991)
4636 - "open" RPCs with transnos would confuse recovery counters (1037)
4637 - do proper endian conversion of all wire messages (288, 340, 891)
4638 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4639 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4640 - fix ptlrpc_connection leak in target_handle_connect (1174)
4641 - fix import refcounting bug in OST and MDS cleanup (1134)
4642 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4643 - fix very unlikely obd_types race condition (501)
4644 - remove osc_open hack for echo_client (1187)
4645 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4646 - a failure in read_inode2 leads to deadlock (1139)
4647 - cancel ack-locks as soon as transaction is committed (1072)
4648 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4649 - make sure to commitrw after any preprw to avoid deadlock (1162)
4650 - failing to execute a file in a lustre FS would lock inode (1203)
4651 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4652 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4653 - obd_uuid2tgt would walk off the end of the list (1255)
4654 - on IA64 the osc would give portals incorrect bulk size (1258)
4655 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4656 - fix lock inversion caused by new llite matching code (1282)
4657 - limit the number of dirty pages on a client to 10MB (1286)
4658 - timed out locks were not being corrected cancelled (1289)
4659 - fix O_DIRECT above 4GB on IA-32 (1292)
4660 * major user-visible changes
4661 - fail out/fail over policy now controlled by the upcall (993)
4663 - add OBD_PING to check server availability and failure (954)
4664 - lustre messages are now sent in sending host order (288, 340, 891)
4665 - add eadatalen to MDS getattr reply (340)
4666 - OST read replies may contain second buffer, with per-page status (593)
4668 2003-03-11 Phil Schwan <phil@clusterfs.com>
4671 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4672 - failed open()s could cause deadlock; fixed (867, 869)
4673 - stop cancelling OST locks when files are closed (481)
4674 - overlapping XID spaces caused network corruption (851, 853)
4675 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4676 - setattr_raw added, to do single-RPC, server-side setattrs
4677 - lmc/lconf syntax change for OST UUIDs
4678 - fix crashy race condition between ptlrpc_free_req and osc_close
4679 - don't use request in mdc_enqueue if we hit a timeout (889)
4680 - don't set the inode i_size for regular files from the MDS (896)
4681 - handle out of order completion AST (842)
4682 - don't LBUG if a lock request times out after receiving AST (913)
4683 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4684 - if a bad lock AST arrives, send an error instead of dropping entirely
4685 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4686 - fix leak in bulk IO when only partially completed (899, 900, 926)
4687 - fix O_DIRECT for ia64 (55)
4688 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4689 - C-z after timeout could hang a process forever; fixed (977)
4691 - client-side I/O cache (678, 924, 929, 941, 970)
4693 - READPAGE and SETATTRs which don't take server-side locks get
4696 2003-02-11 Phil Schwan <phil@clusterfs.com>
4699 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4700 - Fix fsfilt_extN_readpage() to read a full page of directory
4701 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4702 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4703 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4704 - Most intent operations were removed, in favour of a new RPC mode
4705 that does a single RPC to the server and bypasses most of the VFS
4706 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4707 - Aggressively cancel local locks on DLM servers
4708 - mds_reint_unlink sends EA to the client if it's the last nlink.
4709 client uses that EA to unlink OST objects.
4710 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4711 - recursive symlinks were fixed (439)
4712 - fixed NULL deref in DEBUG_REQ
4713 - filter_update_lastobjid no longer calls sync, which annoyed extN
4714 - fixed multi-client small-writes to a single file problem (445)
4715 - fixed mtime updates during file writes (607)
4716 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4717 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4718 - fixed Linux OST size reporting problem (444, 656)
4719 - OST now updates object mtime with writes or setattr (607, 619)
4720 - client verifies file size before zeroing page past EOF (445)
4721 - OST now writes last allocated objid to disk with allocation (108)
4722 - LOV on echo now works (409)
4724 - mds_reint_unlink sends a new buffer, with the EA included. this
4725 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4726 set if a regular file was being unlinked, and it was the last link
4727 - use PtlGet from the target for bulk writes (315)
4728 - OST now updates object mtime with writes or setattr (607, 619)
4729 - LDLM now has a grant-time callback to revalidate locked items, if
4731 - Many MDS operations were reorganized to combat race conditions
4733 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4734 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4736 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4739 - Fully reactivate OST imports after reconnection (512, others)
4740 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4741 - More graceful error handling for truncating on dead OST (515)
4742 - Don't error out unless we're actually accessing dead stripes (474)
4743 - Fix garbage sizes when stripes are missing (410)
4744 - LRU counters were broken, causing constant lock purge (433, 432)
4745 - garbage on read from stripes with failed OSTs (441)
4746 - mark OSCs as active before reconnecting during recovery (438)
4747 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4748 - lfind did not preserve OST order in output (443)
4749 - symlinks cause hung clients, incorrect data (439)
4750 - stop dereferencing request after dropping refcount (457)
4751 - don't LASSERT(spin_is_locked) on non-SMP (455)
4752 - fixes for many rename() bugs
4753 - fstat didn't correctly synchronize attributes (399)
4754 - server must handle lock cancellation during blocking AST prep (487)
4755 - bulk descriptors were free()d too soon (511)
4756 - fix paths in lconf, which would load incorrect modules (451, 507)
4757 - fix confusing lconf 'host not found' error message (386)
4758 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4759 - fix race condition in mdc_blocking_ast() for inode access (526)
4760 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4761 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4762 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4763 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4764 - fix lstripe in several ways (kernel side) (527)
4765 - fix request leak in ldlm_cli_enqueue (262)
4766 - incorrect OSC was marked inactive after OST failure
4767 - call mds_fs_cleanup before unmounting filesystem (524)
4768 - fix races between taking ns_lock and ldlm_lock_change_resource
4769 - fix races updating LOV export open file list
4770 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4771 - fix recovery NULL deref in ldlm_cli_cancel_unused
4772 - fix some DLM races by using new hash table for lock handles (419)
4773 - permit the client to specify desired inodes, at replay
4774 - duplicate requests when we queue them for replay reintegration
4775 - fix last_rcvd offset calculation
4776 - sync after each recovered transaction, so we always make progress
4777 - never, not always, ERESTART requests without transnos
4778 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4780 - skip replay if the MDS says that the client is already connected
4781 - don't check for a recovery-enabled export to match lctl's UUID
4782 - don't INC_USE_COUNT for phantom exports
4783 - don't crash when cleaning up phantom exports (567)
4784 - don't double-finish or set replay data for errored mdc_open requests
4785 - abort requests when they time out, so we don't get old replies
4786 - send/receive replies for AST messages again
4787 - if the client says that it doesn't have the lock, cancel it on the
4789 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4790 mark it as destroyed, it will all work out when decref is called
4791 - fix module use counts (22, 581)
4793 - ASTs now expect a reply (server cancels lock on error reply)
4795 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4798 - fix many simultaneous client startup (392)
4799 - fix dentry->d_it clobbering
4800 - credentials weren't being shipped for readdir/getattr operations
4801 - remove invalid assertions triggered during some concurrent MD
4803 - proper Lustre versions added (336, 389)
4804 - fix memory leak for create error case (398)
4805 - fix LOV locking bug that would get cli/srv out of sync
4806 - fix echo client over LOV (409)
4807 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4808 - fix double-O_EXCL intent crash (424)
4809 - avoid sending multiple lock CANCELs (352)
4811 - MDS can do multi-client recovery (modulo bugs in new code)
4813 - many updates, edits, cleanups
4815 2002-11-18 Phil Schwan <phil@clusterfs.com>
4818 - fix null d_it dereference (346)
4819 - fix full OST/dbench hang (333)
4820 - fix permission problem with file removal (286)
4821 - fix removal of OSCs from LOV when they fail
4822 - fix NULL deref during bulk timeout (214)
4823 - fix problems related to multiple filesystems on one MDS (241)
4824 - fixed serious subtle metadata locking bugs
4825 - free locks on clients when inodes are removed due to memory
4827 - fix inode pointer in lock data (285)
4828 - partial support for multiple MDS on a single host (241)
4829 - data locks weren't cancelled at clear_inode time (290, 311)
4830 - intent locks could lead to unbounded lock growth (205)
4831 - added a maximum lock count, an LRU list, and a flusher
4832 - fix multiple rename (365)
4833 - properly abstracted the echo client
4834 - OSC locked 1 byte too many; fixed
4835 - rewrote brw callback code:
4836 - fixed recovery bugs related to LOVs (306)
4837 - fixed too-many-pages-in-one-write crash (191)
4838 - fixed (again) crash in sync_io_timeout (214)
4839 - probably fixed callback-related race (385)
4841 - Add capability to MDS protocol
4842 - LDLM cancellations and callbacks on different portals
4844 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4847 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4848 - defer open object destruction to close time (601981, 138)
4849 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4850 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4851 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4852 - don't call obd_setattr in ll_file_release for destroyed objects
4853 * protocol change to lustre_msg: move |version| and add |flags|
4854 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4855 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4856 * added replay of create, unlink, link and rename operations during
4857 MDS failover; recovery should be much more robust now
4858 * remove failed OSCs from LOVs (only lov_create uses this so far)
4859 * the lustre-HOWTO was brought (more) up to date (582544)
4861 2002-10-23 Phil Schwan <phil@clusterfs.com>
4864 - in-use dentries weren't being reused properly (617851)
4865 - prevent multiple LDLM setup (599178)
4866 - fix LOV size calculations for truncate (617853)
4867 - fix client handling of MDS intent errors (POSIX)
4868 - fix permission bug in lovstripe.c test (624321)
4869 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4870 - truncate past end of file could corrupt data
4871 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4872 - a race in recovery could return ETIMEDOUT to apps (623947)
4873 - building outside the source directory was fixed
4874 * the lustre-HOWTO was brought (more) up to date (582544)
4875 * major progress was made on recovery functionality
4877 2002-10-10 Phil Schwan <phil@clusterfs.com>
4880 - recovery deadlock fix
4881 - rm -rf causes LBUG fix (617817)
4882 - file open by multiple tasks fix (618962)
4883 - directory permissions bugs (602707 and 620007)
4884 - journal_stop fixed with locking (611313)
4885 - O_APPEND failures resolved (618273, perhaps 614459)
4886 - lconf PATH fix (619770)
4887 - IA64 build fix (621450)
4888 - RPC buffer sizes scale with amount of memory
4890 2002-10-01 Phil Schwan <phil@clusterfs.com>
4893 - locks would be cancelled without throwing away data pages,
4894 resulting in inconsistent data (605627)
4895 - inode attributes were not always being refreshed (605627, 612449)
4896 - lconf now continues to cleanup after lctl reports an error
4897 - MDS now enforces user permissions (602707)
4898 - lprocfs cleanup fixed, but not yet enabled (614157)
4899 - fixed infinite server hang, should a client not respond to an AST
4900 - avoid going into recovery if user calls readlink() with a buffer
4901 that's too small (613941)
4902 - AST RPCs no longer require replies (614867) -- this may be changed
4903 - don't crash server if client sends an IOV that's too big (611336)
4904 - fixed lock conversion deadlock (611892)
4905 - fixed the following of symlinks (614622)
4906 * recovery: the server can remove locks from a client that dies, other
4907 clients can make progress
4908 * more extN patch fixes
4909 * compile-time configurable ptlrpc buffer allocations
4911 - collaborative read cache document
4912 - Lustre Lite Performance CDR document-in-progress
4914 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4917 - fix typo in patch-2.4.18
4919 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4922 - clear ptlrpc request each time in handle_incoming_request()
4923 - unlink of files now destroys the object on the OST
4925 2002-09-19 Peter Braam <braam@clusterfs.com>
4927 * add hard link support
4928 * change obdfile creation method
4929 * kernel patch changed
4931 2002-09-19 Peter Braam <braam@clusterfs.com>
4934 - stack overflow bug in extN fixed
4936 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4938 * documentation updates
4939 - add man pages for config tools
4940 - update tests/README to describe testing with new config tools
4941 - finish metadata API descriptions
4942 * bug fixes and cleanups
4943 - statfs workaround for 16TB limit
4944 - LOV stripe allocation improved, can stripe on subset of OSTs
4945 - LOV file size/IO offset was wrong for files > 4GB in size
4946 - object EA data was being dropped, caused files to be unreadable
4947 - memory overflow with non-LOV OST caused memory corruption
4948 - fixed regression tests to work with new config tools, obdfilter
4949 - fixed bug when directory size became larger than 1 block
4950 - fixed bug (for single client case) when PWD was deleted
4951 - invalidate local directory pages when doing intent-based ops
4952 - avoid LDLM oops when lock callback contained bad data
4954 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4956 * documentation updates
4957 * bug fixes and cleanups
4958 - configuration tools
4961 - 64-bit compile warnings
4962 - 64-bit internal statfs data
4964 * test_brw on persistent OST devices
4966 * lprocfs (disabled)
4968 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4970 * documentation updates
4971 * bug fixes and cleanups
4972 * configuration tools
4974 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4977 * many small fixes to 0_5_4
4978 * io/network handling
4979 * thinkos in MDS operations
4981 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4984 * crucial basic fixes to 0.5.3
4985 * IOR, Iozone work over Elan
4988 2002-08-07 Phil Schwan <phil@clusterfs.com>
4989 * version 0_5_3, our first alpha
4990 * we use the new Portals iovs
4991 * documentation updates
4992 * bug fixes and cleanups
4993 * small changes in the DLM wire protocol
4995 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4996 * version 0_5_1 with some initial stability,
4997 * locking on MD and file I/O.
4998 * documentation updates
4999 * several bug fixes since 0.5.0
5000 * small changes in wire protocol
5002 2002-07-18 Phil Schwan <phil@clusterfs.com>
5004 * delivered as Lustre Light Alpha
5005 * fixed a crash after handling invalid MDS requests
5006 * fixed directory pages for architectures with non-4k pages sizes
5008 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5009 * release version v0_4_4
5010 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5012 * Fixes a number of interruption problems with OST operations.
5013 * Update documentation for portals header changes
5014 * Move all wire protocol structs/defines to lustre_idl.h
5015 * Fixes symlink length bug.
5016 * Add tcpdump to repository.
5018 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5019 * release version v0_4_3
5020 * Fixes statfs for inodes on extN.
5021 * Fixes bug in runtests which would delete /etc/hosts.
5022 * Use 64-bit object IDs wherever possible (not into VFS though)
5023 Remove ost_get_info, which is unused by lustre, and out of date.
5025 2002-07-03 Peter Braam <braam@clusterfs.com>
5026 * release version v0_4_2 Fixes a lookup error (type not passed)
5027 * move forward to head of Portals
5028 * move forward to latest Lustre kernel
5030 2002-06-25 Peter Braam <braam@clusterfs.com>
5031 * release version v0_4_1. Hopefully stable on single node use.