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: Properly propagate oinfo flags from lov to osc for statfs
18 Details : restore missing copy oi_flags to lov requests.
20 Severity : enhancement
22 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
23 : l_getgroups(8), lst(8), routerstat(8)
24 Details : included man pages for llobdstat(8), llstat(8),
25 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
27 Severity : enhancement
29 Description: configure's --enable-quota should check the
30 : kernel .config for CONFIG_QUOTA
31 Details : configure is terminated if --enable-quota is passed but
32 : no quota support is in kernel
36 Description: Remove portals compatibility
37 Details : Remove portals compatibility, not interoperable with releases
42 Description: Resolve device initialization race
43 Details : Prevent proc handler from accessing devices added to the
44 obd_devs array but yet be intialized.
46 Severity : enhancement
48 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
50 Severity : enhancement
52 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
57 Description: Using wrong pointer in osc_brw_prep_request
58 Details : Access to array[-1] can produce panic if kernel compiled with
59 CONFIG_PAGE_ALLOC enabled
63 Description: Client runs out of low memory
64 Details : Consider only lowmem when counting initial number of llap pages
68 Description: *optional* service tags registration
69 Details : if the "service tags" package is installed on a Lustre node
70 When the filesystem is mounted, a local-node service tag will
71 be created. See http://inventory.sun.com/ for more information
72 about the Service Tags asset management system.
76 Description: Kernel BUG tries to release flock
77 Details : Lustre does not destroy flock lock before last reference goes
78 away. So always drop flock locks when client is evicted and
79 perform unlock regardless of successfulness of speaking to MDS.
83 Description: add recount protection for osc callbacks, so avoid panic on shutdown
87 Description: sanity test 65a fails if stripecount of -1 is set
88 Details : handle -1 striping on filesystem in ll_dirstripe_verify
93 Description: ASSERTION(CheckWriteback(page,cmd)) failed
94 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
97 Severity : enhancement
99 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
103 Description: do not process already freed flock
104 Details : flock can possibly be freed by another thread before it reaches
105 to ldlm_flock_completion_ast.
109 Description: LBUG during stress test
110 Details : Need properly lock accesses the flock deadlock detection list.
114 Description: oops in page fault handler
115 Details : kernel page fault handler can return two special 'pages' in error case, don't
116 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
120 Description: timeout with invalidate import.
121 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
122 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
123 ptlrpc_invalidate_import and as result LASSERT.
125 Severity : enhancement
127 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
131 Description: enable MGS and MDT services start separately
132 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
133 a co-located MGS without starting the MGS, which is a complement
134 to 'nosvc' mount option.
138 Frequency : after recovery
139 Description: precreate to many object's after del orphan.
140 Details : del orphan st in oscc last_id == next_id and this triger growing
141 count of precreated objects. Set flag LOW to skip increase count
142 of precreated objects.
146 Frequency : rare, on clear nid stats
147 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
148 Details : when clean nid stats sometimes try destroy live entry,
149 and this produce panic in free.
153 Description: Stack overflow during MDS log replay
154 ease stack pressure by using a thread dealing llog_process.
158 Description: wait until IO finished before start new when do lock cancel.
159 Details : VM protocol want old IO finished before start new, in this case
160 need wait until PG_writeback is cleared until check dirty flag and
161 call writepages in lock cancel callback.
163 Severity : enhancement
165 Description: using special macro for print time and cleanup in includes.
169 Description: mds_mfd_close() ASSERTION(rc == 0)
170 Details : In mds_mfd_close(), we need protect inode's writecount change
171 within its orphan write semaphore to prevent possible races.
175 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
179 Frequency : rare, on shutdown ost
180 Description: don't hit live lock with umount ost.
181 Details : shrink_dcache_parent can be in long loop with destroy dentries,
182 use shrink_dcache_sb instead.
186 Description: don't panic with use echo client
187 Details : echo client pass NULL as client nid pointer and this produce null
192 Description: fix build on ppc32
193 Details : compile code with -m64 flag produce wrong object file for ppc32.
197 Description: add message levels for liblustreapi
201 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
202 Details : Move the 'good_osts' check before the 'total_bavail' check. This
203 will result in an -EAGAIN and in the exit call path we call
204 alloc_rr() which will with increasing aggressiveness attempt to
205 aquire precreated objects on the minimum number of required OSCs.
209 Description: Use old size assignment to avoid deadlock
210 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
211 the scheduling while holding a spinlock. We do not need locking
212 for size in ll_update_inode() because size is only updated from
213 the MDS for directories or files without objects, so there is no
214 other place to do the update, and concurrent access to such inodes
215 are protected by the inode lock.
219 Description: resolve "_IOWR redefined" build error on SLES10
223 Description: dump the memory debugging after all modules are unloaded to
224 suppress false negative in conf_sanity test 39
226 Severity : enhancement
228 Description: build kernel-ib packages for OFED 1.3 in our release cycle
233 Description: fix SLES kernel versioning
234 Details : the kernel version for our SLES 10 kernel did not include a "-"
235 before the "smp" at the end. while this was not a problem in
236 general, it did mean that software trying to use the kernel
237 version to try to detect a vendor specific kernel would fail.
238 this was most evident by the OFED build scripts.
242 Description: Don't update lov_desc members until making sure they are valid
243 Details : When updating lov_desc members via proc fs, need fix their
244 validities before doing the real update.
248 Description: don't put request into delay list while invalidate in flight.
249 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
250 import in flight. this produce timeout for invalidate and sometimes
251 can cause stale data.
255 Frequency : on ppc only
256 Description: not convert ost objects for directory because it's not exist.
257 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
261 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
263 Severity : enhancement
265 Description: Update to support 2.6.22.14 vanilla kernel.
269 Frequency : rare, on recovery
270 Description: read procfs can produce deadlock in some situation
271 Details : Holding lprocfs lock which send rpc can produce block for destroy
272 obd objects and this also block reconnect with -EALREADY. This isn't
273 fix all lprocfs bugs - but make it rare.
275 Severity : enhancement
277 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
280 Frequency : frequent on X2 node
282 Description: mdc_set_open_replay_data LBUG
283 Details : Set replay data for requests that are eligible for replay.
287 Description: lustre_mgs: operation 101 on unconnected MGS
288 Details : When MGC is disconnected from MGS long enough, MGS will evict the
289 MGC, and late on MGC cannot successfully connect to MGS and a lot
290 of the error messages complaining that MGS is not connected.
294 Frequency : on network error
295 Description: panic with double free request if network error
296 Details : mdc_finish_enqueue is finish request if any network error ocuring,
297 but it's true only for synchronus enqueue, for async enqueue
298 (via ptlrpcd) this incorrect and ptlrpcd want finish request
301 Severity : enhancement
303 Description: client-side metadata stat-ahead during readdir(directory readahead)
304 Details : perform client-side metadata stat-ahead when the client detects
305 readdir and sequential stat of dir entries therein
308 Frequency : on start mds
310 Description: Implement get_info(last_id) in obdfilter.
313 Frequency : occasional
315 Description: Correctly check stale fid, not start epoch if ost not support SOM
316 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
317 overwrited with new generated fid, but mds can anwer with one of these
318 two fids and both is not stale. setattr incorectly start epoch and
319 assume will be called done_writeting, but without SOM done_writing
323 Frequency : rare, depends on device drivers and load
325 Description: MDS or OSS nodes crash due to stack overflow
326 Details : Code changes in 1.8.0 increased the stack usage of some functions.
327 In some cases, in conjunction with device drivers that use a lot
328 of stack the MDS (or possibly OSS) service threads could overflow
329 the stack. One change which was identified to consume additional
330 stack has been reworked to avoid the extra stack usage.
333 Frequency : occasional
335 Description: Do not fail import if osc_interpret_create gets -EAGAIN
336 Details : If osc_interpret_create got -EAGAIN it immediately exits and
337 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
338 oscc_has_objects and see OSC has no objests and call
339 oscc_internal_create to resend create request.
341 Severity : enhancement
343 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
345 Severity : enhancement
347 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
350 Frequency : very rare
352 Description: Fix replay if there is an un-replied request and open
353 Details : In some cases, older replay request will revert the
354 mcd->mcd_last_xid on MDS which is used to record the client's
357 Severity : enhancement
359 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
361 Severity : enhancement
363 Description: Add rhel5 support to HEAD.
365 Serverity : enhancement
367 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
372 Description: Don't allow skipping OSTs if index has been specified.
373 Details : Don't allow skipping OSTs if index has been specified, make locking
374 in internal create lots better.
378 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
379 Details : access to bitfield in structure is always rounded to long
380 and this produce problem with not atomic change any bit.
384 Description: Lustre make rpms failed.
385 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
386 builds Lustre rpms from tarball.
389 Frequency : rare on shutdown ost
391 Description: If llog cancel was not send before clean_exports phase, this can
392 produce deadlock in llog code.
393 Details : If llog thread has last reference to obd and call class_import_put
394 this produce deadlock because llog_cleanup_commit_master wait when
395 last llog_commit_thread exited, but this never success because was
396 called from llog_commit_thread.
400 Description: allow userland application know is lost one of stripes.
401 Details : fill lvb_blocks with error code on ost and return it to
402 application if error flag found.
406 Description: NULL lov_tgts causing MDS oops
407 Details : more safe checks for NULL lov_tgts for avoid oops.
409 Severity : enhancement
411 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
415 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
416 Details : If an OST is down the MDS will hang indefinitely in
417 obd_statfs_rqset() waiting for the statfs data. While for
418 MDS QOS usage of statfs, it should not stuck in waiting.
420 Severity : enhancement
422 Description: remote_acl support
423 Details : Support ACL-based permission check for remote user.
424 Support setfacl/getfacl for remote user with the utils
425 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
426 the system "{s,g}etfacl" utils.
428 Severity : enhancement
430 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
432 Severity : enhancement
434 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
438 Description: make mgs_setparam() handle fsname containing dash
439 Details : fsname containing a dash does not work with lctl conf_param
441 Severity : enhancement
443 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
445 Severity : enhancement
447 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
451 Frequency : rare, at shutdown
452 Description: access already free / zero obd_namespace.
453 Details : if client_disconnect_export was called without force flag set,
454 and exist connect request in flight, this can produce access to
455 NULL pointer (or already free pointer) when connect_interpret
456 store ocd flags in obd_namespace.
460 Frequency : only at startup
461 Description: not alloc memory with spinlock held.
462 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
463 if any spinlock held.
465 Severity : enhancement
467 Description: make lustre randomly fail allocating memory
468 Details : Make lustre randomly failed allocating memory for testing purpose.
470 Severity : enhancement
472 Description: lost problems with lov objid file
473 Details : Fixes some scability and access to not inited memory problems
474 in work with lov objdid file.
479 Description: lfs find does not continue on file error
480 Details : Continue other files processing when a file/dir is absent.
484 Description: Inconsistent usage of lustre_pack_reply()
485 Details : Standardize the usage of lustre_pack_reply() such that it
486 always generate a CERROR on failure.
489 Frequency : occasional
491 Description: MDS hang or stay in waiting lock
492 Details : If client receive lock with CBPENDING flag ldlm need send lock
493 cancel as separate rpc, to avoid situation when cancel request
494 can't processed due all i/o threads stay in wait lock.
498 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
502 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
506 Description: Update to SLES9 kernel-2.6.5-7.287.3.
510 Description: Update to RHEL4 latest kernel.
512 Severity : enhancement
514 Description: Build SLES10 patchless client fails
515 Details : The configure was broken by run ./configure with
516 --with-linux-obj=.... argument for patchless client. When the
517 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
518 can't search header adequately. Use absolute path such as
519 -I($LINUX)/include instead.
523 Description: interrupt oig_wait produce painc on resend.
524 Details : brw_redo_request can be used for resend requests from ptlrpcd and
525 private set, and this produce situation when rq_ptlrpcd_data not
526 copyed to new allocated request and triggered LBUG on assert
527 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
528 ptlrpcd set if request is changed and can be safety changed to use
533 Description: LASSERT_{REQ,REP}SWAB macros are buggy
534 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
535 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
536 these macros and replace them with their difinitions instead.
540 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
541 Details : Update which_patch & target file for SLES10 latest kernel.
544 Description: Remove client patches from SLES 10 kernel.
545 Details : This causes SLES 10 clients to behave as patchless clients
546 even on a Lustre-patched (server) kernel.
548 Severity : enhancement
550 Description: self-adjustable client's lru lists
551 Details : use adaptive algorithm for managing client cached locks lru
552 lists according to current server load, other client's work
553 pattern, memory activities, etc. Both, server and client
554 side namespaces provide number of proc tunables for controlling
557 Severity : enhancement
559 Description: light-weight GSS support
560 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
561 and reduce performance overhead.
563 Severity : enhancement
565 Description: Linux keyring support
566 Details : Support using service of Linux keyring for Lustre GSS internal
567 context refresh/cache mechanism.
571 Description: Fix errors in lfs documentation
572 Details : Fixes man pages
576 Description: don't use GFP_* in generic Lustre code.
577 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
581 Description: obdclass is limited by single OBD_ALLOC(idarray)
582 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
586 Description: warnings with build patchless client with vanila 2.6.19 and up
587 Details : change old ctl_table style and replace ctl_table/ctl_table_header
588 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
592 Frequency : Only for SLES
593 Description: sanity_quota fail test_1
594 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
595 file, which makes TSTID[2] inunique.
599 Frequency : Only for SLES9
600 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
601 Details : Update target/ChangeLog/which_patch .
605 Frequency : Only for test_10 in sanity-quota.sh
606 Description: a bug in quota test code
607 Details : checking a test flag in a wrong place
611 Frequency : Only for test_13 in sanity-quota.sh
612 Description: a bug in quota test script
613 Details : shouldn't check overfull allocated quota
617 Description: jbd statistics
618 Details : Port older jbd statistics patch for sles10
620 Severity : enhancement
622 Description: Add printing inode info into message about error in writepage.
627 Description: reply_lock_interpret crash due to race with it and lock cancel.
628 Details : Do not replay locks that are being cancelled. Do not reference
629 locks by their address during replay, just by their handle.
633 Frequency : When flocks are used.
634 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
635 Details : Flock locks might destroy just granted lock if it could be merged
636 with another existing flock, this is done in completion handler,
637 so teach ldlm_cli_enquque_fini that this is a valid case for
643 Description: Oops in read and write path when failing to allocate lock.
644 Details : Check if lock allocation failed and return error back.
648 Description: lstripe command fails for valid OST index
649 Details : The stripe offset is compared to lov->desc.ld_tgt_count
650 instead of lov->desc.ld_active_tgt_count.
654 Description: sanity.sh failed test 103
655 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
660 Description: df doesn't work properly if diskfs blocksize != 4K
661 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
665 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
666 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
668 Severity : enhancement
670 Description: Adaptive timeouts
671 Details : RPC timeouts adapt to changing server load and network
672 conditions to reduce resend attempts and improve recovery time.
676 Description: llapi_file_create() does not allow some changes
677 Details : add llapi_file_open() that allows specifying the mode and
678 open flags, and also returns an open file handle.
682 Description: merge and cleanup kernel patches.
683 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
687 Description: Add journal checksum support.(Kernel part)
688 Details : The journal checksum feature adds two new flags i.e
689 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
690 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
691 indicates that the commit block contains the checksum for
692 the blocks described by the descriptor blocks. Now commit
693 record can be sent to disk without waiting for descriptor
694 blocks to be written to disk. This behavior is controlled
695 using JBD2_FEATURE_ASYNC_COMMIT flag.
699 Description: OSS needs mutliple precreate threads
700 Details : Add ability to start more than one create thread per OSS.
704 Description: Sanity.sh test_65k failed: setstripe should have succeeded
705 Details : execute lfs setstripe on client
709 Description: mds_obd_create error creating tmp object
710 Details : When the user sets quota on root, llog will be affected and can't
711 create files and write files.
714 Frequency : Always on ia64 patchless client, and possibly others.
716 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
717 Details : This allows the patchless client to be loaded on architectures
722 Description: RedHat Update kernel for RHEL5
723 Details : Add the kernel config file for RHEL5.
727 Description: RedHat Update kernel for RHEL5
728 Details : Modify the kernel config file more closer RHEL5.
732 Description: Build failure against Centos5 (RHEL5)
733 Details : Define PAGE_SIZE when it isn't present.
737 Description: test_7 on sanity-quota.sh can't be used on separate machines
738 Details : add facet to handle it
742 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
743 Details : with NFS, the anon dentry's parent was set to itself in
744 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
745 corresponding dentry other than use rec->ur_name.
747 Severity : enhancement
749 Description: lfs setstripe enhancement
750 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
754 Description: enable data checksumming by default
755 Details : enable checksum by default, allow --disable-checksum
756 configure option and "-o nochecksum" mount option. Checksums
757 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
761 Description: lustre support for RHEL5
762 Details : Add support for RHEL5.
766 Description: Client eviction due to failover config
767 Details : after a connection loss, the lustre client should attempt to
768 reconnect to the last active server first before trying the
769 other potential connections.
773 Description: when mds and osts use different quota unit(32bit and 64bit),
774 quota will be released repeatly.
775 Details : void sending multiple quota reqs to mds, which will keep the
776 status between the reqs.
780 Description: osts not allocated evenly to files
781 Details : change the condition to increase offset_idx
784 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
786 Description: Data corruption for OSTs that are formatted larger than 2TB
788 Details : When generating the bio request for lustre file writes the
789 sector number would overflow a temporary variable before being
790 used for the IO. The data reads correctly from Lustre (which
791 will overflow in a similar manner) but other file data or
792 filesystem metadata may be corrupted in some cases.
796 Description: Tune the kernel for good SCSI performance.
797 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
798 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
803 Description: rewrite ext2-derived code in obdclass/uuid.c
804 Details : rewrite inherited code (uuid parsing code from ext2 utils)
805 from scratch preserving functionality.
809 Description: "lfs find -obd UUID" prints directories
810 Details : "lfs find -obd UUID" will return all directory names instead
811 of just file names. It is incorrect because the directories
812 do not reside on the OSTs.
816 Description: use i_size_read and i_size_write in 2.6 port
817 Details : replace inode->i_size access with i_size_read/write()
820 Frequency : when removing large files
822 Description: scheduling issue during removal of large Lustre files
823 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
824 It causes scheduling issues when removing large files (17TB in the
828 Frequency : only with liblustre clients on XT3
830 Description: evictions taking too long
831 Details : allow llrd to evict clients directly on OSTs
834 Frequency : only on ppc
836 Description: /proc/fs/lustre/devices broken on ppc
837 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
838 We should make sure the type of 'index' is loff_t and then cast
839 explicitly as needed below. Do not assign an explicitly cast
843 Frequency : only for rhel5
845 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
846 Details : Modify the target file & which_kernel.
848 Severity : enhancement
850 Description: omit set fsid for export NFS
851 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
852 as FS_REQUIRES_DEV for avoid problems with generate fsid.
856 Frequency : Always, for kernels after 2.6.16
857 Description: Fix warning idr_remove called for id=.. which is not allocated.
858 Details : Last kernels save old s_dev before kill super and not allow
859 to restore from callback - restore it before call kill_anon_super.
863 Description: buffer overruns could theoretically occur
864 Details : llapi_semantic_traverse() modifies the "path" argument by
865 appending values to the end of the origin string, and a
866 overrun may occur. Adding buffer overrun check in liblustreapi.
870 Description: Fix error on 'ls .' at the top of the Lustre mount.
871 Details : Don't revalidate dentry if it is a root dentry.
875 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
876 Details : Modify vm-tunables-rhel4.patch.
880 Description: Kernel config for 2.6.18-vanilla.
881 Details : Modify targets/2.6-vanilla.target.in.
882 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
883 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
884 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
885 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
889 Description: improve handling recoverable errors
890 Details : if request processig with error which can be recoverable on server
891 request should be resend, otherwise page released from cache and
896 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
897 Details : Modify target file & which_patch.
898 A flaw was found in the IA32 system call emulation provided
899 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
900 value could be stored in the %RAX register, which could trigger an
901 out-of-bounds system call table access. An untrusted local user
902 could exploit this flaw to run code in the kernel
903 (ie a root privilege escalation). (CVE-2007-4573).
907 Description: O_DIRECT bypasses client statistics.
908 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
909 read_ahead_stats were updated. copied the stats section from
910 osc_send_oap_rpc() into async_internal().
914 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
918 Description: change order of libsysio includes
919 Details : '#include sysio.h' should always come before '#include xtio.h'
923 Description: handle "serious error: objid * already exists" more gracefully
924 Details : If LAST_ID value on disk is smaller than the objects existing in
925 the O/0/d* directories, it indicates disk corruption and causes an
926 LBUG(). If the object is 0-length, then we should use the existing
927 object. This will help to avoid a full fsck in most cases.
929 Severity : enhancement
931 Description: adapt the lustre_config script to support the upgrade case
932 Details : Add "-u" option for lustre_config script to support upgrading 1.4
933 server targets to 1.6 in parallel.
937 Description: To avoid grant space > avaible space when the disk is almost
938 full. Without this patch you might see the error "grant XXXX >
939 available" or some LBUG about grant, when the disk is almost
941 Details : In filter_check_grant, for non_grant cache write, we should
942 check the left space by if (*left > ungranted + bytes), instead
943 of (*left > ungranted), because only we are sure the left space
944 is enough for another "bytes", then the ungrant space should be
945 increase. In client, we should update cl_avail_grant only there
946 is OBD_MD_FLGRANT in the reply.
950 Description: Update RHEL 4 kernel to fix local root privilege escalation.
951 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
952 described in CVE-2007-4573. This problem could allow untrusted
953 local users to gain root access.
956 Frequency : when using O_DIRECT and quotas
958 Description: Incorrect file ownership on O_DIRECT output files
959 Details : block usage reported by 'lfs quota' does not take into account
960 files that have been written with O_DIRECT.
965 Description: touch file failed when fs is not full
966 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
967 otherwise we can get ENOSP while fs is not full.
971 Description: parallel lock callbacks
972 Details : Instead of sending blocking and completion callbacks as separated
973 requests, adding them to a set and sending in parallel.
976 Frequency : only for Cray XT3
977 Bugzilla : 12829/13455
978 Description: Changing primary group doesn't change the group lustre assigns to
980 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
981 overridden with the primary group provided by the group upcall,
982 whereas the supplied fsgid can be trusted if it is in the list of
983 supplementary groups returned by the group upcall.
985 Severity : enhancement
987 Description: Allow masking D_WARNING, D_ERROR messages from console
988 Details : Console messages can now be disabled via lnet.printk.
992 Description: User code with malformed file open parameter crashes client node
993 Details : Before packing join_file req, all the related reference should be
994 checked carefully in case some malformed flags cause fake
995 join_file req on client.
999 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1000 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1001 mode is equal to the granted mode. However, it should also check
1002 that LDLM_FL_CP_REQD is not set.
1006 Description: Heavy nfs access might result in deadlocks
1007 Details : After ELC code landed, it is now improper to enqueue any mds
1008 locks under och_sem, because enqueue might want to decide to
1009 cancel open locks for same inode we are holding och_sem for.
1013 Description: Client eviction while running blogbench
1014 Details : A lot of unlink operations with concurrent I/O can lead to a
1015 deadlock causing evictions. To address the problem, the number of
1016 oustanding OST_DESTROY requests is now throttled to
1017 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1018 callbacks are processed in priority.
1022 Description: enable ACLs on MDS by default
1023 Details : ACLs must be enabled on MDS by default.
1026 Frequency : PPC/PPC64 only
1028 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1029 Details : fix duplicated definitions between asm-ppc64/types.h and
1030 lustre_types.h on PPC.
1033 Frequency : PPC/PPC64 only
1035 Description: asm-ppc/segment.h does not exist
1036 Details : fix compile issue on PPC.
1040 Description: better handle error messages in extents code
1043 Frequency : RHEL4 only
1045 Description: mkfs is very slow on IA64/RHEL4
1046 Details : A performance regression has been discovered in the MPT Fusion
1047 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1048 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1049 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1051 Severity : enhancement
1053 Description: SNMP support enhancement
1054 Details : Adding total number of sampled request for an MDS node in snmp
1057 Severity : enhancement
1059 Description: Optimize ldlm waiting list processing for PR extent locks
1060 Details : When processing waiting list for read extent lock and meeting read
1061 lock that is same or wider to it that is not contended, skip
1062 processing rest of the list and immediatelly return current
1063 status of conflictness, since we are guaranteed there are no
1064 conflicting locks in the rest of the list.
1068 Description: Time out and refuse to reconnect
1069 Details : When the failover node is the primary node, it is possible
1070 to have two identical connections in imp_conn_list. We must
1071 compare not conn's pointers but NIDs, otherwise we can defeat
1072 connection throttling.
1076 Description: port llog fixes from b1_6 into HEAD
1077 Details : Port llog reference couting and some llog cleanups from b1_6
1078 (bug 10800) into HEAD, for protect from panic and access to already
1079 free llog structures.
1083 Description: Detect stride IO mode in read-ahead
1084 Details : When a client does stride read, read-ahead should detect that and
1085 read-ahead pages according to the detected stride pattern.
1089 Description: data checksumming impacts single node performance
1090 Details : add support for several checksum algorithm. Currently, only CRC32
1091 and Adler-32 are supported. The checksum type can be changed on
1092 the fly via /proc/fs/lustre/osc/*/checksum_type.
1096 Description: use adler32 for page checksums
1097 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1102 Description: build for x2 fails
1103 Details : fix compile issue on Cray systems.
1107 Description: Properly match for duplicate locks
1108 Details : Due to different lock order from skiplists code, we need to
1109 traverse entire list for now
1112 Frequency : only on PPC/SLES10
1114 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1115 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1116 be defined. Add a hack in mkfs_lustre.c to work around this compile
1121 Description: LASSERT on MDS when client holding flock lock dies
1122 Details : ldlm pool logic depends on number of granted locks equal to
1123 number of released locks which is not true for flock locks, so
1124 just exclude such locks from consideration.
1128 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1129 Details : Use fsfilt_sync() for both the whole filesystem sync and
1130 individual file sync to eliminate dangerous inode locking
1131 with I_LOCK that can lead to a deadlock.
1135 Description: performance in 1.6.3
1136 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1137 sure that 1MB requests are not fragmented by the block layer.
1139 Severity : enhancement
1141 Description: organize the server-side client stats on per-nid basis
1142 Details : Change the structure of stats under obdfilter and mds to
1150 The "uuid"s file would list the uuids of _active_ exports.
1151 And the clear entry is to clear all stats and stale nids.
1153 Severity : enhancement
1155 Description: eliminate client locks in face of contention
1156 Details : file contention detection and lockless i/o implementation
1157 for contended files.
1161 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1164 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1166 Description: MDS may lose file striping (and hence file data) in some cases
1167 Details : If there are additional extended attributes stored on the MDS,
1168 in particular ACLs, SELinux, or user attributes (if user_xattr
1169 is specified for the client mount options) then there is a risk
1170 of attribute loss. Additionally, the Lustre file striping
1171 needs to be larger than default (e.g. striped over all OSTs),
1172 and an additional attribute must be stored initially in the
1173 inode and then increase in size enough to be moved to the
1174 external attribute block (e.g. ACL growing in size) for file
1179 Description: skiplist implementation simplification
1180 Details : skiplists are used to group compatible locks on granted list
1181 that was implemented as tracking first and last lock of each lock group
1182 the patch changes that to using doubly linked lists
1186 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1187 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1188 because LASSERT on both the data supplied by a client, and the data
1189 on disk is dangerous and incorrect.
1191 Severity : enhancement
1193 Description: Slow truncate/writes to huge files at high offsets.
1194 Details : Directly associate cached pages to lock that protect those pages,
1195 this allows us to quickly find what pages to write and remove
1196 once lock callback is received.
1200 Description: more ldlm soft lockups
1201 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1202 starve other threads from the resource lock for a long time in
1203 case of long waiting queue, so change the debug level from
1204 D_OTHER to the less frequently used D_INFO.
1206 Severity : enhancement
1208 Description: add -gid, -group, -uid, -user options to lfs find
1212 Description: Hung threads in invalidate_inode_pages2_range
1213 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1214 one is completed. As a result, some RPCs are stuck in the queue
1219 Description: filter threads hungs on waiting journal commit
1220 Details : Cleanup filter group llog code, then only filter group llog will
1221 be only created in the MDS/OST syncing process.
1225 Description: Procfs and llog threads access destoryed import sometimes.
1226 Details : Sync the import destoryed process with procfs and llog threads by
1227 the import refcount and semaphore.
1229 Severity : enhancement
1231 Description: openlock cache of b1_6 port to HEAD
1236 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1237 Details : If insertion of an extent fails, then discard the inode
1238 preallocation and free data blocks else it can lead to duplicate
1243 Description: don't always update ctime in ext3_xattr_set_handle()
1244 Details : Current xattr code updates the inode ctime in ext3_xattr_set_handle.
1245 In some cases the ctime should not be updated, for example for
1246 2.0->1.8 compatibility it is necessary to delete an xattr and it
1247 should not update the ctime.
1251 Bugzilla : 15713/16362
1252 Description: Assertion in iopen_connect_dentry in 1.6.3
1253 Details : looking up an inode via iopen with the wrong generation number can
1254 populate the dcache with a disconneced dentry while the inode
1255 number is in the process of being reallocated. This causes an
1256 assertion failure in iopen since the inode's dentry list contains
1257 both a connected and disconnected dentry.
1261 Description: assertion failure in ldlm_handle2lock()
1262 Details : fix a race between class_handle_unhash() and class_handle2object()
1263 introduced in lustre 1.6.5 by bug 13622.
1268 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1269 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1274 Description: oops in sd_iostats_seq_show()
1275 Details : unloading/reloading the scsi low level driver triggers a kernel
1276 bug when trying to access the sd iostat file.
1281 Description: Kernel panics during QLogic driver reload
1282 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1283 patch, causing memory corruption.
1288 Description: journal_dev option does not work in b1_6
1289 Details : pass mount option during pre-mount.
1291 --------------------------------------------------------------------------------
1293 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1295 * Support for kernels:
1296 2.6.5-7.283 (SLES 9),
1297 2.6.9-55.EL (RHEL 4),
1298 2.6.16.46-0.14 (SLES 10),
1299 2.6.18.8 vanilla (kernel.org)
1300 * Client support for unpatched kernels:
1301 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1302 2.6.16 - 2.6.22 vanilla (kernel.org)
1303 * Due to recently discovered recovery problems, we do not recommend
1304 using patchless RHEL 4 clients with this or any earlier release.
1305 * Recommended e2fsprogs version: 1.39.cfs8
1306 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1307 * Starting with this release, the ldiskfs backing filesystem required
1308 by Lustre is now in its own package, lustre-ldiskfs. This package
1309 should be installed. It is versioned separately from Lustre and
1310 may be released separately in future.
1314 Description: block reactivating mgc import until all deactivates complete
1315 Details : Fix race when failing back MDT/MGS to itself (testing)
1318 Frequency : at statup only
1320 Description: mds_lov_synchronize race leads to various problems
1321 Details : simultaneous MDT->OST connections at startup can cause the
1322 sync to abort, leaving the OSC in a bad state.
1324 Severity : enhancement
1326 Description: add optional extra BUILD_VERSION info
1327 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1328 to override the lustre version.
1331 Frequency : 2.6.18 servers only
1333 Description: ll_kern_mount() doesn't release the module reference
1334 Details : The ldiskfs module reference count never drops down to 0
1335 because ll_kern_mount() doesn't release the module reference.
1340 Description: server LBUG when using old ost_num_threads parameter
1341 Details : Accept the old ost_num_threads parameter but warn that it
1342 is deprecated, and fix an off-by-one error that caused an LBUG.
1347 Description: Transient SCSI error results in persistent IO issue
1348 Details : iobuf->dr_error is not reinitialized to 0 between two
1352 Frequency : sometimes when underlying device returns I/O errors
1354 Description: OSTs not going read-only during write failures
1355 Details : OSTs are not remounted read-only when the journal commit threads
1356 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1357 instead of the ext3 wrappers.
1361 Description: poor connect scaling with increasing client count
1362 Details : Don't run filter_grant_sanity_check for more than 100 exports
1363 to improve scaling for large numbers of clients.
1366 Frequency : SLES10 only
1368 Description: sanity-quota.sh quotacheck failed: rc = -22
1369 Details : Quotas cannot be enabled on SLES10.
1372 Frequency : liblustre clients only
1374 Description: getdirentries does not give error when run on compute nodes
1375 Details : getdirentries does not fail when the size specified as an argument
1376 is too small to contain at least one entry
1378 Severity : enhancement
1380 Description: Add LNET router traceability for debug purposes
1381 Details : If a checksum failure occurs with a router as part of the
1382 IO path, the NID of the last router that forwarded the bulk data
1383 is printed so it can be identified.
1388 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1389 Details : Due to a race condition, liblustre clients were occasionally
1390 evicted incorrectly.
1392 Severity : enhancement
1394 Description: lfs setstripe use optional parameters instead of postional
1397 Severity : enhancement
1399 Description: Nanosecond timestamp support for ldiskfs
1400 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1401 resolution timestamps. There is not yet support for this at
1402 the Lustre filesystem level.
1405 Frequency : during server recovery
1407 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1408 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1412 Frequency : nfs export on patchless client
1414 Description: connectathon hang when test nfs export over patchless client
1415 Details : Disconnected dentry cannot be found with lookup, so we do not need
1416 to unhash it or make it invalid
1419 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1420 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1421 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1426 Description: open req refcounting wrong on reconnect
1427 Details : If reconnect happened between getting open reply from server and
1428 call to mdc_set_replay_data in ll_file_open, we will schedule
1429 replay for unreferenced request that we are about to free.
1430 Subsequent close will crash in variety of ways.
1431 Check that request is still eligible for replay in
1432 mdc_set_replay_data().
1437 Description: disable writes to filesystem when reading health_check file
1438 Details : the default for reading the health_check proc file has changed
1439 to NOT do a journal transaction and write to disk, because this
1440 can cause reads of the /proc file to hang and block HA state
1441 checking on a healthy but otherwise heavily loaded system. It
1442 is possible to return to the previous behaviour during configure
1443 with --enable-health-write.
1445 Severity : enhancement
1447 Description: 64-bit inode version
1448 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1449 to the inode. This will be required for version-based recovery.
1454 Description: MDS fails to start if a duplicate client export is detected
1455 Details : in some rare cases it was possible for a client to connect to
1456 an MDS multiple times. Upon recovery the MDS would detect this
1457 and fail during startup. Handle this more gracefully.
1459 Severity : enhancement
1461 Description: Add -o localflock option to simulate old noflock
1463 Details : This will achieve local-only flock/fcntl locks
1469 Description: log_commit_thread vs filter_destroy race leads to crash
1470 Details : Take import reference before releasing llog record semaphore
1475 Description: Wrong request locking in request set processing
1476 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1477 imp_delayed_list, in this place should be used imp_lock.
1480 Frequency : when reconnection
1482 Description: Grant Leak when osc reconnect to OST
1483 Details : When osc reconnect ost, OST(filter) should check whether it
1484 should grant more space to client by comparing fed_grant and
1485 cl_avail_grant, and return the granted space to client instead
1486 of "new granted" space, because client will call osc_init_grant
1487 to update the client grant space info.
1490 Frequency : when client reconnect to OST
1492 Description: Grant Leak when osc do resend and replay bulk write
1493 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1494 bulk write request, because the grant info while be sync between
1495 OSC and OST when reconnect, and we should ignore the grant info
1496 these of resend/replay write req.
1501 Description: Grant space more than avaiable left space sometimes.
1502 Details : When then OST is about to be full, if two bulk writing from
1503 different clients came to OST. Accord the avaliable space of the
1504 OST, the first req should be permitted, and the second one
1505 should be denied by ENOSPC. But if the seconde arrived before
1506 the first one is commited. The OST might wrongly permit second
1507 writing, which will cause grant space > avaiable space.
1510 Frequency : when client is evicted
1512 Description: Grant might be wrongly erased when osc is evicted by OST
1513 Details : when the import is evicted by server, it will fork another
1514 thread ptlrpc_invalidate_import_thread to invalidate the
1515 import, where the grant will be set to 0. While the original
1516 thread will update the grant it got when connecting. So if
1517 the former happened latter, the grant will be wrongly errased
1518 because of this race.
1523 Description: Checking Stale with correct fid
1524 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1525 to check whether it is stale, because sometimes, we want the
1526 enqueue happened anyway, and op_data.fid2 will not be initialized.
1528 Severity : enhancement
1530 Description: update patchless client
1531 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1534 Frequency : only with 2.4 kernel
1536 Description: random memory corruption
1537 Details : size of struct ll_inode_info is to big for union inode.u and this
1538 can be cause of random memory corruption.
1543 Description: Memory leak in recovery
1544 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1545 It should also check obd_fail before fsfilt_start, otherwise if
1546 fsfilt_start return -EROFS,(failover mds during mds recovery).
1547 then the req will return with repmsg->transno = 0 and rc = EROFS.
1548 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1549 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1550 be freed no matter whether fsfilt_commit success or not.
1553 Frequency : only with huge count clients
1555 Description: Prevents from taking the superblock lock in llap_from_page for
1557 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1558 for llap_from_page call in ll_removepage prevents from taking the
1559 superblock lock for a soon died page.
1564 Description: Not check open intent error before release open handle
1565 Details : in some rare cases, the open intent error is not checked before
1566 release open handle, which may cause
1567 ASSERTION(open_req->rq_transno != 0), because it tries to release
1568 the failed open handle.
1573 Description: Set cat log bitmap only after create log success.
1574 Details : in some rare cases, the cat log bitmap is set too early. and it
1575 should be set only after create log success.
1579 Description: Accessing a block bevice can re-enable I/O when Lustre is
1580 tearing down a device.
1581 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1586 Description: service threads may hog cpus when there are a lot of requests
1587 Details : Insert cond_resched to give other threads a chance to use some CPU
1592 Description: the cat log was not initialized in recovery
1593 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1594 unlink log on mds will not be initialized until mds post
1595 recovery. And also in mds post recovery, the unlink log will
1596 initialization will be done asynchronausly, so there will be race
1597 between add unlink log and unlink log initialization.
1601 Description: brw_stats were being printed incorrectly
1602 Details : brw_stats were being printed as log2 but all of them were not
1603 recorded as log2. Also remove some code duplication arising from
1604 filter_tally_{read,write}.
1608 Frequency : rare, only in recovery.
1609 Description: ASSERTION(req->rq_type != LI_POISON) failed
1610 Details : imp_lock should be held while iterating over imp_sending_list for
1611 prevent destroy request after get timeout in ptlrpc_queue_wait.
1615 Description: replay-single.sh test 52 fails
1616 Details : A lock's skiplist need to be cleanup when it being unlinked
1617 from its resource list.
1621 Description: Short directio read returns full requested size rather than
1623 Details : Direct I/O operations should return actual amount of bytes
1624 transferred rather than requested size.
1626 Severity : enhancement
1628 Description: metadata RPC reduction (e.g. for rm performance)
1629 Details : decrease the amount of synchronous RPC between clients and servers
1630 by canceling conflicing lock before the operation on the client
1631 and packing thier handles into the main operation RPC to server.
1633 Severity : enhancement
1635 Description: Async OSC create to avoid the blocking unnecessarily.
1636 Details : If a OST has no remain object, system will block on the creating
1637 when need to create a new object on this OST. Now, ways use
1638 pre-created objects when available, instead of blocking on an
1639 empty osc while others are not empty. If we must block, we block
1640 for the shortest possible period of time.
1642 Severity : enhancement
1644 Description: refine locking for avoid write wrong info into lov_objid file
1645 Details : fix possible races with add new target and write/update data in
1648 --------------------------------------------------------------------------------
1650 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1655 Frequency : on some architectures
1657 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1658 Details : Uninitialized flags sometimes cause configuration commands to
1662 Frequency : patchless clients only
1664 Description: missing __iget() symbol export
1665 Details : The __iget() symbol export is missing. To avoid the need for
1666 this on patchless clients the deathrow inode reaper is turned
1667 off, and we depend on the VM to clean up old inodes. This
1668 dependency was during via the fix for bug 12181.
1670 --------------------------------------------------------------------------------
1672 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1674 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1675 INTEROPERATE with older versions automatically. In many cases a
1676 special upgrade step is needed. Please read the
1677 user documentation before upgrading any part of a live system.
1678 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1679 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1680 * WARNING: Lustre configuration and startup changes are required with
1681 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1687 Severity : enhancement
1689 Description: MountConf
1690 Details : Lustre configuration is now managed via mkfs and mount
1691 commands instead of lmc and lconf. New obd types (MGS, MGC)
1692 are added for dynamic configuration management. See
1693 https://mail.clusterfs.com/wikis/lustre/MountConf for
1696 Severity : enhancement
1698 Description: dynamic OST addition
1699 Details : OSTs can now be added to a live filesystem
1701 Severity : enhancement
1703 Description: startup order invariance
1704 Details : MDTs and OSTs can be started in any order. Clients only
1705 require the MDT to complete startup.
1707 Severity : enhancement
1709 Description: parallel, asynchronous orphan cleanup
1710 Details : orphan cleanup is now performed in separate threads for each
1711 OST, allowing parallel non-blocking operation.
1713 Severity : enhancement
1715 Description: optimized stripe assignment
1716 Details : stripe assignments are now made based on ost space available,
1717 ost previous usage, and OSS previous usage, in order to try
1718 to optimize storage space and networking resources.
1720 Severity : enhancement
1722 Description: Permanently set tunables
1723 Details : All writable /proc/fs/lustre tunables can now be permanently
1724 set on a per-server basis, at mkfs time or on a live system.
1726 Severity : enhancement
1728 Description: Lustre message v2
1729 Details : Add lustre message format v2.
1731 Severity : enhancement
1733 Description: client OST exclusion list
1734 Details : Clients can be started with a list of OSTs that should be
1735 declared "inactive" for known non-responsive OSTs.
1739 Description: SPEC SFS validation failure on NFS v2 over lustre.
1740 Details : Changes the blocksize for regular files to be 2x RPC size,
1741 and not depend on stripe size.
1743 Severity : enhancement
1745 Description: Multiple MD RPCs in flight.
1746 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1747 To avoid overly-overloading MDS, introduce a limit on number of
1748 MDS RPCs in flight for a single client and add /proc controls
1749 to adjust this limit.
1751 Severity : enhancement
1753 Description: client read/write statistics
1754 Details : Add client read/write call usage stats for performance
1755 analysis of user processes.
1756 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1757 file access. extents_stats shows chunk size distribution.
1758 extents_stats_per_process show chunk size distribution per
1761 Severity : enhancement
1763 Description: mds statistics
1764 Details : Add detailed mds operations statistics in
1765 /proc/fs/lustre/mds/*/stats.
1769 Description: Failure of copying files with lustre special EAs.
1770 Details : Client side always return success for setxattr call for lustre
1771 special xattr (currently only "trusted.lov").
1776 Description: Refcount LNET uuids
1777 Details : The global LNET uuid list grew linearly with every startup;
1778 refcount repeated list entries instead of always adding to
1781 Severity : enhancement
1783 Description: Dynamic service threads
1784 Details : Within a small range, start extra service threads
1785 automatically when the request queue builds up.
1788 Frequency : mixed-endian client/server environments
1790 Description: mixed-endian crashes
1791 Details : The new msg_v2 system had some failures in mixed-endian
1794 Severity : enhancement
1796 Description: Easy OST removal
1797 Details : OSTs can be permanently deactivated with e.g. 'lctl
1798 conf_param lustre-OST0001.osc.active=0'
1800 Severity : enhancement
1802 Description: MGS proc entries
1803 Details : Added basic proc entries for the MGS showing what filesystems
1806 Severity : enhancement
1808 Description: provide MGS failover
1809 Details : Added config lock reacquisition after MGS server failover.
1811 Severity : enhancement
1813 Description: add Linux 2.4 support
1814 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1818 Description: a large application tries to do I/O to the same resource and dies
1819 in the middle of it.
1820 Details : Check the req->rq_arrival time after the call to
1821 ost_brw_lock_get(), but before we do anything about
1822 processing it & sending the BULK transfer request. This
1823 should help move old stale pending locks off the queue as
1824 quickly as obd_timeout.
1827 Frequency : when an incorrect nid is specified during startup
1829 Description: ptlrpc connect to non-existant node causes kernel crash
1830 Details : LNET can't be re-entered from an event callback, which
1831 happened when we expire a message after the export has been
1832 cleaned up. Instead, hand the zombie cleanup off to another
1835 Severity : enhancement
1837 Description: plain/inodebits lock performance improvement
1838 Details : Grouping plain/inodebits in granted list by their request modes
1839 and bits policy, thus improving the performance of search through
1843 Frequency : only if OST filesystem is corrupted
1845 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1846 Details : for a short time RPCs with bulk IO are in the replay list,
1847 but replay of bulk IOs is unimplemented. If the OST filesystem
1848 is corrupted due to disk cache incoherency and then replay is
1849 started it is possible to trip an assertion. Avoid putting
1850 committed RPCs into the replay list at all to avoid this issue.
1853 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1855 Description: System hang on startup
1856 Details : This bug allowed the liblustre (e.g. catamount) client to
1857 return to the app before handling all startup RPCs. This
1858 could leave the node unresponsive to lustre network traffic
1859 and manifested as a server ptllnd timeout.
1861 Severity : enhancement
1863 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1864 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1865 boolean to control whether to print peer debug info when a
1866 client's RPC times out.
1869 Frequency : only for kernels with patches from Lustre below 1.4.3
1871 Description: Remove old rdonly API
1872 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1875 Frequency : only for devices with external journals
1877 Description: Set external device read-only also
1878 Details : During a commanded failover stop, we set the disk device
1879 read-only while the server shuts down. We now also set any
1880 external journal device read-only at the same time.
1883 Frequency : when upgrading from 1.4 while trying to change parameters
1885 Description: The wrong (new) MDC name was used when setting parameters for
1886 upgraded MDT's. Also allows changing of OSC (and MDC)
1887 parameters if --writeconf is specified at tunefs upgrade time.
1890 Frequency : when setting specific ost indicies
1892 Description: QOS code breaks on skipped indicies
1893 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1894 created with --index need not be sequential.
1896 Severity : enhancement
1898 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1899 Details : The uninit_groups feature works in conjunction with the kernel
1900 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1901 up the pass1 processing of e2fsck. This is a read-only feature
1902 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1903 mount filesystems that have had this feature enabled.
1905 Severity : enhancement
1907 Description: Improve multi-block allocation algorithm to avoid fragmentation
1908 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1909 allocation locality and avoid filesystem fragmentation.
1911 ------------------------------------------------------------------------------
1913 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1915 * Support for kernels:
1916 2.4.21-47.0.1.EL (RHEL 3)
1917 2.6.5-7.283 (SLES 9)
1918 2.6.9-42.0.10.EL (RHEL 4)
1919 2.6.12.6 vanilla (kernel.org)
1920 2.6.16.27-0.9 (SLES 10)
1921 * Recommended e2fsprogs version: 1.39.cfs5
1923 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1927 Frequency : occasional, depends on client load and configuration
1928 Bugzilla : 12181, 12203
1929 Description: data loss for recently-modified files
1931 Details : In some cases it is possible that recently written or created
1932 files may not be written to disk in a timely manner (this should
1933 normally be within 30s unless client IO load is very high).
1934 The problem appears as zero-length files or files that are a
1935 multiple of 1MB in size after a client crash or client eviction
1936 that are missing data at the end of the file.
1938 This problem is more likely to be hit on clients where files are
1939 repeatedly created and unlinked in the same directory, clients
1940 have a large amount of RAM, have many CPUs, the filesystem has
1941 many OSTs, the clients are rebooted frequently, and/or the files
1942 are not accessed by other nodes after being written.
1944 The presence of the problem can be detected by looking at
1945 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1946 smaller than the second (nr_unused) then dirty files will not
1947 be flushed automatically to disk. "sync; sleep 10" should be
1948 run several times on the node before unmounting it to update
1949 Lustre (this is also safe to run on nodes without this problem).
1951 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1952 that can cause this same problem, so customers using that kernel
1953 also need to update the kernel in addition to Lustre. In order
1954 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1956 It is normal that files written just before a client crash (less
1957 than 30s) may not yet have been flushed to disk, even for local
1961 Frequency : frequent on thin XT3 nodes
1963 Description: UUID collision on thin XT3 Linux nodes
1964 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1965 randomly, since we relied on an insufficiently-seeded PRNG.
1970 Description: OSS hangs after "All ost request buffers busy"
1971 Details : A deadlock between quota and journal operations caused OSS
1972 hangs after printing "All ost request buffers busy."
1975 Frequency : always on liblustre builds
1977 Description: Cleanup compiler warnings on liblustre
1980 Frequency : always on liblustre builds on XT3
1982 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1986 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1987 > 32000 subdirectories
1988 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1989 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1990 subdirectory count crosses 32000. This will aid e2fsck to
1991 correctly handle more than 32000 subdirectories.
1994 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1996 Description: System hang on startup
1997 Details : This bug allowed the liblustre (e.g. catamount) client to
1998 return to the app before handling all startup RPCs. This
1999 could leave the node unresponsive to lustre network traffic
2000 and manifested as a server ptllnd timeout.
2002 Severity : enhancement
2004 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2005 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2006 boolean to control whether to print peer debug info when a
2007 client's RPC times out.
2012 Description: make O_SYNC working on 2.6 kernels
2013 Details : 2.6 kernels use different method for mark pages for write,
2014 so need add a code to lustre for O_SYNC work.
2019 Description: Failure to close file and release space on NFS
2020 Details : Put inode details into lock acquired in ll_intent_file_open.
2021 Use mdc_intent_lock in ll_intent_open to properly
2022 detect all kind of errors unhandled by mdc_enqueue.
2027 Description: proc file read during shutdown sometimes raced obd removal,
2029 Details : Add lock to prevent obd access after proc file removal.
2032 Frequency : Only for files larger than 4GB on 32-bit clients.
2034 Description: improperly doing page alignment of locks
2035 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2036 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2041 Description: RPCs being resent when they shouldn't be
2042 Details : Some RPCs that should not be resent are being resent. This
2043 can cause inconsistencies in the RPC state machine. Do not
2044 resend such requests.
2047 Frequency : rare, only with NFS export
2049 Description: Crash on NFS re-export node
2050 Details : under very unusual load conditions an assertion is hit in
2051 ll_intent_file_open()
2054 Frequency : only if OST filesystem is corrupted
2056 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2057 Details : for a short time RPCs with bulk IO are in the replay list,
2058 but replay of bulk IOs is unimplemented. If the OST filesystem
2059 is corrupted due to disk cache incoherency and then replay is
2060 started it is possible to trip an assertion. Avoid putting
2061 committed RPCs into the replay list at all to avoid this issue.
2066 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2067 Details : Large single O_DIRECT read and write calls can fail to allocate
2068 a sufficiently large buffer to process the request. In case of
2069 allocation failure the allocation is retried with a smaller
2070 buffer and broken into smaller requests.
2072 Severity : enhancement
2074 Description: Add -o localflock option to simulate old noflock behaviour.
2075 Details : This will achieve local-only flock/fcntl locks coherentness.
2080 Description: versioning check is incomplete
2081 Details : Checking the version difference of client vs. server, report
2082 error if the gap is too big.
2087 Description: add support PG_writeback bit
2088 Details : add support for PG_writeback bit for Lustre, for more carefull
2089 work with page cache in 2.6 kernel. This also fix some deadlocks
2090 and remove hack for work O_SYNC with 2.6 kernel.
2092 Severity : enhancement
2094 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2095 Details : The uninit_groups feature works in conjunction with the kernel
2096 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2097 up the pass1 processing of e2fsck. This is a read-only feature
2098 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2099 mount filesystems that have had this feature enabled.
2101 Severity : enhancement
2103 Description: Improve multi-block allocation algorithm to avoid fragmentation
2104 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2105 allocation locality and avoid filesystem fragmentation.
2107 ------------------------------------------------------------------------------
2109 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
2111 * Support for kernels:
2112 2.6.9-42.0.3.EL (RHEL 4)
2113 2.6.5-7.276 (SLES 9)
2114 2.4.21-47.0.1.EL (RHEL 3)
2115 2.6.12.6 vanilla (kernel.org)
2116 2.6.16.21-0.8 (SLES10)
2117 * Recommended e2fsprogs version: 1.39.cfs2-0
2119 * The backwards-compatible /proc/sys/portals symlink has been removed
2120 in this release. Before upgrading, please ensure that you change
2121 any configuration scripts or /etc/sysctl.conf files that access
2122 /proc/sys/portals/* or sysctl portals.* to use the corresponding
2123 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
2124 in advance of the upgrade on any system running Lustre 1.4.6 or
2125 newer, since /proc/sys/lnet was added in that version.
2126 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2130 Frequency : only when quota is used
2132 Description: avoid scanning export list for quota master
2133 Details : Change the algorithms to avoid scanning export list in order
2134 to improve the efficiency.
2137 Frequency : MDS failover only, very rarely
2139 Description: "went back in time" messages on mds failover
2140 Details : The greatest transno may be lost when the current operation
2141 finishes with an error (transno==0) and the client's last_rcvd
2142 record is over-written. Save the greatest transno in the
2143 mds_last_transno for this case.
2146 Frequency : always for specific kernels and striping counts
2148 Description: client may get "Matching packet too big" without ACL support
2149 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2150 when trying to access files in certain configurations. The
2151 clients should in fact be denied when mounting because they do
2152 not understand ACLs.
2155 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2157 Description: many clients connecting with IO in progress causes connect timeouts
2158 Details : Avoid synchronous journal commits to avoid delays caused by many
2159 clients connecting/disconnecting when bulk IO is in progress.
2160 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2161 OST_IO_PORTAL to avoid delays behind potentially many pending
2165 Frequency : occasionally with multiple writers to a single file
2167 Description: shared writes to file may result in wrong size reported by stat()
2168 Details : Allow growing of kms when extent lock is cancelled
2171 Frequency : always with random mmap IO to multi-striped file
2173 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2174 Details : Only if the hole is at the end of OST object so that kms is too
2175 small. Fix is to increase kms accordingly in ll_nopage.
2178 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2180 Description: writes to a missing object would leak memory on the OST
2181 Details : If there is an inconsistency between the MDS and OST filesystems,
2182 such that the MDS references an object that doesn't exist, writes
2183 to that object will leak memory due to incorrect cleanup in the
2184 error handling path, eventually running out of memory on the OST.
2189 Description: Creating too long symlink causes lustre errors
2190 Details : Check symlink and name lengths before sending requests to MDS.
2193 Frequency : only if flock is enabled (not on by default)
2195 Description: posix locks not released on fd closure on 2.6.9+
2196 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2197 kernels, this caused such locks not to be released on fd close and
2198 then assertions on fs unmount about still used locks.
2201 Frequency : MDS failover only, very rarely
2203 Description: clients may get ASSERTION(granted_lock != NULL)
2204 Details : When request was taking a long time, and a client was resending
2205 a getattr by name lock request. The were multiple lock requests
2206 with the same client lock handle and
2207 mds_getattr_name->fixup_handle_for_resent_request found one of the
2208 lock handles but later failed with ASSERTION(granted_lock != NULL).
2213 Description: handle->h_buffer_credits > 0, assertion failure
2214 Details : h_buffer_credits is zero after truncate, causing assertion
2215 failure. This patch extends the transaction or creates a new
2219 Frequency : NFS re-export or patchless client
2220 Bugzilla : 11179, 10796
2221 Description: Crash on NFS re-export node (__d_move)
2222 Details : We do not want to hash the dentry if we don't have a lock.
2223 But if this dentry is later used in d_move, we'd hit uninitialised
2224 list head d_hash, so we just do this to init d_hash field but
2225 leave dentry unhashed.
2228 Frequency : NFS re-export or patchless client
2230 Description: NFS exports has problem with symbolic link
2231 Details : lustre client didn't properly install dentry when re-exported
2232 to NFS or running patchless client.
2235 Frequency : NFS re-export or patchless client
2237 Description: Various nfs/patchless fixes.
2238 Details : fixes reuse disconected alias for lookup process - this fixes
2239 warning "find_exported_dentry: npd != pd",
2240 fix permission error with open files at nfs.
2241 fix apply umask when do revalidate.
2244 Frequency : occasional
2246 Description: Crash on NFS re-export node
2247 Details : calling clear_page() on the wrong pointer triggered oops in
2248 generic_mapping_read().
2251 Frequency : rarely, using O_DIRECT IO
2253 Description: unaligned directio crashes client with LASSERT
2254 Details : check for unaligned buffers before trying any requests.
2257 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2259 Description: stale data returned from RAID cache
2260 Details : If only a small amount of IO is done to the RAID device before
2261 reading it again it is possible to get stale data from the RAID
2262 cache instead of reading it from disk.
2265 Frequency : always for sles10 kernel
2267 Description: sles10 support
2268 Details : ll_follow_link: compile fixes and using of nd_set_link
2269 under newer kernels.
2272 Frequency : depends on arch, kernel and compiler version, always on sles10
2275 Description: recursive or deep enough symlinks cause stack overflow
2276 Details : getting rid of large stack-allocated variable in
2280 Frequency : depends on hardware
2282 Description: lustre write performance loss in the SLES10 kernel
2283 Details : the performance loss is caused by using of write barriers in the
2284 ext3 code. The SLES10 kernel turns barrier support on by
2285 default. The fix is to undo that change for ldiskfs.
2287 ------------------------------------------------------------------------------
2289 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2291 * Support for kernels:
2292 2.6.9-42.0.3EL (RHEL 4)
2293 2.6.5-7.276 (SLES 9)
2294 2.4.21-47.0.1.EL (RHEL 3)
2295 2.6.12.6 vanilla (kernel.org)
2299 Frequency : quota enabled and large files being deleted
2301 Description: releasing more than 4GB of quota at once hangs OST
2302 Details : If a user deletes more than 4GB of files on a single OST it
2303 will cause the OST to spin in an infinite loop. Release
2304 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2309 Description: statfs data retrieved from /proc may be stale or zero
2310 Details : When reading per-device statfs data from /proc, in the
2311 {kbytes,files}_{total,free,avail} files, it may appear
2312 as zero or be out of date.
2315 Frequency : systems with MD RAID1 external journal devices
2317 Description: lconf's call to blkid is confused by RAID1 journal devices
2318 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2319 of returning all block devices that match the journal UUID.
2322 Frequency : always, for aggregate stripe size over 4GB
2324 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2325 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2326 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2329 Frequency : always if "lfs find" used on a local file/directory
2331 Description: "lfs find" segfaults if used on a local file/directory
2332 Details : The case where a directory component was not specified wasn't
2333 handled correctly. Handle this properly.
2336 Frequency : always on ppc64
2338 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2339 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2340 perform endianness conversion.
2343 Frequency : rarely (truncate to non-zero file size after write under load)
2344 Bugzilla : 10730, 10687
2345 Description: Files padded with zeros to next 4K multiple
2346 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2347 it is possible that files that are truncated to a non-zero size
2348 immediately after being written are filled with zero bytes beyond
2349 the truncated size. No file data is lost.
2351 Severity : enhancement
2353 Description: Allow recovery/failover for liblustre clients.
2354 Details : liblustre clients were unaware of failover configurations until
2357 Severity : enhancement
2359 Description: user file locks should fail when not mounting with flock option
2360 Details : Set up an error-returning stub in ll_file_operations.lock field
2361 to prevent incorrect behaviour when client is mounted without
2362 flock option. Also, set up properly f_op->flock field for
2366 Frequency : always on ia64
2368 Description: "lfs df" loops on printing out MDS statfs information
2369 Details : The obd_ioctl_data was not initialized and in some systems
2370 this caused a failure during the ioctl that did not return
2371 an error. Initialize the struct and return an error on failure.
2374 Frequency : SLES 9 only
2376 Description: Error of copying files with lustre special EAs as root
2377 Details : Client side always return success for setxattr call for lustre
2378 special xattr (currently only "trusted.lov").
2381 Frequency : rarely on clusters with both ia64+i386 clients
2383 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2384 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2385 clients are doing concurrent writes to the same file it is
2386 possible that the ia64 clients may LASSERT because the OST
2387 extent locks are not PAGE_SIZE aligned. Ensure that grown
2388 locks are always aligned on the request boundary.
2391 Frequency : specific use, occasional
2393 Description: Overwriting in use executable truncates on-disk binary image
2394 Details : If one node attempts to overwrite an executable in use by
2395 another node, we now correctly return ETXTBSY instead of
2396 truncating the file.
2398 Severity : enhancement
2400 Description: Async OSC create to avoid the blocking unnecessarily.
2401 Details : If a OST has no remain object, system will block on the creating
2402 when need to create a new object on this OST. Now, ways use
2403 pre-created objects when available, instead of blocking on an
2404 empty osc while others are not empty. If we must block, we block
2405 for the shortest possible period of time.
2410 Description: chmod on Lustre root is propagated to other clients
2411 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2412 invalid by the follow_mount time.
2415 Frequency : liblustre clients only
2417 Description: Race in 'instant cancel' lock handling could lead to such locks
2418 never to be granted in case of SMP MDS
2419 Details : Do not destroy not yet granted but cbpending locks in
2423 Frequency : replay/resend of open
2425 Description: non null lock assetion failure in mds_intent_policy
2426 Details : Trying to replay/resend lockless open requests resulted in
2427 mds_open() returning 0 with no lock. Now it sets a flag if
2428 a lock is going to be returned.
2430 Severity : enhancement
2432 Description: Checksum enhancements
2433 Details : New checksum enhancements allow for resending RPCs that failed
2436 Severity : enhancement
2438 Description: Tunables on number of dirty pages in cacche
2439 Details : Allow to set limit on number of dirty pages cached.
2444 Description: client crash on unmount - lock still has references
2445 Details : In some error handling cases it was possible to leak a lock
2446 reference on a client while accessing a file. This was not
2447 harmful to the client during operation, but would cause the
2448 client to crash when the filesystem is unmounted.
2451 Frequency : specific case, rare
2453 Description: ETXTBSY on mds though file not in use by client
2454 Details : ETXTBSY is no longer incorrectly returned when attempting to
2455 chmod or chown a directory that the user previously tried to
2456 execute or a currently-executing binary.
2459 Frequency : extremely rare except on liblustre-based clients
2461 Description: Lustre space not freed when files are deleted
2462 Details : Clean up open-unlinked files after client eviction. Previously
2463 the unlink was skipped and the files remained as orphans.
2468 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2469 Details : In some cases it was possible to send a blocking callback to
2470 a client doing a glimpse, even though that client didn't get
2471 a lock granted. When the glimpse lock is cancelled on the OST
2472 the freed lock is left on the waiting list and corrupted the list.
2475 Frequency : all core dumps
2477 Description: Broke core dumps to lustre
2478 Details : Negative dentry may be unhashed if parent does not have UPDATE
2479 lock, but some callers, e.g. do_coredump, expect dentry to be
2480 hashed after successful create, hash it in ll_create_it.
2482 ------------------------------------------------------------------------------
2484 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2486 * Support for kernels:
2487 2.6.9-42.0.2.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)
2494 Frequency : always on RHEL 3
2496 Description: Number of open files grows over time
2497 Details : The number of open files grows over time, whether or not
2498 Lustre is started. This was due to a filp leak introduced
2499 by one of our kernel patches.
2501 ------------------------------------------------------------------------------
2503 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2505 * Support for kernels:
2506 2.6.9-42.EL (RHEL 4)
2507 2.6.5-7.276 (SLES 9)
2508 2.4.21-40.EL (RHEL 3)
2509 2.6.12.6 vanilla (kernel.org)
2514 Bugzilla : 5719, 9635, 9792, 9684
2515 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2516 Details : If a server is under heavy load and cannot reply to new
2517 connection requests before the client resends the (re)connect,
2518 the connection handling code can behave badly if two service
2519 threads are concurrently handing separate (re)connections from
2520 the same client. Add better locking to the connection handling
2521 code, and ensure that only a single connection will be processed
2522 for a given client UUID, even if the lock is dropped.
2524 Severity : enhancement
2526 Description: add TCP zero-copy support to kernel
2527 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2528 sends if the hardware supports scatter-gather and checksumming.
2529 This allows socklnd to do client-write and server-read more
2530 efficiently and reduce CPU utilization from skbuf copying.
2533 Frequency : only if NFS exporting from client
2535 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2536 Details : If setting attributes on a file created under NFS that had
2537 never been opened it would be possible to oops the client
2538 if the file had no objects.
2541 Frequency : always for liblustre
2543 Description: liblustre client does MDS+OSTs setattr RPC for each write
2544 Details : When doing a write from a liblustre client, the client
2545 incorrectly issued an RPC to the MDS and each OST the file was
2546 striped over in order to update the timestamps. When writing
2547 with small chunks and many clients this could overwhelm the MDS
2548 with RPCs. In all cases it would slow down the write because
2549 these RPCs are unnecessary.
2551 Severity : enhancement
2553 Description: allow number of MDS service threads to be changed at module load
2554 Details : It is now possible to change the number of MDS service threads
2555 running. Adding "options mds mds_num_threads={N}" to the MDS's
2556 /etc/modprobe.conf will set the number of threads for the next
2557 time Lustre is restarted (assuming the "mds" module is also
2558 reloaded at that time). The default number of threads will
2559 stay the same, 32 for most systems.
2564 Description: OST crash if filesystem is unformatted or corrupt
2565 Details : If an OST is started on a device that has never been formatted
2566 or if the filesystem is corrupt and cannot even mount then the
2567 error handling cleanup routines would dereference a NULL pointer.
2572 Description: NULL pointer deref in llap_from_page.
2573 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2574 incorrect) mapping if the page was truncated/reclaimed while it was
2575 searched for. Check for this condition and skip such pages when
2576 doing readahead. Introduce extra check to llap_from_page() to
2577 verify page->mapping->host is non-NULL (so page is not anonymous).
2580 Frequency : Sometimes when using sys_sendfile
2582 Description: "page not covered by a lock" warnings from ll_readpage
2583 Details : sendfile called ll_readpage without right page locks present.
2584 Now we introduced ll_file_sendfile that does necessary locking
2585 around call to generic_file_sendfile() much like we do in
2589 Frequency : with certain MDS communication failures at client mount time
2591 Description: NULL pointer deref after failed client mount
2592 Details : a client connection request may delayed by the network layer
2593 and not be sent until after the PTLRPC layer has timed out the
2594 request. If the client fails the mount immediately it will try
2595 to clean up before the network times out the request. Add a
2596 reference from the request import to the obd device and delay
2597 the cleanup until the network drops the request.
2600 Frequency : occasionally during client (re)connect
2602 Description: assertion failure during client (re)connect
2603 Details : processing a client connection request may be delayed by the
2604 client or server longer than the client connect timeout. This
2605 causes the client to resend the connection request. If the
2606 original connection request is replied in this interval, the
2607 client may trip an assertion failure in ptlrpc_connect_interpret()
2608 which thought it would be the only running connect process.
2611 Frequency : only with obd_echo servers and clients that are rebooted
2613 Description: kernel BUG accessing uninitialized data structure
2614 Details : When running an obd_echo server it did not start the ping_evictor
2615 thread, and when a client was evicted an uninitialized data
2616 structure was accessed. Start the ping_evictor in the RPC
2617 service startup instead of the OBD startup.
2619 Severity : enhancement
2620 Bugzilla : 10193 (patchless)
2621 Description: Remove dependency on various unexported kernel interfaces.
2622 Details : No longer need reparent_to_init, exit_mm, exit_files,
2623 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2626 Frequency : rare (only users of deprecated and unsupported LDAP config)
2628 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2629 Details : LDAP apparently contains 'lustreName' attributes instead of
2630 'name'. A simple remapping of the name is sufficient.
2633 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2635 Description: waiting_locks_callback trips kernel BUG if client is evicted
2636 Details : Running with the dump_on_timeout debug flag turned on makes
2637 it possible that the waiting_locks_callback() can try to dump
2638 the Lustre kernel debug logs from an interrupt handler. Defer
2639 this log dumping to the expired_lock_main() thread.
2641 Severity : enhancement
2643 Description: Support NFS exporting on 2.6 kernels.
2644 Details : Implement non-rawops metadata methods for NFS server to use without
2645 changing NFS server code.
2648 Frequency : very rare (synthetic metadata workload only)
2650 Description: two racing renames might cause an MDS thread to deadlock
2651 Details : Running the "racer" program may cause one MDS thread to rename
2652 a file from being the source of a rename to being the target of
2653 a rename at exactly the same time that another thread is doing
2654 so, and the second thread has already enqueued these locks after
2655 doing a lookup of the target and is trying to relock them in
2656 order. Ensure that we don't try to re-lock the same resource.
2659 Frequency : only very large systems with liblustre clients
2661 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2662 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2663 all OSTs in parallel. This allows the eviction of stale liblustre
2664 clients to proceed much faster than if they were done in series,
2665 and also offers similar improvements for other set_info RPCs.
2670 Description: excessive CPU usage during initial read phase on client
2671 Details : During the initial read phase on a client, it would agressively
2672 retry readahead on the file, consuming too much CPU and impacting
2673 performance (since 1.4.5.8). Improve the readahead algorithm
2674 to avoid this, and also improve some other common cases (read
2675 of small files in particular, where "small" is files smaller than
2676 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2681 Description: MDS crash when receiving packet with unknown intent.
2682 Details : Do not LBUG in unknown intent case, just return -EFAULT
2684 Severity : enhancement
2685 Bugzilla : 9293, 9385
2686 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2687 Details : Do not serialize getattr (non-intent version) and statfs.
2690 Frequency : occasional, when OST network is overloaded/intermittent
2692 Description: client evicted by OST after bulk IO timeout
2693 Details : If a client sends a bulk IO request (read or write) the OST
2694 may evict the client if it is unresposive to its data GET/PUT
2695 request. This is incorrect if the network is overloaded (takes
2696 too long to transfer the RPC data) or dropped the OST GET/PUT
2697 request. There is no need to evict the client at all, since
2698 the pinger and/or lock callbacks will handle this, and the
2699 client can restart the bulk request.
2702 Frequency : Always when mmapping file with no objects
2704 Description: client crashes when mmapping file with no objects
2705 Details : Check that we actually have objects in a file before doing any
2706 operations on objects in ll_vm_open, ll_vm_close and
2712 Description: Request leak when working with deleted CWD
2713 Details : Introduce advanced request refcount tracking for requests
2714 referenced from lustre intent.
2716 Severity : Enhancement
2718 Description: Cache open file handles on client.
2719 Details : MDS now will return special lock along with openhandle, if
2720 requested and client is allowed to hold openhandle, even if unused,
2721 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2722 closing files for every read/write openration.
2724 Severity : Enhancement
2726 Description: Cache open negative dentries on client when possible.
2727 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2728 negative dentries on lock revocation.
2733 Description: Remounting a client read-only wasn't possible with a zconf mount
2734 Details : It wasn't possible to remount a client read-only with llmount.
2736 Severity : enhancement
2737 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2738 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2739 MPICH is included to add Lustre-specific ADIO interfaces.
2740 This is based closely on the UFS ADIO layer and only differs
2741 in file creation, in order to allow the OST striping to be set.
2742 This is user-contributed code and not supported by CFS.
2747 Description: extended inode attributes (immutable, append-only) work improperly
2748 when 2.4 and 2.6 kernels are used on client/server or vice versa
2749 Details : Introduce kernel-independent values for these flags.
2751 Severity : enhancement
2754 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2755 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2756 granularity. Now, Fractional MB granularity is be supported,
2757 this is very useful for low memory system.
2759 Severity : enhancement
2761 Description: Getattr by fid
2762 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2763 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2764 have locks and avoids getattr rpc requests for it on every stat.
2767 Frequency : Always, for filesystems larger than 2TB
2769 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2770 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2771 filesystems larger than 2TB to be mounted with mballoc enabled.
2774 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2776 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2777 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2778 filesystems larger than 2TB is mounted then the kernel will
2779 silently corrupt the start of the filesystem. CONFIG_LBD is
2780 enabled for all CFS-supported kernels, but the possibility of
2781 this happening with a modified kernel config exists.
2783 Severity : enhancement
2785 Description: add client O_DIRECT support for 2.6 kernels
2786 Details : It is now possible to do O_DIRECT reads and writes to files
2787 in the Lustre client mountpoint on 2.6 kernel clients.
2789 Severity : enhancement
2791 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2792 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2793 parallel, not waiting for response from every OST before sending
2794 a rpc to the next OST.
2799 Description: setattr vs write race when updating file timestamps
2800 Details : Client processes that update a file timestamp into the past
2801 right after writing to the file (e.g. tar) it is possible that
2802 the updated file modification time can be reset to the current
2803 time due to a race between processing the setattr and write RPC.
2805 Severity : enhancement
2807 Description: Bring 'lfs find' closer in line with regular Linux find.
2808 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2809 -print0 options and obtains all the needed info through the lustre
2812 Severity : enhancement
2814 Description: support up to 1024 configured devices on one node
2815 Details : change obd_dev array from statically allocated to dynamically
2816 allocated structs as they are first used to reduce memory usage
2821 Description: Flush dirty partially truncated pages during truncate
2822 Details : Immediatelly flush partially truncated pages in filter_setattr,
2823 this way we completely avoid having any pages in page cache on OST
2824 and can retire ugly workarounds during writes to flush such pages.
2829 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2830 Details : protect inode from truncation within vfs_unlink() context
2831 just take a reference before calling vfs_unlink() and release it
2832 when parent's i_sem is free.
2837 Description: last_id value checked outside lock on OST caused LASSERT failure
2838 Details : If there were multiple MDS->OST object precreate requests in
2839 flight, it was possible that the OST's last object id was checked
2840 outside a lock and incorrectly tripped an assertion. Move checks
2841 inside locks, and discard old precreate requests.
2844 Frequency : always, if extents are used on OSTs
2846 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2847 Details : OSTs using the extents format would not zero the high 16 bits of
2848 the index physical block number. This is not a problem for any
2849 OST filesystems smaller than 16TB, and no kernels support ext3
2850 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2851 new/modified files) and can be fixed for existing filesystems
2852 with e2fsprogs-1.39-cfs1.
2857 Description: import connection selection may be incorrect if timer wraps
2858 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2859 import connections to be ignored if the 32-bit jiffies counter
2860 wraps. Use a 64-bit jiffies counter.
2863 Frequency : very large clusters immediately after boot
2865 Description: LNET request buffers exhausted under heavy short-term load
2866 Details : If a large number of client requests are generated on a service
2867 that has previously never seen so many requests it is possible
2868 that the request buffer growth cannot keep up with the spike in
2869 demand. Instead of dropping incoming requests, they are held in
2870 the LND until the RPC service can accept more requests.
2873 Frequency : Sometimes during replay
2875 Description: Assertion failure in ll_local_open after replay.
2876 Details : If replay happened on an open request reply before we were able
2877 to set replay handler, reply will become not swabbed tripping the
2878 assertion in ll_local_open. Now we set the handler right after
2879 recognising of open request
2882 Frequency : very rare
2884 Description: kernel reports "badness in vsnprintf"
2885 Details : Reading from the "recovery_status" /proc file in small chunks
2886 may cause a negative length in lprocfs_obd_rd_recovery_status()
2887 call to vsnprintf() (which is otherwise harmless). Exit early
2888 if there is no more space in the output buffer.
2890 Severity : enhancement
2892 Description: clear OBD RPC statistics by writing to them
2893 Details : It is now possible to clear the OBD RPC statistics by writing
2894 to the "stats" file.
2899 Description: Client mtime is not the same on different clients after utimes
2900 Details : In some cases, the client was using the utimes() syscall on
2901 a file cached on another node. The clients now validate the
2902 ctime from the MDS + OSTs to determine which one is right.
2907 Description: Inability to activate failout mode
2908 Details : lconf script incorrectly assumed that in pythong string's numeric
2909 value is used in comparisons.
2912 Frequency : always with multiple stripes per file
2914 Description: Inefficient object allocation for mutli-stripe files
2915 Details : When selecting which OSTs to stripe files over, for files with
2916 a stripe count that divides evenly into the number of OSTs,
2917 the MDS is always picking the same starting OST for each file.
2918 Return the OST selection heuristic to the original design.
2923 Description: mount failures may take full timeout to return an error
2924 Details : Under some heavy load conditions it is possible that a
2925 failed mount can wait for the full obd_timeout interval,
2926 possibly several minutes, before reporting an error.
2927 Instead return an error as soon as the status is known.
2929 Frequency : quota enabled and large files being deleted
2931 Description: releasing more than 4GB of quota at once hangs OST
2932 Details : If a user deletes more than 4GB of files on a single OST it
2933 will cause the OST to spin in an infinite loop. Release
2934 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2939 Description: statfs data retrieved from /proc may be stale or zero
2940 Details : When reading per-device statfs data from /proc, in the
2941 {kbytes,files}_{total,free,avail} files, it may appear
2942 as zero or be out of date.
2945 Frequency : systems with MD RAID1 external journal devices
2947 Description: lconf's call to blkid is confused by RAID1 journal devices
2948 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2949 of returning all block devices that match the journal UUID.
2952 Frequency : always, for aggregate stripe size over 4GB
2954 Description: assertion fails when trying to use 4GB stripe size
2955 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2956 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2959 Frequency : always on ppc64
2961 Description: the first write on an ext3 filesystem with mballoc got stuck
2962 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2963 perform endianness conversion.
2965 ------------------------------------------------------------------------------
2967 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2969 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2970 INTEROPERATE with older versions automatically. Please read the
2971 user documentation before upgrading any part of a live system.
2972 * WARNING: Lustre networking configuration changes are required with
2973 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2976 * Support for kernels:
2977 2.6.9-22.0.2.EL (RHEL 4)
2978 2.6.5-7.244 (SLES 9)
2979 2.6.12.6 vanilla (kernel.org)
2982 Severity : enhancement
2983 Bugzilla : 7981/8208
2984 Description: Introduced Lustre Networking (LNET)
2985 Details : LNET is new networking infrastructure for Lustre, it includes
2986 a reorganized network configuration mode (see the user
2987 documentation for full details) as well as support for routing
2988 between different network fabrics. Lustre Networking Devices
2989 (LNDs) for the supported network fabrics have also been
2990 created for this new infrastructure.
2992 Severity : enhancement
2993 Description: Introduced Access control lists
2994 Details : clients can set ACLs on files and directories in order to have
2995 more fine-grained permissions than the standard Unix UGO+RWX.
2996 The MDS must be started with the "-o acl" mount option.
2998 Severity : enhancement
2999 Description: Introduced filesystem quotas
3000 Details : Administrators may now establish per-user quotas on the
3003 Severity : enhancement
3005 Description: Configuration change for the XT3
3006 The PTLLND is now used to run Lustre over Portals on the XT3
3007 The configure option(s) --with-cray-portals are no longer used.
3008 Rather --with-portals=<path-to-portals-includes> is used to
3009 enable building on the XT3. In addition to enable XT3 specific
3010 features the option --enable-cray-xt3 must be used.
3015 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3016 Details : A race between allocating a new llcd and re-getting the llcd_lock
3017 allowed another thread to grab newly-allocated llcd.
3019 Severity : enhancement
3021 Description: 2.6 OST async journal commit and locking fix to improve performance
3022 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3023 2.6 kernels are now async as they already were in 2.4 kernels so
3024 that they can commit concurrently with the network bulk transfer.
3025 For block-allocated files the filter allocation semaphore is held
3026 to avoid filesystem fragmentation during allocation. BKL lock
3027 removed for 2.6 xattr operations where it is no longer needed.
3032 Description: lconf incorrectly determined whether two IP networks could talk
3033 Details : In some more complicated routing and multiple-network
3034 configurations, lconf will avoid trying to make a network
3035 connection to a disjoint part of the IP space. It was doing the
3036 math incorrectly for one set of cases.
3041 Description: Fix for potential infinite loop processing records in an llog.
3042 Details : If an llog record is corrupted/zeroed, it is possible to loop
3043 forever in llog_process(). Validate the llog record length
3044 and skip the remainder of the block on error.
3047 Frequency : occasional (liblustre only)
3049 Description: liblustre could not open files whose last component is a symlink
3050 Details : sysio_path_walk() would incorrectly pass the open intent to
3051 intermediate path components.
3054 Frequency : rare (liblustre only with non-standard tuning)
3055 Bugzilla : 7201 (7350)
3056 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3057 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3058 after it has been released, only do this for referenced locks
3060 Severity : enhancement
3062 Description: specifying an (invalid) directory default stripe_size of -1
3063 would reset the directory default striping
3064 Details : stripe_size -1 was used internally to signal directory stripe
3065 removal, now use "all default" to signal dir stripe removal
3066 as a directory striping of "all default" is not useful
3069 Frequency : common for large clusters running liblustre clients
3071 Description: doing an ls when liblustre clients are running is slow
3072 Details : sending a glimpse AST to a liblustre client waits for every AST
3073 to time out, as liblustre clients will not respond. Since they
3074 cannot cache data we refresh the OST lock LVB from disk instead.
3076 Severity : enhancement
3078 Description: doing an ls at the same time as file IO can be slow
3079 Details : enqueue and other "small" requests can be blocked behind many
3080 large IO requests. Create a new OST IO portal for non-IO
3081 requests so they can be processed faster.
3084 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
3086 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
3087 created file with mode 000
3088 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
3089 allow an ftruncate() as a non-root user to a file with mode 000.
3090 Lustre now respects this flag to disable mode checks when
3091 truncating a file owned by the user
3094 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
3097 Description: Revoking locks from clients that went dead or catatonic might take
3099 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
3100 cancellation of such locks instant on servers without waiting for
3101 any reply from clients. Clients drops these locks when cancel
3102 notification from server is received without replying.
3105 Frequency : liblustre-only, when liblustre client dies or becomes busy
3107 Description: Doing ls on Linux clients can take a long time with active
3109 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
3110 granting such locks to it in the first place if possible. Locks
3111 are taken by proxy on the OST during the read or write and
3112 dropped immediately afterward. Add connect flags handling, do
3113 not grant locks to liblustre clients for glimpse ASTs.
3115 Severity : enhancement
3117 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
3118 Details : Existing read-ahead algorithm is tuned for the case of streamlined
3119 sequential reads and behaves badly with applications doing random
3120 reads. Improve it by reading ahead at least read region, and
3121 avoiding excessive large RPC for small reads.
3123 Severity : enhancement
3125 Description: Creating more than 1000 files for a single job may cause a load
3126 imbalance on the OSTs if there are also a large number of OSTs.
3127 Details : qos_prep_create() uses an OST index reseed value that is an
3128 even multiple of the number of available OSTs so that if the
3129 reseed happens in the middle of the object allocation it will
3130 still utilize the OSTs as uniformly as possible.
3135 Description: OST or MDS may oops in ping_evictor_main()
3136 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3137 but doesn't restart at beginning of obd_exports_timed list
3140 Severity : enhancement
3142 Description: improve by-nid export eviction on the MDS and OST
3143 Details : allow multiple exports with the same NID to be evicted at one
3144 time without re-searching the exports list.
3147 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3149 Description: MDS may oops in groups_free()
3150 Details : in rare race conditions a newly allocated group_info struct is
3151 freed again, and this can be NULL. The 2.4 compatibility code
3152 for groups_free() checked for a NULL pointer, but 2.6 did not.
3155 Frequency : common for liblustre clients doing little filesystem IO
3156 Bugzilla : 9352, 7313
3157 Description: server may evict liblustre clients accessing contended locks
3158 Details : if a client is granted a lock or receives a completion AST
3159 with a blocking AST already set it would not reply to the AST
3160 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3163 Frequency : lfs setstripe, only systems with more than 160 OSTs
3165 Description: unable to set striping with a starting offset beyond OST 160
3166 Details : llapi_create_file() incorrectly limited the starting stripe
3167 index to the maximum single-file stripe count.
3170 Frequency : LDAP users only
3172 Description: lconf did not handle in-kernel recovery with LDAP properly
3173 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3175 Severity : enhancement
3177 Description: bind OST threads to NUMA nodes to improve performance
3178 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3179 node and do their allocations there to localize memory access
3181 Severity : enhancement
3183 Description: llmount can determine client NID directly from Myrinet (GM)
3184 Details : the client NID code from gmnalnid was moved directly into
3185 llmount, removing the need to use this or specifying the
3186 client NID explicitly when mounting GM clients with zeroconf
3189 Frequency : if client is started with down MDS
3191 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3192 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3193 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3194 ptlrpc_import_delayed_req() because we didn't check if the
3195 request was interrupted, and we also didn't break out of the
3196 event loop if there was no timeout
3201 Description: data loss during non-page-aligned writes to a single file from
3202 both multiple nodes and multiple threads on one node at same time
3203 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3204 inconsistency led to false short-reads, that were cached and later
3205 used by ->prepare_write() to fill in partially written page,
3206 leading to data loss.
3209 Frequency : always, if lconf --abort_recovery used
3211 Description: lconf --abort_recovery fails with 'Operation not supported'
3212 Details : lconf was attempting to abort recovery on the MDT device and not
3215 Severity : enhancement
3217 Description: remove cleanup logs
3218 Details : replace lconf-generated cleanup logs with lustre internal
3219 cleanup routines. Eliminates the need for client-cleanup and
3222 Severity : enhancement
3224 Description: add support for EAs (user and system) on lustre filesystems
3225 Details : it is now possible to store extended attributes in the Lustre
3226 client filesystem, and with the user_xattr mount option it
3227 is possible to allow users to store EAs on their files also
3229 Severity : enhancement
3231 Description: Add possibility (config option) to show minimal available OST free
3233 Details : When compiled with --enable-mindf configure option, statfs(2)
3234 (and so, df) will return least minimal free space available from
3235 all OSTs as amount of free space on FS, instead of summary of
3236 free spaces of all OSTs.
3238 Severity : enhancement
3240 Description: do not expand extent locks acquired on OST-side
3241 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3242 by server: they are not cached anyway.
3245 Frequency : when mmap is used/binaries executed from Lustre
3247 Description: Unmmap pages before throwing them away from read cache.
3248 Details : llap_shrink cache now attempts to unmap pages before discarding
3249 them (if unmapping failed - do not discard). SLES9 kernel has
3250 extra checks that trigger if this unmapping is not done first.
3255 Description: lconf didn't resolve symlinks before checking to see whether a
3256 given mountpoint was already in use
3259 Frequency : when migrating failover services
3260 Bugzilla : 6395, 9514
3261 Description: When migrating a subset of services from a node (e.g. failback
3262 from a failover service node) the remaining services would
3263 time out and evict clients.
3264 Details : lconf --force (implied by --failover) sets the global obd_timeout
3265 to 5 seconds in order to quickly disconnect, but this caused
3266 other RPCs to time out too quickly. Do not change the global
3267 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3269 Severity : enhancement
3270 Frequency : if MDS is started with down OST
3271 Bugzilla : 9439,5706
3272 Description: Allow startup/shutdown of an MDS without depending on the
3273 availability of the OSTs.
3274 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3275 Add appropriate locking and lov-osc refcounts for safe
3276 cleaning. Add osc abort_inflight calls in case the
3277 synchronize never started.
3280 Frequency : occasional (Cray XT3 only)
3282 Description: root not authorized to access files in CRAY_PORTALS environment
3283 Details : The client process capabilities were not honoured on the MDS in
3284 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3285 been accessed by an authorized user then root was able to access
3286 the file on the local client also. The root user capabilities
3287 are now allowed on the MDS, as this environment has secure UID.
3290 Frequency : occasional
3292 Description: ldiskfs "too long searching" message happens too often
3293 Details : A debugging message (otherwise harmless) prints too often on
3294 the OST console. This has been reduced to only happen when
3295 there are fragmentation problems on the filesystem.
3300 Description: Division by zero in statfs when all OSCs are inactive
3301 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3302 lov_statfs divides by value returned by lov_get_stripecnt().
3306 Bugzilla : 9489, 3273
3307 Description: First write from each client to each OST was only 4kB in size,
3308 to initialize client writeback cache, which caused sub-optimal
3309 RPCs and poor layout on disk for the first writen file.
3310 Details : Clients now request an initial cache grant at (re)connect time
3311 and so that they can start streaming writes to the cache right
3312 away and always do full-sized RPCs if there is enough data.
3313 If the OST is rebooted the client also re-establishes its grant
3314 so that client cached writes will be honoured under the grant.
3319 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3320 Details : Now I/O RPCs go to different portal number and (presumably) fast
3321 lock requests (and glimses) and other RPCs get their own service
3322 threads pool that should be able to service those RPCs
3325 Severity : enhancement
3327 Description: Ability to exchange lustre version between client and servers and
3328 issue warnings at client side if client is too old. Also for
3329 liblustre clients there is ability to refuse connection of too old
3331 Details : New 'version' field is added to connect data structure that is
3332 filled with version info. That info is later checked by server and
3336 Frequency : rare, liblustre only.
3337 Bugzilla : 9296, 9581
3338 Description: Two simultaneous writes from liblustre at offset within same page
3339 might proceed at the same time overwriting eachother with stale
3341 Details : I/O lock withing llu_file_prwv was released too early, before data
3342 actually was hitting the wire. Extended lock-holding time until
3343 server acknowledges receiving data.
3346 Frequency : extremely rare. Never observed in practice.
3348 Description: avoid generating lustre_handle cookie of 0.
3349 Details : class_handle_hash() generates handle cookies by incrementing
3350 global counter, and can hit 0 occasionaly (this is unlikely, but
3351 not impossible, because initial value of cookie counter is
3352 selected randonly). Value of 0 is used as a sentinel meaning
3353 "unassigned handle" --- avoid it. Also coalesce two critical
3354 sections in this function into one.
3356 Severity : enhancement
3358 Description: allow liblustre clients to delegate truncate locking to OST
3359 Details : To avoid overhead of locking, liblustre client instructs OST to
3360 take extent lock in ost_punch() on client's behalf. New connection
3361 flag is added to handle backward compatibility.
3363 Severity : enhancement
3364 Bugzilla : 4928, 7341, 9758
3365 Description: allow number of OST service threads to be specified
3366 Details : a module parameter allows the number of OST service threads
3367 to be specified via "options ost ost_num_threads={N}" in the
3368 OSS's /etc/modules.conf or /etc/modprobe.conf.
3372 Bugzilla : 6146, 9635, 9895
3373 Description: servers crash with bad pointer in target_handle_connect()
3374 Details : In rare cases when a client is reconnecting it was possible that
3375 the connection request was the last reference for that export.
3376 We would temporarily drop the export reference and get a new
3377 one, but this may have been the last reference and the export
3378 was just destroyed. Get new reference before dropping old one.
3380 Severity : enhancement
3381 Frequency : if client is started with failover MDS
3383 Description: Allow multiple MDS hostnames in the mount command
3384 Details : Try to read the configuration from all specified MDS
3385 hostnames during a client mount in case the "primary"
3388 Severity : enhancement
3390 Description: Stop sending data to evicted clients as soon as possible.
3391 Details : Check if the client we are about to send or are sending data to
3392 was evicted already. (Check is done every second of waiting,
3393 for which l_wait_event interface was extended to allow checking
3394 of exit condition at specified intervals).
3397 Frequency : rare, normally only when NFS exporting is done from client
3399 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3401 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3404 Frequency : always, if srand() is called before liblustre initialization
3406 Description: Liblustre uses system PRNG disturbing its usage by user application
3407 Details : Introduce internal to lustre fast and high-quality PRNG for
3408 lustre usage and make liblustre and some other places in generic
3409 lustre code to use it.
3411 Severity : enhancement
3412 Bugzilla : 9477, 9557, 9870
3413 Description: Verify that the MDS configuration logs are updated when xml is
3414 Details : Check if the .xml configuration logs are newer than the config
3415 logs stored on the MDS and report an error if this is the case.
3416 Request --write-conf, or allow starting with --old_conf.
3418 Severity : enhancement
3420 Description: Handle symlinks in the path when checking if Lustre is mounted.
3421 Details : Resolve intermediate symlinks when checking if a client has
3422 mounted a filesystem to avoid duplicate client mounts.
3427 Description: lconf can hit an error exception but still return success.
3428 Details : The lconf command catches the Command error exception at the top
3429 level script context and will exit with the associated exit
3430 status, but doesn't ensure that this exit status is non-zero.
3435 Description: failure of ptlrpc thread startup can cause oops
3436 Details : Starting a ptlrpc service thread can fail if there are a large
3437 number of threads or the server memory is very fragmented.
3438 Handle this without oopsing.
3441 Frequency : always, only if liblustre and non-default acceptor port was used
3443 Description: liblustre cannot connect to servers with non-default acceptor port
3444 Details : tcpnal_set_default_params() was not called and was therefore
3445 ignoring the environment varaible TCPNAL_PORT, as well as other
3446 TCPNAL_ environment variables
3451 Description: two objects could be created on the same OST for a single file
3452 Details : If an OST is down, in some cases it was possible to create two
3453 objects on a single OST for a single file. No problems other
3454 than potential performance impact and spurious error messages.
3458 Bugzilla : 5681, 9562
3459 Description: Client may oops in ll_unhash_aliases
3460 Details : Client dcache may become inconsistent in race condition.
3461 In some cases "getcwd" can fail if the current directory is
3467 Description: Inode refcounting problems in NFS export code
3468 Details : link_raw functions used to call d_instantiate without obtaining
3469 extra inode reference first.
3473 Bugzilla : 9942, 9903
3474 Description: Referencing freed requests leading to crash, memleaks with NFS.
3475 Details : We used to require that call to ll_revalidate_it was always
3476 followed by ll_lookup_it. Also with revalidate_special() it is
3477 possible to call ll_revalidate_it() twice for the same dentry
3478 even if first occurence returned success. This fix changes semantic
3479 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3480 reference on a request referred from the intent.
3481 ll_intent_release() then releases such a request.
3484 Frequency : rare, normally benchmark loads only
3486 Description: unlinked inodes were kept in memory on the client
3487 Details : If a client is repeatedly creating and unlinking files it
3488 can accumulate a lot of stale inodes in the inode slab cache.
3489 If there is no other client load running this can cause the
3490 client node to run out of memory. Instead flush old inodes
3491 from client cache that have the same inode number as a new inode.
3494 Frequency : SLES9 2.6.5 kernel and long filenames only
3495 Bugzilla : 9969, 10379
3496 Description: utime reports stale NFS file handle
3497 Details : SLES9 uses out-of-dentry names in some cases, which confused
3498 the lustre dentry revalidation. Change it to always use the
3502 Frequency : rare, unless heavy write-truncate concurrency is continuous
3503 Bugzilla : 4180, 6984, 7171, 9963, 9331
3504 Description: OST becomes very slow and/or deadlocked during object unlink
3505 Details : filter_destroy() was holding onto the parent directory lock
3506 while truncating+unlinking objects. For very large objects this
3507 may block other threads for a long time and slow overall OST
3508 responsiveness. It may also be possible to get a lock ordering
3509 deadlock in this case, or run out of journal credits because of
3510 the combined truncate+unlink. Solution is to do object truncate
3511 first in one transaction without parent lock, and then do the
3512 final unlink in a new transaction with the parent lock. This
3513 reduces the lock hold time dramatically.
3516 Frequency : rare, 2.4 kernels only
3518 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3519 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3520 before unmounting their filesystem in order to avoid deadlock.
3521 The kernel_locked() function in 2.4 kernels only checks whether
3522 the kernel lock is held, not whether it is this process that is
3523 holding it as 2.6 kernels do.
3528 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3529 Details : The client ptlrpc code may be trying to reconnect to a down
3530 server before a previous connection attempt has timed out.
3531 Increase the reconnect interval to be longer than the connection
3532 timeout interval to avoid sending duplicate connections to
3536 Frequency : echo_client brw_test command
3538 Description: fix echo_client to work with OST preallocated code
3539 Details : OST preallocation code (5137) didn't take echo_client IO path
3540 into account: echo_client calls filter methods outside of any
3541 OST thread and, hence, there is no per-thread preallocated
3542 pages and buffers to use. Solution: hijack pga pages for IO. As
3543 a byproduct, this avoids unnecessary data copying.
3547 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3548 Description: Client can oops in mdc_commit_close() after open replay
3549 Details : It was possible for the MDS to return an open request with no
3550 transaction number in mds_finish_transno() if the client was
3551 evicted, but without actually returning an error. Clients
3552 would later try to replay that open and may trip an assertion
3553 Simplify the client close codepath, and always return an error
3554 from the MDS in case the open is not successful.
3557 Frequency : rare, 2.6 OSTs only
3559 Description: OST may deadlock under high load on fragmented files
3560 Details : If there was a heavy load and highly-fragmented OST filesystems
3561 it was possible to have all the OST threads deadlock waiting on
3562 allocation of biovecs, because the biovecs were not released
3563 until the entire RPC IO was completed. Instead, release biovecs
3564 as soon as they are complete to ensure forward IO progress.
3566 Severity : enhancement
3568 Description: Support for specifying external journal device at mount
3569 Details : If an OST or MDS device is formatted with an external journal
3570 device, this device major/minor is stored in the ext3 superblock
3571 and may not be valid for failover. Allow detecting and
3572 specifying the external journal at mount time.
3577 Description: Mounting an MDS with pending unlinked files may cause oops
3578 Details : target_finish_recovery() calls mds_postrecov() which returned
3579 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3580 considers this an error and immediately begins cleaning up the
3581 lov, just after starting the mds_lov process
3583 Severity : enhancement
3585 Description: Implement 'lfs df' to report actual free space on per-OST basis
3586 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3587 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3588 usage of objects; '-h' to report in human readable format.
3590 ------------------------------------------------------------------------------
3592 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3599 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3600 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3601 at startup during orphan unlinking. Instead, skip buddy update
3602 before setup, as it will be regenerated after recovery is complete.
3607 Description: If an OST is inactive, its locks might reference stale inodes.
3608 Details : lov_change_cbdata() must iterate over all namespaces, even if
3609 they are inactive to clear inode references from the lock.
3611 Severity : enhancement
3612 Frequency : occasional, if non-standard max_dirty_mb used
3614 Description: Client will block write RPCs if not enough grant
3615 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3616 then the client will block writes while waiting for another RPC
3617 to complete instead of consuming its dirty limit. With change
3618 we get improved performance when max_dirty_mb is small.
3620 Severity : enhancement
3621 Bugzilla : 3389, 6253
3622 Description: Add support for supplementary groups on the MDS.
3623 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3624 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3625 lookups for user supplementary groups into a cache.
3629 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3630 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3631 for liblustre O_EXCL is enforced by MDS.
3635 Description: readdir never returns NULL in liblustre.
3636 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3637 dentry in struct dirent.
3641 Frequency : liblustre only, depends on application IO pattern
3642 Description: liblustre clients evicted if not contacting servers
3643 Details : Don't put liblustre clients into the ping_evictor list, so
3644 they will not be evicted by the pinger ever.
3646 Severity : enhancement
3648 Description: Add ability to evict clients by NID from MDS.
3649 Details : By echoing "nid:$NID" string into
3650 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3651 $NID would be instantly evicted from this MDS and from all active
3652 OSTs connected to it.
3656 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3657 Details : lookup_it_finish() used to query file size from OSTs that was not
3662 Description: service threads change working directory to that of init
3663 Details : Starting lustre service threads may pin the working directory
3664 of the parent thread, making that filesystem busy. Threads
3665 now change to the working directory of init to avoid this.
3669 Frequency : during shutdown only
3670 Description: shutdown with a failed MDS or OST can cause unmount to hang
3671 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3676 Frequency : 2.6 only
3677 Description: chmod/chown may include an extra supplementary group
3678 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3679 supplementary group and if none is specified this is used.
3682 Bugzilla : 5479 (6816)
3684 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3685 Details : If lookup is in progress on a file that is unlinked we might try
3686 to revalidate the inode and fail in revalidate after lookup is
3687 complete and ll_file_open() enqueues the open again but
3688 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3691 Frequency : always, if lconf --abort_recovery used
3693 Description: lconf --abort_recovery fails with 'Operation not supported'
3694 Details : lconf was attempting to abort recovery on the MDT device and not
3697 ------------------------------------------------------------------------------
3699 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3704 Frequency : rare (only unsupported configurations with a node running as an
3706 Bugzilla : 6514, 5137
3707 Description: Mounting a Lustre file system on a node running as an OST could
3709 Details : OSTs now preallocates memory needed to write out data at
3710 startup, instead of when needed, to avoid having to
3711 allocate memory in possibly low memory situations.
3712 Specifically, if the file system is mounted on on OST,
3713 memory pressure could force it to try to write out data,
3714 which it needed to allocate memory to do. Due to the low
3715 memory, it would be unable to do so and the node would
3716 become unresponsive.
3718 Severity : enhancement
3720 Description: Addition of lconf --service command line option
3721 Details : lconf now accepts a '--service <arg>' option, which is
3722 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3724 Severity : enhancement
3726 Description: Failover mode is now the default for OSTs.
3727 Details : By default, OSTs will now run in failover mode. To return to
3728 the old behaviour, add '--failout' to the lmc line for OSTs.
3730 Severity : enhancement
3732 Description: Health checks are now provided for MDS and OSTs
3733 Details : Additional detailed health check information on MSD and OSTs
3734 is now provided through the procfs health_check value.
3737 Frequency : occasional, depends on IO load
3739 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3740 after numerous create/delete cycles
3741 Details : The ext3 inode allocation policy would not allocate new inodes
3742 very well on the OSTs because there are no new directories
3743 being created. Instead we look for groups with free space if
3744 the parent directories are nearly full.
3749 Description: Network or server problems during mount may cause partially
3750 mounted clients instead of returning an error.
3751 Details : The config llog parsing code may overwrite the error return
3752 code during mount error handling, returning success instead
3758 Description: MDS can fail to allocate large reply buffers
3759 Details : After long uptimes the MDS can fail to allocate large reply
3760 buffers (e.g. zconf client mount config records) due to memory
3761 fragmentation or consumption by the buffer cache. Preallocate
3762 some large reply buffers so that these replies can be sent even
3763 under memory pressure.
3767 Frequency : rare (liblustre)
3768 Description: fsx running with liblustre complained that using truncate() to
3769 extend the file doesn't work. This patch corrects that issue.
3770 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3771 ATTR_SIZE and lsm use in llu_setattr_raw.
3775 Frequency : rare, only 2.6 kernels
3776 Description: Unusual file access patterns on the MDS may result in inode
3777 data being lost in very rare circumstances.
3778 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3779 code in the 2.6 kernel caused the inode and/or EA data not to
3780 be read from disk, causing single-file corruption.
3784 Frequency : rare, only 2.6 filesystems using extents
3785 Description: Heavy concurrent write and delete load may cause data corruption.
3786 Details : It was possible under high-load situations to have an extent
3787 metadata block in the block device cache from a just-unlinked
3788 file overwrite a newly-allocated data block. We now unmap any
3789 metadata buffers that alias just-allocated data blocks.
3793 Frequency : filesystems with default stripe_count larger than 77
3794 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3795 Details : lconf specifies an inode size of 4096 bytes when the default
3796 stripe_count is larger than 77. This conflicts with the default
3797 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3798 this case to avoid pinning too much memory for large EAs.
3800 ------------------------------------------------------------------------------
3802 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3807 Frequency : rare (extremely heavy IO load with hundreds of clients)
3809 Description: Client is evicted, gets IO error writing to file
3810 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3811 caused clients to be evicted for AST timeouts. The fixes in
3812 bug 5192 mean we no longer need to have such short AST timeouts
3813 so ldlm_timeout has been increased.
3816 Frequency : occasional during --force or --failover shutdown under load
3817 Bugzilla : 5949, 4834
3818 Description: Server oops/LBUG if stopped with --force or --failover under load
3819 Details : a collection of import/export refcount and cleanup ordering
3820 issues fixed for safer force cleanup
3823 Frequency : only filesystems larger than 120 OSTs
3824 Bugzilla : 5990, 6223
3825 Description: lfs getstripe would oops on a very large filesystem
3826 Details : lov_getconfig used kfree on vmalloc'd memory
3829 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3830 Bugzilla : 6242, 6243
3831 Description: reading from files that had been truncated to a non-zero size
3832 but never opened returned no data
3833 Details : ll_file_read() reads zeros from no-object files to EOF
3838 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3839 Details : There's little harm in aborting MDS/OSS recovery and letting it
3840 try again, so I removed the LASSERT and return an error instead.
3842 Severity : enhancement
3844 Description: New debugging infrastructure for tracking down data corruption
3845 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3846 (b) cover more of the client-side code path, and (c) try to narrow
3847 down where problems occurred
3851 Bugzilla : 3819, 4364, 4397, 6313
3852 Description: Racing close and eviction MDS could cause assertion in mds_close
3853 Details : It was possible to get multiple mfd references during close and
3854 client eviction, leading to one thread referencing a freed mfd.
3856 Severity: : enhancement
3857 Bugzilla : 3262, 6359
3858 Description: Attempts to reconnect to servers are now more aggressive.
3859 Details : This builds on the enhanced upcall-less recovery that was added
3860 in 1.4.2. When trying to reconnect to servers, clients will
3861 now try each server in the failover group every 10 seconds. By
3862 default, clients would previously try one server every 25 seconds.
3867 Description: After recovery, certain operations trigger a failed
3868 assertion on a client.
3869 Details : Failing over an mds, using lconf -d --failover, while a
3870 client was doing a readdir() call would cause the client to
3871 LBUG after recovery completed and the readdir() was resent.
3873 Severity : enhancement
3875 Description: Default groups are now added by lconf
3876 Details : You can now run lconf --group <servicename> without having to
3877 manually add groups with lmc.
3880 Frequency : occasional
3882 Description: Nodes with an elan id of 0 trigger a failed assertion
3885 Frequency : always when accessing e.g. tty/console device nodes
3887 Description: tty and some other devices nodes cannot be used on lustre
3888 Details : file's private_data field is used by device data and lustre
3889 values in there got lost. New field was added to struct file to
3890 store fs-specific private data.
3893 Frequency : when exporting Lustre via NFS
3895 Description: NFSD failed occasionally when looking up a path component
3896 Details : NFSD is looking up ".." which was broken in ext3 directories
3897 that had grown large enough to become hashed.
3900 Frequency : Clusters with multiple interfaces not on the same subnet
3902 Description: Nodes will repeatedly try to reconnect to an interface which it
3903 cannot reach and report an error to the log.
3904 Details : Extra peer list entries will be created by lconf with some peers
3905 unreachable. lconf now validates the peer before adding it.
3908 Frequency : Only if a default stripe is set on the filesystem root.
3910 Description: Setting a default stripe on the filesystem root prevented the
3911 filesystem from being remounted.
3912 Details : The client was sending extra request flags in the root getattr
3913 request and did not allocate a reply buffer for the dir EA.
3916 Frequency : occasional, higher if lots of files are accessed by one client
3917 Bugzilla : 6159, 6097
3918 Description: Client trips assertion regarding lsm mismatch/magic
3919 Details : While revalidating inodes the VFS looks up inodes with ifind()
3920 and in rare cases can find an inode that is being freed.
3921 The ll_test_inode() code will free the lsm during ifind()
3922 when it finds an existing inode and then the VFS later attaches
3923 this free lsm to a new inode.
3927 Bugzilla : 6422, 7030
3928 Description: MDS deadlock between mkdir and client eviction
3929 Details : Creating a new file via mkdir or mknod (starting a transaction
3930 and getting the ns lock) can deadlock with client eviction
3931 (gets ns lock and trying to finish a synchronous transaction).
3934 Frequency : occasional
3935 Description: While starting a server, the fsfilt_ext3 module could not be
3937 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3938 kernels. Previously, lconf would still use the ext3 name
3939 when trying to load modules. Now, it will correctly use
3940 ext3 on 2.4 and ldiskfs on 2.6.
3942 Severity : enhancement
3943 Description: The default stripe count has been changed to 1
3944 Details : The interpretation of the default stripe count (0, to lfs
3945 or lmc) has been changed to mean striping across a single
3946 OST, rather than all available. For general usage we have
3947 found a stripe count of 1 or 2 works best.
3949 Severity : enhancement
3950 Description: Add support for compiling against Cray portals.
3951 Details : Conditional compiling for some areas that are different
3955 Frequency : occasional
3956 Bugzilla : 6409, 6834
3957 Description: Creating files with an explicit stripe count may lead to
3958 a failed assertion on the MDS
3959 Details : If some OSTs are full or unavailable, creating files may
3960 trigger a failed assertion on the MDS. Now, Lustre will
3961 try to use other servers or return an error to the
3965 Frequency : occasional
3967 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3968 caused lock timeout during readahead (since 1.4.2).
3969 Details : Processes doing readahead might match a lock that hasn't been
3970 granted yet if there are overlapping and conflicting lock
3971 requests. The readahead process waits on ungranted lock
3972 (original lock is CBPENDING), while OST waits for that process
3973 to cancel CBPENDING read lock and eventually evicts client.
3975 Severity : enhancement
3977 Description: Initial enabling of flock support for clients
3978 Details : Implements fcntl advisory locking and file status functions.
3979 This feature is provided as an optional mount flag (default
3980 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3981 locking are implemented yet, and those that are are not guaranteed
3982 to be completely correct in production environments.
3983 mount -t lustre -o [flock|noflock] ...
3986 Frequency : occasional
3988 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3989 assertion in the ext3 JBD (journaling) layer.
3990 Details : The b_committed_data struct is protected by the big kernel lock
3991 in 2.4 kernels, serializing journal_commit_transaction() and
3992 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3993 access to this struct. In 2.6 kernels there is finer grained
3994 locking to improve SMP performance of the JBD layer.
3998 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4003 Description: merge and cleanup kernel patches.
4006 -----------------------------------------------------------------------------
4008 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4010 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4011 versions of Lustre. Please update all servers and clients to
4012 version 1.4.2 or later at the same time. You must also run
4013 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4014 to update the configuration logs.
4016 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4017 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4018 - swab llog records of type '0' so we get proper header size/idx (5861)
4019 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4020 - fix rename of one directory over another leaking an inode (5953)
4021 - avoid SetPageDirty on 2.6 (5981)
4022 - don't re-add just-being-destroyed locks to the waiting list (5653)
4023 - when creating new directories, inherit the parent's custom
4024 striping settings if present parent (3048)
4025 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4026 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4027 - don't hold client locks on temporary worklist from l_lru (5666)
4028 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4029 - automatically evict dead clients (5921)
4030 - Update file size properly in create+truncate+fstat case (6196)
4031 - Do not unhash mountpoint dentries, do not allow removal of
4033 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4034 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4035 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4036 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4037 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4038 - Correctly handle reads of files with no objects (6243)
4039 - lctl recover will also mark a device active if deactivate used (5933)
4041 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4042 - allow --write-conf on an MDS with different nettype than client (5619)
4043 - don't write config llogs to MDS for mounts not from that MDS (5617)
4044 - lconf should create multiple TCP connections from a client (5201)
4045 - init scripts are now turned off by default; run chkconfig --on
4046 lustre and chkconfig --on lustrefs to use them
4047 - upcalls are no longer needed for clients to recover to failover
4049 - add --abort-recovery option to lconf to abort recovery on device
4051 - add support for an arbitrary number of OSTs (3026)
4052 - Quota support protocol changes.
4053 - forward compatibility changes to wire structs (6007)
4054 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4055 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4056 - improved "lustre status" script
4057 - initialize blocksize for non-regular files (6062)
4058 - added --disable-server and --disable-client configure options (5782)
4059 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4060 - Vanilla 2.4.29 support
4061 - increase maximum number of obd devices to 520 (6242)
4062 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4063 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4064 SLES 9 2.6.5 kernels
4065 - limit stripes per file to 160 (the maximum EA size) (6093)
4067 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4070 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
4071 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
4072 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
4073 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
4074 - small ext3 extents cleanups and fixes (5733)
4075 - improved mballoc code, several small races and bugs fixed (5733, 5638)
4076 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
4077 - avoid needless client->OST connect, fix handle mismatch (5317)
4078 - fix DLM error path that led to out-of-sync client, long delays (5779)
4079 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
4080 - fix several locking issues related to i_size (5492,5624,5654,5672)
4081 - don't move pending lock onto export if it is already evicted (5683)
4082 - fix kernel oops when creating .foo in unlinked directory (5548)
4083 - fix deadlock in obdfilter statistics vs. object create (5811)
4084 - use time_{before,after} to avoid timer jiffies wrap (5882)
4085 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
4086 - Do not leak request if server was not able to process it (5154)
4087 - If mds_open unable to find parent dir, make that negative lookup(5154)
4088 - don't create new directories with extent-mapping (5909, 5936)
4090 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
4091 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
4092 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
4093 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
4094 - support for RHEL4 kernel on i686 (5773)
4095 - provide error messages when incompatible logs are encountered (5898)
4097 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
4098 * version 1.4.0.10 (1.4.1 release candidate 1)
4100 - don't keep a lock reference when lock is not granted (4238)
4101 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4102 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
4103 - drop import inflight refcount on signal_completed_replay error (5255)
4104 - unlock page after async write error during send (3677)
4105 - handle missing objects in filter_preprw_read properly (5265)
4106 - no transno return for symlink open, don't save no-trasno open (3440)
4107 - don't try to complete elan receive that already failed (4012)
4108 - free RPC server reply state on error (5406)
4109 - clean up thread from ptlrpc_start_thread() on error (5160)
4110 - readahead could read extra page into cache that wasn't ejected (5388)
4111 - prevent races in class_attach/setup/cleanup/detach (5260)
4112 - don't dereference de->d_inode after l_dput of de (5458)
4113 - use "int" for stripe value returned from lock_to_stripe (5544)
4114 - mballoc allocation and error-checking fixes in 2.6 (5504)
4115 - block device patches to fix I/O request sizes in 2.6 (5482)
4116 - look up hostnames for IB nals (5602)
4117 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
4118 - don't start multiple acceptors for the same port (5277)
4119 - fix incorrect LASSERT in mds_getattr_name (5635)
4120 - export a proc file for general "ping" checking (5628)
4121 - fix "lfs check" to not block when the MDS is down (5628)
4123 - service request history (4965)
4124 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4125 - create an "evict_client" /proc entry on OSTs, like the MDS has
4126 - fix mount usage message, return errors per mount(8) (5168)
4127 - change grep [] to grep "[]" in tests so they work in more UMLs
4128 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4129 - remove ancient LOV_MAGIC_V0 EA support (5047)
4130 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4131 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4132 - allow readahead allocations to fail when low on memory (5383)
4133 - mmap locking landed again, after considerable improvement (2828)
4134 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4136 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4139 - send OST transaction number in read/write reply to free req (4966)
4140 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4141 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4142 - the watchdog thread now runs as interruptible (5246)
4143 - flock/lockf fixes (but it's still disabled, pending 5135)
4144 - don't use EXT3 constants in llite code (5094)
4145 - memory shortage at startup could cause assertion (5176)
4147 - reorganization of lov code
4148 - single portals codebase
4150 - add extents/mballoc support (5025)
4151 - direct I/O reads in the obdfilter (4048)
4152 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4154 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4157 - send OST transaction number in read/write reply to free req (4966)
4158 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4159 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4160 - the watchdog thread now runs as interruptible (5246)
4161 - handle missing objects in filter_preprw_read properly (5265)
4162 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4163 - drop import inflight refcount on signal_completed_replay error (5255)
4164 - unlock page after async write error during send (3677)
4165 - return original error code on reconstructed replies (3761)
4166 - no transno return for symlink open, don't save no-trasno open (3440)
4168 - add pid to ldlm debugging output (4922)
4169 - bump the watchdog timeouts -- we can't handle 30sec yet
4170 - extra debugging for orphan dentry/inode bug (5259)
4172 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4175 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4176 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4177 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4178 - fix NULL dereference in /proc/sys/portals/routes (4827)
4179 - allow failed mdc_close() operations to be interrupted (4561)
4180 - stop precreate on OST before MDS would time out on it (4778)
4181 - don't send partial-page writes before EOF from client (4410)
4182 - discard client grant for sub-page writes on large-page clients (4520)
4183 - don't free dentries not owned by NFS code, check generation (4806)
4184 - fix lsm leak if mds_create_objects() fails (4801)
4185 - limit debug_daemon file size, always print CERROR messages (4789)
4186 - use transno after validating reply (3892)
4187 - process timed out requests if import state changes (3754)
4188 - update mtime on OST during writes, return in glimpse (4829)
4189 - add mkfsoptions to LDAP (4679)
4190 - use ->max_readahead method instead of zapping global ra (5039)
4191 - don't interrupt __l_wait_event() during strace
4193 - add software watchdogs to catch hung threads quickly (4941)
4194 - make lustrefs init script start after nfs is mounted
4195 - fix CWARN/ERROR duplication (4930)
4196 - return async write errors to application if possible (2248)
4197 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4198 - print NAL number in %x format (4645)
4199 - update barely-supported suse-2.4.21-171 series (4842)
4200 - support for sles 9 %post scripts
4201 - support for building 2.6 kernel-source packages
4202 - support for sles km_* packages
4204 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4207 - ignore -ENOENT errors in osc_destroy (3639)
4208 - notify osc create thread that OSC is being cleaned up (4600)
4209 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4210 - reconstruct ost_handle() like mds_handle() (4657)
4211 - create a new thread to do import eviction to avoid deadlock (3969)
4212 - let lconf resolve symlinked-to devices (4629)
4213 - don't unlink "objects" from directory with default EA (4554)
4214 - hold socknal file ref over connect in case target is down (4394)
4215 - allow more than 32000 subdirectories in a single directory (3244)
4216 - fix blocks count for O_DIRECT writes (3751)
4217 - OST returns ENOSPC from object create when no space left (4539)
4218 - don't send truncate RPC if file size isn't changing (4410)
4219 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4220 - bind to privileged port in socknal and tcpnal (3689)
4222 - rate limit CERROR/CWARN console message to avoid overload (4519)
4223 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4224 - basic mmap support (3918)
4225 - kernel patch series update from b1_4 (4711)
4227 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4230 - avoid crash during MDS cleanup with OST shut down (2775)
4231 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4232 - don't use bad inodes on the MDS (3744)
4233 - dynamic object preallocation to improve recovery speed (4236)
4234 - don't hold spinlock over lock dumping or change debug flags (4401)
4235 - don't zero obd_dev when it is force cleaned (3651)
4236 - print grants to console if they go negative (4431)
4237 - "lctl deactivate" will stop automatic recovery attempts (3406)
4238 - look for existing locks in ldlm_handle_enqueue() (3764)
4239 - don't resolve lock handle twice in recovery avoiding race (4401)
4240 - revalidate should check working dir is a directory (4134)
4242 - don't always mark "slow" obdfilter messages as errors (4418)
4244 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4247 - don't close LustreDB during write_conf until it is done (3860)
4248 - fix typo in lconf for_each_profile (3821)
4249 - allow dumping logs from multiple threads at one time (3820)
4250 - don't allow multiple threads in OSC recovery (3812)
4251 - fix debug_size parameters (3864)
4252 - fix mds_postrecov to initialize import for llog ctxt (3121)
4253 - replace config semaphore with spinlock (3306)
4254 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4255 - don't allow enqueue to complete on a destroyed export (3822)
4256 - down write_lock before checking llog header bitmap (3825)
4257 - recover from lock replay timeout (3764)
4258 - up llog sem before sending rpc (3652)
4259 - reduce ns lock hold times when setting kms (3267)
4260 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4261 - fix NULL deref and obd_dev leak on setup error (3312)
4262 - replace some LBUG about llog ops with error handling (3841)
4263 - don't match INVALID dentries from d_lookup and spin (3784)
4264 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4265 - fix invalid assertion in ptlrpc_set_wait (3880)
4267 - add libwrap support for the TCP acceptor (3996)
4268 - add /proc/sys/portals/routes for non-root route listing (3994)
4269 - allow setting MDS UUID in .xml (2580)
4270 - print the stack of a process that LBUGs (4228)
4272 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4275 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4276 - make sure to unset replay flag from failed open requests (3440)
4277 - if default stripe count is 0, use OST count for inode size (3636)
4278 - update parent mtime/ctime on client for create/unlink (2611)
4279 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4280 - free recovery state on server during a forced cleanup (3571)
4281 - unregister_reply for resent reqs (3063)
4282 - loop back devices mounting and status check on 2.6 (3563)
4283 - fix resource-creation race that can provoke i_size == 0 (3513)
4284 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4285 - more debugging for page-accounting assertion (3746)
4286 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4287 - avoid deadlock after precreation failure (3758)
4288 - fix race and lock order deadlock in orphan handling (3450, 3750)
4289 - add validity checks when grabbing inodes from l_ast_data (3599)
4291 - add /proc/.../recovery_status to obdfilter (3428)
4292 - lightweight CDEBUG infrastructure, debug daemon (3668)
4293 - change default OSC RPC parameters to be better on small clusters
4294 - turn off OST read cache for files smaller than 32MB
4295 - install man pages and include them in rpms (3100)
4296 - add new init script for (un)mounting lustre filesystems (2593)
4297 - run chkconfig in %post for init scripts (3701)
4298 - drop scimac NAL (unmaintained)
4300 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4303 - clean kiobufs before and after use (3485)
4304 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4305 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4306 - change init script to fail more clearly if not run as root (1528)
4307 - allow clients to reconnect during replay (1742)
4308 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4309 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4310 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4311 - fix cond_resched (3554)
4312 - only evict unfinished clients after recovery (3515)
4313 - allow bulk resend, prevent data loss (3570)
4314 - dynamic ptlrpc request buffer allocation (2102)
4315 - don't allow unlinking open directory if it isn't empty (2904)
4316 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4317 - remove extraneous obd dereference causing LASSERT failure (3334)
4318 - don't use get_cycles() when creating temp. files on the mds (3156)
4319 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4320 - handle EEXIST for set-stripe, set proper directory name (3336)
4322 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4323 - fix ksocknal_fmb_callback() error messages (2918)
4325 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4328 - don't copy lvb into (possibly NULL) reply on error (2983)
4329 - don't deref dentry after dput, don't free lvb on error (2922)
4330 - use the kms to determine writeback rpc length (2947)
4331 - increment oti_logcookies when osc is inactive (2948)
4332 - update client's i_blocks count via lvb messages (2543)
4333 - handle intent open/close of special files properly (1557)
4334 - mount MDS with errors=remount-ro, like obdfilter (2009)
4335 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4336 - don't use cancelling-locks' kms values (2947)
4337 - use highest lock extent for kms, not last one (2925)
4338 - don't dereference ERR_PTR() dentry in error handling path (3107)
4339 - fix thread race in portals_debug_dumplog() (3122)
4340 - create lprocfs device entries at setup instead of at attach (1519)
4341 - common AST error handler, don't evict client on completion race (3145)
4342 - zero nameidata in detach_mnt in 2.6 (3118)
4343 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4344 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4345 - initialize RPC timeout timer earlier for 2.6 (3219)
4346 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4347 - print nal/nid for unknown nid (3258)
4348 - additional checks for oscc recovery before doing precreate (3284)
4349 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4350 - don't crash in mdc_close for bad permissions on open (3285)
4351 - zero i_rdev for non-device files (3147)
4352 - clear page->private before handing to FS, better assertion (3119)
4353 - tune the read pipeline (3236)
4354 - fix incorrect decref of invalidated dentry (2350)
4355 - provide read-ahead stats and refine rpc in flight stats (3328)
4356 - don't hold journal transaction open across create RPC (3313)
4357 - update atime on MDS at close time (3265)
4358 - close LDAP connection when recovering to avoid server load (3315)
4359 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4360 - don't leak open file on MDS after open resend (3325)
4361 - serialize filter_precreate and filter_destroy_precreated (3329)
4362 - loop device shouldn't call sync_dev() for nul device (3092)
4363 - clear page cache after eviction (2766)
4364 - resynchronize MDS->OST in background (2824)
4365 - refuse to mount the same filesystem twice on same mountpoint (3394)
4366 - allow llmount to create routes for mounting behind routers (3320)
4367 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4368 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4369 - fix rare problem with rename on htree directories (3417)
4371 - allow default OST striping configuration per directory (1414)
4372 - fix compilation for qswnal for 2.6 kernels (3125)
4373 - increase maximum number of MDS request buffers for large systems
4374 - change liblustreapi to be useful for external progs like lfsck (3098)
4375 - increase local configuration timeout for slow disks (3353)
4376 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4378 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4381 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4382 - don't overwrite extent policy data in reply if lock was blocked (2901)
4383 - drop filter export grants atomically with removal from device (2663)
4384 - del obd_self_export from work_list in class_disconnect_exports (2908)
4385 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4386 - swab reply message in mdc_close, other PPC fixes (2464)
4387 - fix destroying of named logs (2325)
4388 - overwrite old logs when running lconf --write_conf (2264)
4389 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4390 - fix race in target_handle_connect (2898)
4391 - mds_reint_create() should take same inode create lock (2926)
4392 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4393 - don't close files for self_export to avoid uninitialized obd (2936)
4394 - allow MDS with the same name as client node (2939)
4395 - hold dentry reference for closed log files for unlink (2325)
4396 - reserve space for all logs during transactions (2059)
4397 - don't evict page beyond end of stripe extent (2925)
4398 - don't oops on a deleted current working directory (2399)
4399 - handle hard links to targets without a parent properly (2517)
4400 - don't dereference NULL lock when racing during eviction (2867)
4401 - don't grow lock extents when lots of conflicting locks (2919)
4403 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4406 - account for cache space usage on clients to avoid data loss (974)
4407 - lfsck support in lustre kernel code (2349)
4408 - reduce journal credits needed for BRW writes (2370)
4409 - orphan handling to avoid losing space on client/server crashes
4410 - ptlrpcd can be blocked, stopping ALL progress (2477)
4411 - use lock value blocks to assist in proper KMS, faster stat (1021)
4412 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4413 - recovery for initial connections (2355)
4414 - fixes for mds_cleanup_orphans (1934)
4415 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4416 - block all file creations until orphan recovery completes (1901)
4417 - client remove rq_connection from request struct (2423)
4418 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4419 - recovery timer race (2670)
4420 - mdc_close recovey bug (2532)
4421 - ptlrpc cleanup bug (2710)
4422 - mds timeout on local locks (2588)
4423 - namespace lock held during RPCs (2431)
4424 - handle interrupted sync write properly (2503)
4425 - don't try to handle a message that hasn't been replied to (2699)
4426 - client assert failure during cleanup after abort recovery (2701)
4427 - leak mdc device after failed mount (2712)
4428 - ptlrpc_check_set allows timedout requests to complete (2714)
4429 - wait for inflight reqs when ptlrpcd finishes (2710)
4430 - make sure unregistered services are removed from the srv_list
4431 - reset bulk XID's when resending them (caught by 1138 test)
4432 - unregister_bulk after timeout
4433 - fix lconf error (2694)
4434 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4435 - readahead locks pages, leaves pending causing memory pressure (2673)
4436 - increase OST request buffers to 4096 on large machines (2729)
4437 - fix up permission of existing directories in simple_mkdir (2661)
4438 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4439 - don't assign transno to errored transactions (2742)
4440 - don't delete objects on OST if given a bogus objid from MDS (2751)
4441 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4442 - if rq_no_resend, then timeout request after recovery (2432)
4443 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4444 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4445 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4446 - MDS recovery shouldn't delete objects that it has given out (2730)
4447 - if enqueue arrives after completion, don't clobber LVB (2819)
4448 - don't unlock pages twice when trigger_group_io returns error (2814)
4449 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4450 - don't write pages to disk if there was an error (1450)
4451 - don't ping imports that have recovery disabled (2676)
4452 - take buffered bytes into account when balancing socknal conn (2817)
4453 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4454 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4455 - remove little-endian swabbing of llog records (1987)
4456 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4457 - retry reposting request buffers if they fail (1191)
4458 - grow extent at grant time to avoid granting a revoked lock (2809)
4459 - lock revoke doesn't evict page if covered by a second lock (2765)
4460 - disable VM readahead to avoid reading outside lock extents (2805)
4462 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4463 - updated kernel patches for hp-2.4.20 kernel (2681)
4465 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4468 - fix truncated write corruption (2366)
4469 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4471 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4472 - ensure len(uuid) < 37 in lmc (1171)
4473 - fix ia64 OOPS in llog_test (2255)
4474 - zero end of page at obdfilter for partial page writes (2648)
4475 - don't leave stale dentries around after renames (bug 2428)
4476 - fix timeouts when evicting a client with a single lock held (2642)
4477 - set deadline for the initial HELLO message to drain (2634)
4478 - print out dotted-quad IP addresses in the socknal (2302)
4480 - additional debugging for MDS client eviction problem (2443)
4481 - fix mkfsoptions support for osts (2603, 2604)
4483 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4486 - add series for the vanilla 2.6.0 kernel
4487 - add series for the vanilla 2.4.24 kernel
4488 - add series for a cray x86/64 UL kernel drop
4489 - fix xattr patches for the vanilla 2.4.19 series
4491 - generate true UUIDs in lmc (1171)
4492 - have portals stack dumping break in UML (2466)
4493 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4494 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4495 - wake callback waiting threads on client eviction (2460)
4496 - Add --ptldebug and --subsystem to lmc (1719)
4497 - update assertion to allow safe interrupt allocation
4498 - set rq_no_resend for cancel requests (2432)
4499 - recalculate ptlrpcd timeout after resend (2494)
4500 - call vfs_rmdir when removing pending directories (2368)
4501 - fix renaming a file to itself (2429)
4502 - lmc creates a default one-stripe lov (2454)
4503 - expand procfs space to handle large clusters (2326)
4504 - increase UML stack to avoid overflow
4505 - update lconf's list of debug and subsystem masks
4506 - fix lfs find --obd (2510)
4507 - /proc tunable for disabling filter read caching (2591)
4508 - stop rpm packages from altering slapd.conf (2301)
4509 - disable nagle in the socknal under 0conf (2578)
4510 - choose mds inode size based on stripe count (2572)
4511 - fix kernel-source rpm problems (2516)
4513 - add --disable-doc to avoid pdf generation (2421)
4514 - update documentation, tests, type-os, comments
4515 - avoid format warnings on ia64
4516 - remove the TOE NAL
4517 - tiny code cleanups by removing unused fields
4519 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4522 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4523 - remove the most blatant lies from BUILDING file (2371)
4524 - change default debug level to reasonable production setting
4525 - reduce client side cache size to reduce cache flush time
4526 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4527 - make TCP zerocopy and pinger support enabled by default (2476)
4528 - sync writes completed after process exits caused crashes (2319)
4529 - maintain correct mount count on the MDS (2356)
4530 - backout 1557, because 2316 wasn't really fixed
4531 - better file I/O statistics gathering in /proc
4532 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4533 - another kernel patch to fix zero-copy TCP function export
4534 - don't take duplicate lock when processing re-sent getattr (2420)
4535 - lctl uses obd_self_export instead of creating new conn (2353)
4536 - MDS/OST recovery case which requires object creation asserted (2425)
4537 - move lfs from /usr/sbin to /usr/bin in packages
4538 - fix race between mds_client_add and mds_client_free (2417)
4539 - use kmalloc instead of slabs in portals (2430)
4540 - don't create duplicate records when a failover MDS is present (2442)
4541 - remove unnecessary mount age check (2332)
4542 - don't remove directory inodes from locks prematurely (2451)
4543 - don't break if MDS service name is the same as hostname (2103)
4544 - fix races in client write RPC generation when cache full (2482)
4546 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4549 - remove now-unused request->rq_obd (278)
4550 - if an allocation fails, print out how much memory we've used (1933)
4551 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4552 - add the "configurable stack size" patch to most series files (1256)
4553 - ability to write large log records, for 100+ OST configs (2306)
4554 - fix NULL deref when filter_prep fails (2314)
4555 - fix operator precedence error in filter_sync
4556 - dynamic allocation of socknal TX descriptors (2315)
4557 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4558 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4559 - ensure that CWARN messages in Portals always get to the syslog
4560 - __init/__exit are not for prototype decls (ldlm_init/exit)
4561 - x86-64 compile warning fixes
4562 - fix gateway LMC keyword conflict (2318)
4563 - fix MDS lock inversions in getattr/reint paths (1844)
4564 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4565 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4566 - DLM race condition prevented some lock evictions (2328)
4567 - ENOMEM detection and retry on socknal sends (2230)
4568 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4569 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4570 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4571 - better error messages when a client is rejected during recovery (1505)
4572 - avoid cancelling locks which were never granted, after failure (2330)
4573 - fix i_sem/journal inversion in mds_client_add (2333)
4574 - fix truncate/getattr lock cycle deadlock (2334)
4575 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4576 - fix two rare exit paths which could leak an l_lock() ref (2321)
4577 - fencepost error in MDS/OST orphan recovery (2226)
4578 - make log record alignment 8 bytes (1988)
4579 - lstripe now fails when requested offset > ost_count (2237)
4580 - ensure that all kernel series have a complete list.h (1607)
4581 - fix crashes in special-file operations (2316)
4582 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4584 - allow configurable automake binary, for testing new versions
4585 - small update to the lfs documentation
4587 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4589 * fix negative export reference count in fsfilt_sync (2312)
4591 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4592 * release candidate 0.9.1
4594 - orphans are moved into the PENDING directory for possible recovery
4595 - replayed opens now open by fid for orphan/rename safety (1042)
4596 - last close of an orphan inode generates a transno (683)
4597 - chdir() and mount() now pin the directory entry (1020)
4598 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4599 - discard very old requests without processing them (1502)
4600 - remove some common, well-understood CERRORs (1505)
4601 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4602 - clear "grant" flags in OST replies until OST grant code lands (1644)
4603 - fix read performance by not clobbering i_blksize on client (1598)
4604 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4605 - make LRU size a /proc tunable, clears locks when reduced (707)
4606 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4607 - server locks take references on exports now (1558)
4608 - build fixes for 2.4.20-rh trees (1663)
4609 - return an error from lov_create if all OSCs are inactive (1751)
4610 - fix import levels when a reconnect happens without a timeout (1597)
4611 - exit early from mds_open if we get a lookup error (1749)
4612 - partial page read at EOF wouldn't wait for disk before sending (1642)
4613 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4614 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4615 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4616 - write/truncate lock inversion (1639)
4617 - Don't auto-load obdclass, portals modules during cleanup (1495)
4618 - fix timestamps from jumping to "now" (1763)
4619 - extra journal assertions (1648)
4620 - add an extra multiunlink test (1771)
4621 - fix read_record/write_record API (1776)
4622 - fix leak of offset_extent, possible incorrect i_size later (1772)
4623 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4624 - Debugging for the kqswnal_get_idle_tx problems (1820)
4625 - Allow recovery to be attempted multiple times (1536)
4626 - Write out MDS last_rcvd file after it is first created (1600)
4627 - Fix tx_descriptor leak in failed transmit situations (1827)
4628 - ext3 journaling fixes for assertion failure after IO error (1871)
4629 - class_export_put() on freed export after completion AST error (1896)
4630 - Fix revalidate looping in VFS (1322)
4631 - Don't access a freed export during MDS_REINT timeout (1521)
4632 - Add open-unlink recovery support on the MDS (1673,1764)
4633 - Return an error if no MDS data was read from last_rcvd (1946)
4634 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4635 - Don't setup a disk device that doesn't match exported UUID (317)
4636 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4637 - avoid committing NULL handle in force close
4638 - local.sh is now a one-stripe LOV configuration
4639 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4640 - fix ext3 htree duplicate directory entry corruption (1516)
4641 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4642 - update to new LOV EA format (2097)
4643 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4645 2003-06-15 Phil Schwan <phil@clusterfs.com>
4648 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4649 - per-import/export recovery handling (958, 931, 959)
4650 - multiple last-rcvd slots, for serving multiple FSes (949)
4651 - connections are again shared between multiple imp/exports (963, 964)
4652 - "umount -f" would hang if any requests needed to be sent (393, 978)
4653 - avoid pinning large req buffer by copying for queued messages (989)
4654 - add "uuid" to "lctl device" command to help upcalls (991)
4655 - "open" RPCs with transnos would confuse recovery counters (1037)
4656 - do proper endian conversion of all wire messages (288, 340, 891)
4657 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4658 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4659 - fix ptlrpc_connection leak in target_handle_connect (1174)
4660 - fix import refcounting bug in OST and MDS cleanup (1134)
4661 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4662 - fix very unlikely obd_types race condition (501)
4663 - remove osc_open hack for echo_client (1187)
4664 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4665 - a failure in read_inode2 leads to deadlock (1139)
4666 - cancel ack-locks as soon as transaction is committed (1072)
4667 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4668 - make sure to commitrw after any preprw to avoid deadlock (1162)
4669 - failing to execute a file in a lustre FS would lock inode (1203)
4670 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4671 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4672 - obd_uuid2tgt would walk off the end of the list (1255)
4673 - on IA64 the osc would give portals incorrect bulk size (1258)
4674 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4675 - fix lock inversion caused by new llite matching code (1282)
4676 - limit the number of dirty pages on a client to 10MB (1286)
4677 - timed out locks were not being corrected cancelled (1289)
4678 - fix O_DIRECT above 4GB on IA-32 (1292)
4679 * major user-visible changes
4680 - fail out/fail over policy now controlled by the upcall (993)
4682 - add OBD_PING to check server availability and failure (954)
4683 - lustre messages are now sent in sending host order (288, 340, 891)
4684 - add eadatalen to MDS getattr reply (340)
4685 - OST read replies may contain second buffer, with per-page status (593)
4687 2003-03-11 Phil Schwan <phil@clusterfs.com>
4690 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4691 - failed open()s could cause deadlock; fixed (867, 869)
4692 - stop cancelling OST locks when files are closed (481)
4693 - overlapping XID spaces caused network corruption (851, 853)
4694 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4695 - setattr_raw added, to do single-RPC, server-side setattrs
4696 - lmc/lconf syntax change for OST UUIDs
4697 - fix crashy race condition between ptlrpc_free_req and osc_close
4698 - don't use request in mdc_enqueue if we hit a timeout (889)
4699 - don't set the inode i_size for regular files from the MDS (896)
4700 - handle out of order completion AST (842)
4701 - don't LBUG if a lock request times out after receiving AST (913)
4702 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4703 - if a bad lock AST arrives, send an error instead of dropping entirely
4704 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4705 - fix leak in bulk IO when only partially completed (899, 900, 926)
4706 - fix O_DIRECT for ia64 (55)
4707 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4708 - C-z after timeout could hang a process forever; fixed (977)
4710 - client-side I/O cache (678, 924, 929, 941, 970)
4712 - READPAGE and SETATTRs which don't take server-side locks get
4715 2003-02-11 Phil Schwan <phil@clusterfs.com>
4718 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4719 - Fix fsfilt_extN_readpage() to read a full page of directory
4720 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4721 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4722 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4723 - Most intent operations were removed, in favour of a new RPC mode
4724 that does a single RPC to the server and bypasses most of the VFS
4725 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4726 - Aggressively cancel local locks on DLM servers
4727 - mds_reint_unlink sends EA to the client if it's the last nlink.
4728 client uses that EA to unlink OST objects.
4729 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4730 - recursive symlinks were fixed (439)
4731 - fixed NULL deref in DEBUG_REQ
4732 - filter_update_lastobjid no longer calls sync, which annoyed extN
4733 - fixed multi-client small-writes to a single file problem (445)
4734 - fixed mtime updates during file writes (607)
4735 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4736 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4737 - fixed Linux OST size reporting problem (444, 656)
4738 - OST now updates object mtime with writes or setattr (607, 619)
4739 - client verifies file size before zeroing page past EOF (445)
4740 - OST now writes last allocated objid to disk with allocation (108)
4741 - LOV on echo now works (409)
4743 - mds_reint_unlink sends a new buffer, with the EA included. this
4744 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4745 set if a regular file was being unlinked, and it was the last link
4746 - use PtlGet from the target for bulk writes (315)
4747 - OST now updates object mtime with writes or setattr (607, 619)
4748 - LDLM now has a grant-time callback to revalidate locked items, if
4750 - Many MDS operations were reorganized to combat race conditions
4752 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4753 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4755 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4758 - Fully reactivate OST imports after reconnection (512, others)
4759 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4760 - More graceful error handling for truncating on dead OST (515)
4761 - Don't error out unless we're actually accessing dead stripes (474)
4762 - Fix garbage sizes when stripes are missing (410)
4763 - LRU counters were broken, causing constant lock purge (433, 432)
4764 - garbage on read from stripes with failed OSTs (441)
4765 - mark OSCs as active before reconnecting during recovery (438)
4766 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4767 - lfind did not preserve OST order in output (443)
4768 - symlinks cause hung clients, incorrect data (439)
4769 - stop dereferencing request after dropping refcount (457)
4770 - don't LASSERT(spin_is_locked) on non-SMP (455)
4771 - fixes for many rename() bugs
4772 - fstat didn't correctly synchronize attributes (399)
4773 - server must handle lock cancellation during blocking AST prep (487)
4774 - bulk descriptors were free()d too soon (511)
4775 - fix paths in lconf, which would load incorrect modules (451, 507)
4776 - fix confusing lconf 'host not found' error message (386)
4777 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4778 - fix race condition in mdc_blocking_ast() for inode access (526)
4779 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4780 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4781 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4782 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4783 - fix lstripe in several ways (kernel side) (527)
4784 - fix request leak in ldlm_cli_enqueue (262)
4785 - incorrect OSC was marked inactive after OST failure
4786 - call mds_fs_cleanup before unmounting filesystem (524)
4787 - fix races between taking ns_lock and ldlm_lock_change_resource
4788 - fix races updating LOV export open file list
4789 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4790 - fix recovery NULL deref in ldlm_cli_cancel_unused
4791 - fix some DLM races by using new hash table for lock handles (419)
4792 - permit the client to specify desired inodes, at replay
4793 - duplicate requests when we queue them for replay reintegration
4794 - fix last_rcvd offset calculation
4795 - sync after each recovered transaction, so we always make progress
4796 - never, not always, ERESTART requests without transnos
4797 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4799 - skip replay if the MDS says that the client is already connected
4800 - don't check for a recovery-enabled export to match lctl's UUID
4801 - don't INC_USE_COUNT for phantom exports
4802 - don't crash when cleaning up phantom exports (567)
4803 - don't double-finish or set replay data for errored mdc_open requests
4804 - abort requests when they time out, so we don't get old replies
4805 - send/receive replies for AST messages again
4806 - if the client says that it doesn't have the lock, cancel it on the
4808 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4809 mark it as destroyed, it will all work out when decref is called
4810 - fix module use counts (22, 581)
4812 - ASTs now expect a reply (server cancels lock on error reply)
4814 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4817 - fix many simultaneous client startup (392)
4818 - fix dentry->d_it clobbering
4819 - credentials weren't being shipped for readdir/getattr operations
4820 - remove invalid assertions triggered during some concurrent MD
4822 - proper Lustre versions added (336, 389)
4823 - fix memory leak for create error case (398)
4824 - fix LOV locking bug that would get cli/srv out of sync
4825 - fix echo client over LOV (409)
4826 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4827 - fix double-O_EXCL intent crash (424)
4828 - avoid sending multiple lock CANCELs (352)
4830 - MDS can do multi-client recovery (modulo bugs in new code)
4832 - many updates, edits, cleanups
4834 2002-11-18 Phil Schwan <phil@clusterfs.com>
4837 - fix null d_it dereference (346)
4838 - fix full OST/dbench hang (333)
4839 - fix permission problem with file removal (286)
4840 - fix removal of OSCs from LOV when they fail
4841 - fix NULL deref during bulk timeout (214)
4842 - fix problems related to multiple filesystems on one MDS (241)
4843 - fixed serious subtle metadata locking bugs
4844 - free locks on clients when inodes are removed due to memory
4846 - fix inode pointer in lock data (285)
4847 - partial support for multiple MDS on a single host (241)
4848 - data locks weren't cancelled at clear_inode time (290, 311)
4849 - intent locks could lead to unbounded lock growth (205)
4850 - added a maximum lock count, an LRU list, and a flusher
4851 - fix multiple rename (365)
4852 - properly abstracted the echo client
4853 - OSC locked 1 byte too many; fixed
4854 - rewrote brw callback code:
4855 - fixed recovery bugs related to LOVs (306)
4856 - fixed too-many-pages-in-one-write crash (191)
4857 - fixed (again) crash in sync_io_timeout (214)
4858 - probably fixed callback-related race (385)
4860 - Add capability to MDS protocol
4861 - LDLM cancellations and callbacks on different portals
4863 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4866 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4867 - defer open object destruction to close time (601981, 138)
4868 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4869 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4870 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4871 - don't call obd_setattr in ll_file_release for destroyed objects
4872 * protocol change to lustre_msg: move |version| and add |flags|
4873 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4874 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4875 * added replay of create, unlink, link and rename operations during
4876 MDS failover; recovery should be much more robust now
4877 * remove failed OSCs from LOVs (only lov_create uses this so far)
4878 * the lustre-HOWTO was brought (more) up to date (582544)
4880 2002-10-23 Phil Schwan <phil@clusterfs.com>
4883 - in-use dentries weren't being reused properly (617851)
4884 - prevent multiple LDLM setup (599178)
4885 - fix LOV size calculations for truncate (617853)
4886 - fix client handling of MDS intent errors (POSIX)
4887 - fix permission bug in lovstripe.c test (624321)
4888 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4889 - truncate past end of file could corrupt data
4890 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4891 - a race in recovery could return ETIMEDOUT to apps (623947)
4892 - building outside the source directory was fixed
4893 * the lustre-HOWTO was brought (more) up to date (582544)
4894 * major progress was made on recovery functionality
4896 2002-10-10 Phil Schwan <phil@clusterfs.com>
4899 - recovery deadlock fix
4900 - rm -rf causes LBUG fix (617817)
4901 - file open by multiple tasks fix (618962)
4902 - directory permissions bugs (602707 and 620007)
4903 - journal_stop fixed with locking (611313)
4904 - O_APPEND failures resolved (618273, perhaps 614459)
4905 - lconf PATH fix (619770)
4906 - IA64 build fix (621450)
4907 - RPC buffer sizes scale with amount of memory
4909 2002-10-01 Phil Schwan <phil@clusterfs.com>
4912 - locks would be cancelled without throwing away data pages,
4913 resulting in inconsistent data (605627)
4914 - inode attributes were not always being refreshed (605627, 612449)
4915 - lconf now continues to cleanup after lctl reports an error
4916 - MDS now enforces user permissions (602707)
4917 - lprocfs cleanup fixed, but not yet enabled (614157)
4918 - fixed infinite server hang, should a client not respond to an AST
4919 - avoid going into recovery if user calls readlink() with a buffer
4920 that's too small (613941)
4921 - AST RPCs no longer require replies (614867) -- this may be changed
4922 - don't crash server if client sends an IOV that's too big (611336)
4923 - fixed lock conversion deadlock (611892)
4924 - fixed the following of symlinks (614622)
4925 * recovery: the server can remove locks from a client that dies, other
4926 clients can make progress
4927 * more extN patch fixes
4928 * compile-time configurable ptlrpc buffer allocations
4930 - collaborative read cache document
4931 - Lustre Lite Performance CDR document-in-progress
4933 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4936 - fix typo in patch-2.4.18
4938 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4941 - clear ptlrpc request each time in handle_incoming_request()
4942 - unlink of files now destroys the object on the OST
4944 2002-09-19 Peter Braam <braam@clusterfs.com>
4946 * add hard link support
4947 * change obdfile creation method
4948 * kernel patch changed
4950 2002-09-19 Peter Braam <braam@clusterfs.com>
4953 - stack overflow bug in extN fixed
4955 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4957 * documentation updates
4958 - add man pages for config tools
4959 - update tests/README to describe testing with new config tools
4960 - finish metadata API descriptions
4961 * bug fixes and cleanups
4962 - statfs workaround for 16TB limit
4963 - LOV stripe allocation improved, can stripe on subset of OSTs
4964 - LOV file size/IO offset was wrong for files > 4GB in size
4965 - object EA data was being dropped, caused files to be unreadable
4966 - memory overflow with non-LOV OST caused memory corruption
4967 - fixed regression tests to work with new config tools, obdfilter
4968 - fixed bug when directory size became larger than 1 block
4969 - fixed bug (for single client case) when PWD was deleted
4970 - invalidate local directory pages when doing intent-based ops
4971 - avoid LDLM oops when lock callback contained bad data
4973 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4975 * documentation updates
4976 * bug fixes and cleanups
4977 - configuration tools
4980 - 64-bit compile warnings
4981 - 64-bit internal statfs data
4983 * test_brw on persistent OST devices
4985 * lprocfs (disabled)
4987 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4989 * documentation updates
4990 * bug fixes and cleanups
4991 * configuration tools
4993 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4996 * many small fixes to 0_5_4
4997 * io/network handling
4998 * thinkos in MDS operations
5000 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5003 * crucial basic fixes to 0.5.3
5004 * IOR, Iozone work over Elan
5007 2002-08-07 Phil Schwan <phil@clusterfs.com>
5008 * version 0_5_3, our first alpha
5009 * we use the new Portals iovs
5010 * documentation updates
5011 * bug fixes and cleanups
5012 * small changes in the DLM wire protocol
5014 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5015 * version 0_5_1 with some initial stability,
5016 * locking on MD and file I/O.
5017 * documentation updates
5018 * several bug fixes since 0.5.0
5019 * small changes in wire protocol
5021 2002-07-18 Phil Schwan <phil@clusterfs.com>
5023 * delivered as Lustre Light Alpha
5024 * fixed a crash after handling invalid MDS requests
5025 * fixed directory pages for architectures with non-4k pages sizes
5027 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5028 * release version v0_4_4
5029 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5031 * Fixes a number of interruption problems with OST operations.
5032 * Update documentation for portals header changes
5033 * Move all wire protocol structs/defines to lustre_idl.h
5034 * Fixes symlink length bug.
5035 * Add tcpdump to repository.
5037 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5038 * release version v0_4_3
5039 * Fixes statfs for inodes on extN.
5040 * Fixes bug in runtests which would delete /etc/hosts.
5041 * Use 64-bit object IDs wherever possible (not into VFS though)
5042 Remove ost_get_info, which is unused by lustre, and out of date.
5044 2002-07-03 Peter Braam <braam@clusterfs.com>
5045 * release version v0_4_2 Fixes a lookup error (type not passed)
5046 * move forward to head of Portals
5047 * move forward to latest Lustre kernel
5049 2002-06-25 Peter Braam <braam@clusterfs.com>
5050 * release version v0_4_1. Hopefully stable on single node use.