1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.19.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-sun3
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
17 Description: oops in page fault handler
18 Details : kernel page fault handler can return two special 'pages' in error case, don't
19 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
23 Description: timeout with invalidate import.
24 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
25 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
26 ptlrpc_invalidate_import and as result LASSERT.
28 Severity : enhancement
30 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
34 Description: enable MGS and MDT services start separately
35 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
36 a co-located MGS without starting the MGS, which is a complement
37 to 'nosvc' mount option.
40 Frequency : always with o2ib 1.3 and sles10
42 Description: fix build with SLES10 and o2ib v3.
43 Details : sles10 uses diffrent name for Module.symver file but configure
44 assume this file has same name on RHEL/SLES/vanila kernels.
48 Frequency : after recovery
49 Description: precreate to many object's after del orphan.
50 Details : del orphan st in oscc last_id == next_id and this triger growing
51 count of precreated objects. Set flag LOW to skip increase count
52 of precreated objects.
56 Frequency : rare, on clear nid stats
57 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
58 Details : when clean nid stats sometimes try destroy live entry,
59 and this produce panic in free.
63 Description: Stack overflow during MDS log replay
64 ease stack pressure by using a thread dealing llog_process.
68 Description: wait until IO finished before start new when do lock cancel.
69 Details : VM protocol want old IO finished before start new, in this case
70 need wait until PG_writeback is cleared until check dirty flag and
71 call writepages in lock cancel callback.
73 Severity : enhancement
75 Description: using special macro for print time and cleanup in includes.
79 Description: mds_mfd_close() ASSERTION(rc == 0)
80 Details : In mds_mfd_close(), we need protect inode's writecount change
81 within its orphan write semaphore to prevent possible races.
85 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
89 Frequency : rare, on shutdown ost
90 Description: don't hit live lock with umount ost.
91 Details : shrink_dcache_parent can be in long loop with destroy dentries,
92 use shrink_dcache_sb instead.
96 Description: don't panic with use echo client
97 Details : echo client pass NULL as client nid pointer and this produce null
102 Description: fix build on ppc32
103 Details : compile code with -m64 flag produce wrong object file for ppc32.
107 Description: add message levels for liblustreapi
111 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
112 Details : Move the 'good_osts' check before the 'total_bavail' check. This
113 will result in an -EAGAIN and in the exit call path we call
114 alloc_rr() which will with increasing aggressiveness attempt to
115 aquire precreated objects on the minimum number of required OSCs.
119 Description: Use old size assignment to avoid deadlock
120 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
121 the scheduling while holding a spinlock. We do not need locking
122 for size in ll_update_inode() because size is only updated from
123 the MDS for directories or files without objects, so there is no
124 other place to do the update, and concurrent access to such inodes
125 are protected by the inode lock.
129 Description: resolve "_IOWR redefined" build error on SLES10
133 Description: dump the memory debugging after all modules are unloaded to
134 suppress false negative in conf_sanity test 39
136 Severity : enhancement
138 Description: build kernel-ib packages for OFED 1.3 in our release cycle
143 Description: fix SLES kernel versioning
144 Details : the kernel version for our SLES 10 kernel did not include a "-"
145 before the "smp" at the end. while this was not a problem in
146 general, it did mean that software trying to use the kernel
147 version to try to detect a vendor specific kernel would fail.
148 this was most evident by the OFED build scripts.
152 Description: Don't update lov_desc members until making sure they are valid
153 Details : When updating lov_desc members via proc fs, need fix their
154 validities before doing the real update.
158 Description: don't put request into delay list while invalidate in flight.
159 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
160 import in flight. this produce timeout for invalidate and sometimes
161 can cause stale data.
165 Frequency : on ppc only
166 Description: not convert ost objects for directory because it's not exist.
167 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
171 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
173 Severity : enhancement
175 Description: Update to support 2.6.22.14 vanilla kernel.
179 Frequency : rare, on recovery
180 Description: read procfs can produce deadlock in some situation
181 Details : Holding lprocfs lock which send rpc can produce block for destroy
182 obd objects and this also block reconnect with -EALREADY. This isn't
183 fix all lprocfs bugs - but make it rare.
185 Severity : enhancement
187 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
190 Frequency : frequent on X2 node
192 Description: mdc_set_open_replay_data LBUG
193 Details : Set replay data for requests that are eligible for replay.
197 Description: lustre_mgs: operation 101 on unconnected MGS
198 Details : When MGC is disconnected from MGS long enough, MGS will evict the
199 MGC, and late on MGC cannot successfully connect to MGS and a lot
200 of the error messages complaining that MGS is not connected.
204 Frequency : on network error
205 Description: panic with double free request if network error
206 Details : mdc_finish_enqueue is finish request if any network error ocuring,
207 but it's true only for synchronus enqueue, for async enqueue
208 (via ptlrpcd) this incorrect and ptlrpcd want finish request
211 Severity : enhancement
213 Description: client-side metadata stat-ahead during readdir(directory readahead)
214 Details : perform client-side metadata stat-ahead when the client detects
215 readdir and sequential stat of dir entries therein
218 Frequency : on start mds
220 Description: Implement get_info(last_id) in obdfilter.
223 Frequency : occasional
225 Description: Correctly check stale fid, not start epoch if ost not support SOM
226 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
227 overwrited with new generated fid, but mds can anwer with one of these
228 two fids and both is not stale. setattr incorectly start epoch and
229 assume will be called done_writeting, but without SOM done_writing
233 Frequency : rare, depends on device drivers and load
235 Description: MDS or OSS nodes crash due to stack overflow
236 Details : Code changes in 1.8.0 increased the stack usage of some functions.
237 In some cases, in conjunction with device drivers that use a lot
238 of stack the MDS (or possibly OSS) service threads could overflow
239 the stack. One change which was identified to consume additional
240 stack has been reworked to avoid the extra stack usage.
243 Frequency : occasional
245 Description: Do not fail import if osc_interpret_create gets -EAGAIN
246 Details : If osc_interpret_create got -EAGAIN it immediately exits and
247 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
248 oscc_has_objects and see OSC has no objests and call
249 oscc_internal_create to resend create request.
251 Severity : enhancement
253 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
255 Severity : enhancement
257 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
260 Frequency : very rare
262 Description: Fix replay if there is an un-replied request and open
263 Details : In some cases, older replay request will revert the
264 mcd->mcd_last_xid on MDS which is used to record the client's
267 Severity : enhancement
269 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
271 Severity : enhancement
273 Description: Add rhel5 support to HEAD.
275 Serverity : enhancement
277 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
282 Description: Don't allow skipping OSTs if index has been specified.
283 Details : Don't allow skipping OSTs if index has been specified, make locking
284 in internal create lots better.
288 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
289 Details : access to bitfield in structure is always rounded to long
290 and this produce problem with not atomic change any bit.
294 Description: Lustre make rpms failed.
295 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
296 builds Lustre rpms from tarball.
299 Frequency : rare on shutdown ost
301 Description: If llog cancel was not send before clean_exports phase, this can
302 produce deadlock in llog code.
303 Details : If llog thread has last reference to obd and call class_import_put
304 this produce deadlock because llog_cleanup_commit_master wait when
305 last llog_commit_thread exited, but this never success because was
306 called from llog_commit_thread.
310 Description: allow userland application know is lost one of stripes.
311 Details : fill lvb_blocks with error code on ost and return it to
312 application if error flag found.
316 Description: NULL lov_tgts causing MDS oops
317 Details : more safe checks for NULL lov_tgts for avoid oops.
319 Severity : enhancement
321 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
325 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
326 Details : If an OST is down the MDS will hang indefinitely in
327 obd_statfs_rqset() waiting for the statfs data. While for
328 MDS QOS usage of statfs, it should not stuck in waiting.
330 Severity : enhancement
332 Description: remote_acl support
333 Details : Support ACL-based permission check for remote user.
334 Support setfacl/getfacl for remote user with the utils
335 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
336 the system "{s,g}etfacl" utils.
338 Severity : enhancement
340 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
342 Severity : enhancement
344 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
348 Description: make mgs_setparam() handle fsname containing dash
349 Details : fsname containing a dash does not work with lctl conf_param
351 Severity : enhancement
353 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
355 Severity : enhancement
357 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
361 Frequency : rare, at shutdown
362 Description: access already free / zero obd_namespace.
363 Details : if client_disconnect_export was called without force flag set,
364 and exist connect request in flight, this can produce access to
365 NULL pointer (or already free pointer) when connect_interpret
366 store ocd flags in obd_namespace.
370 Frequency : only at startup
371 Description: not alloc memory with spinlock held.
372 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
373 if any spinlock held.
375 Severity : enhancement
377 Description: make lustre randomly fail allocating memory
378 Details : Make lustre randomly failed allocating memory for testing purpose.
380 Severity : enhancement
382 Description: lost problems with lov objid file
383 Details : Fixes some scability and access to not inited memory problems
384 in work with lov objdid file.
389 Description: lfs find does not continue on file error
390 Details : Continue other files processing when a file/dir is absent.
394 Description: Inconsistent usage of lustre_pack_reply()
395 Details : Standardize the usage of lustre_pack_reply() such that it
396 always generate a CERROR on failure.
399 Frequency : occasional
401 Description: MDS hang or stay in waiting lock
402 Details : If client receive lock with CBPENDING flag ldlm need send lock
403 cancel as separate rpc, to avoid situation when cancel request
404 can't processed due all i/o threads stay in wait lock.
408 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
412 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
416 Description: Update to SLES9 kernel-2.6.5-7.287.3.
420 Description: Update to RHEL4 latest kernel.
422 Severity : enhancement
424 Description: Build SLES10 patchless client fails
425 Details : The configure was broken by run ./configure with
426 --with-linux-obj=.... argument for patchless client. When the
427 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
428 can't search header adequately. Use absolute path such as
429 -I($LINUX)/include instead.
433 Description: interrupt oig_wait produce painc on resend.
434 Details : brw_redo_request can be used for resend requests from ptlrpcd and
435 private set, and this produce situation when rq_ptlrpcd_data not
436 copyed to new allocated request and triggered LBUG on assert
437 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
438 ptlrpcd set if request is changed and can be safety changed to use
443 Description: LASSERT_{REQ,REP}SWAB macros are buggy
444 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
445 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
446 these macros and replace them with their difinitions instead.
450 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
451 Details : Update which_patch & target file for SLES10 latest kernel.
454 Description: Remove client patches from SLES 10 kernel.
455 Details : This causes SLES 10 clients to behave as patchless clients
456 even on a Lustre-patched (server) kernel.
458 Severity : enhancement
460 Description: self-adjustable client's lru lists
461 Details : use adaptive algorithm for managing client cached locks lru
462 lists according to current server load, other client's work
463 pattern, memory activities, etc. Both, server and client
464 side namespaces provide number of proc tunables for controlling
467 Severity : enhancement
469 Description: light-weight GSS support
470 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
471 and reduce performance overhead.
473 Severity : enhancement
475 Description: Linux keyring support
476 Details : Support using service of Linux keyring for Lustre GSS internal
477 context refresh/cache mechanism.
481 Description: Fix errors in lfs documentation
482 Details : Fixes man pages
486 Description: don't use GFP_* in generic Lustre code.
487 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
491 Description: obdclass is limited by single OBD_ALLOC(idarray)
492 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
496 Description: warnings with build patchless client with vanila 2.6.19 and up
497 Details : change old ctl_table style and replace ctl_table/ctl_table_header
498 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
502 Frequency : Only for SLES
503 Description: sanity_quota fail test_1
504 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
505 file, which makes TSTID[2] inunique.
509 Frequency : Only for SLES9
510 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
511 Details : Update target/ChangeLog/which_patch .
515 Frequency : Only for test_10 in sanity-quota.sh
516 Description: a bug in quota test code
517 Details : checking a test flag in a wrong place
521 Frequency : Only for test_13 in sanity-quota.sh
522 Description: a bug in quota test script
523 Details : shouldn't check overfull allocated quota
527 Description: jbd statistics
528 Details : Port older jbd statistics patch for sles10
530 Severity : enhancement
532 Description: Add printing inode info into message about error in writepage.
537 Description: reply_lock_interpret crash due to race with it and lock cancel.
538 Details : Do not replay locks that are being cancelled. Do not reference
539 locks by their address during replay, just by their handle.
543 Frequency : When flocks are used.
544 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
545 Details : Flock locks might destroy just granted lock if it could be merged
546 with another existing flock, this is done in completion handler,
547 so teach ldlm_cli_enquque_fini that this is a valid case for
553 Description: Oops in read and write path when failing to allocate lock.
554 Details : Check if lock allocation failed and return error back.
558 Description: lstripe command fails for valid OST index
559 Details : The stripe offset is compared to lov->desc.ld_tgt_count
560 instead of lov->desc.ld_active_tgt_count.
564 Description: sanity.sh failed test 103
565 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
570 Description: df doesn't work properly if diskfs blocksize != 4K
571 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
575 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
576 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
580 Description: obd_health_check_timeout too short
581 Details : set obd_health_check_timeout as 1.5x of obd_timeout
585 Description: llapi_file_create() does not allow some changes
586 Details : add llapi_file_open() that allows specifying the mode and
587 open flags, and also returns an open file handle.
591 Description: merge and cleanup kernel patches.
592 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
596 Description: Add journal checksum support.(Kernel part)
597 Details : The journal checksum feature adds two new flags i.e
598 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
599 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
600 indicates that the commit block contains the checksum for
601 the blocks described by the descriptor blocks. Now commit
602 record can be sent to disk without waiting for descriptor
603 blocks to be written to disk. This behavior is controlled
604 using JBD2_FEATURE_ASYNC_COMMIT flag.
608 Description: OSS needs mutliple precreate threads
609 Details : Add ability to start more than one create thread per OSS.
613 Description: Sanity.sh test_65k failed: setstripe should have succeeded
614 Details : execute lfs setstripe on client
618 Description: mds_obd_create error creating tmp object
619 Details : When the user sets quota on root, llog will be affected and can't
620 create files and write files.
623 Frequency : Always on ia64 patchless client, and possibly others.
625 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
626 Details : This allows the patchless client to be loaded on architectures
631 Description: RedHat Update kernel for RHEL5
632 Details : Add the kernel config file for RHEL5.
636 Description: RedHat Update kernel for RHEL5
637 Details : Modify the kernel config file more closer RHEL5.
641 Description: Build failure against Centos5 (RHEL5)
642 Details : Define PAGE_SIZE when it isn't present.
646 Description: test_7 on sanity-quota.sh can't be used on separate machines
647 Details : add facet to handle it
651 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
652 Details : with NFS, the anon dentry's parent was set to itself in
653 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
654 corresponding dentry other than use rec->ur_name.
656 Severity : enhancement
658 Description: lfs setstripe enhancement
659 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
663 Description: enable data checksumming by default
664 Details : enable checksum by default, allow --disable-checksum
665 configure option and "-o nochecksum" mount option. Checksums
666 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
670 Description: lustre support for RHEL5
671 Details : Add support for RHEL5.
675 Description: Client eviction due to failover config
676 Details : after a connection loss, the lustre client should attempt to
677 reconnect to the last active server first before trying the
678 other potential connections.
682 Description: when mds and osts use different quota unit(32bit and 64bit),
683 quota will be released repeatly.
684 Details : void sending multiple quota reqs to mds, which will keep the
685 status between the reqs.
689 Description: osts not allocated evenly to files
690 Details : change the condition to increase offset_idx
693 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
695 Description: Data corruption for OSTs that are formatted larger than 2TB
697 Details : When generating the bio request for lustre file writes the
698 sector number would overflow a temporary variable before being
699 used for the IO. The data reads correctly from Lustre (which
700 will overflow in a similar manner) but other file data or
701 filesystem metadata may be corrupted in some cases.
705 Description: Tune the kernel for good SCSI performance.
706 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
707 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
712 Description: rewrite ext2-derived code in obdclass/uuid.c
713 Details : rewrite inherited code (uuid parsing code from ext2 utils)
714 from scratch preserving functionality.
718 Description: "lfs find -obd UUID" prints directories
719 Details : "lfs find -obd UUID" will return all directory names instead
720 of just file names. It is incorrect because the directories
721 do not reside on the OSTs.
725 Description: use i_size_read and i_size_write in 2.6 port
726 Details : replace inode->i_size access with i_size_read/write()
729 Frequency : when removing large files
731 Description: scheduling issue during removal of large Lustre files
732 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
733 It causes scheduling issues when removing large files (17TB in the
737 Frequency : only with liblustre clients on XT3
739 Description: evictions taking too long
740 Details : allow llrd to evict clients directly on OSTs
743 Frequency : only on ppc
745 Description: /proc/fs/lustre/devices broken on ppc
746 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
747 We should make sure the type of 'index' is loff_t and then cast
748 explicitly as needed below. Do not assign an explicitly cast
752 Frequency : only for rhel5
754 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
755 Details : Modify the target file & which_kernel.
757 Severity : enhancement
759 Description: omit set fsid for export NFS
760 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
761 as FS_REQUIRES_DEV for avoid problems with generate fsid.
765 Frequency : Always, for kernels after 2.6.16
766 Description: Fix warning idr_remove called for id=.. which is not allocated.
767 Details : Last kernels save old s_dev before kill super and not allow
768 to restore from callback - restore it before call kill_anon_super.
772 Description: buffer overruns could theoretically occur
773 Details : llapi_semantic_traverse() modifies the "path" argument by
774 appending values to the end of the origin string, and a
775 overrun may occur. Adding buffer overrun check in liblustreapi.
779 Description: Fix error on 'ls .' at the top of the Lustre mount.
780 Details : Don't revalidate dentry if it is a root dentry.
784 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
785 Details : Modify vm-tunables-rhel4.patch.
789 Description: Kernel config for 2.6.18-vanilla.
790 Details : Modify targets/2.6-vanilla.target.in.
791 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
792 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
793 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
794 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
798 Description: improve handling recoverable errors
799 Details : if request processig with error which can be recoverable on server
800 request should be resend, otherwise page released from cache and
805 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
806 Details : Modify target file & which_patch.
807 A flaw was found in the IA32 system call emulation provided
808 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
809 value could be stored in the %RAX register, which could trigger an
810 out-of-bounds system call table access. An untrusted local user
811 could exploit this flaw to run code in the kernel
812 (ie a root privilege escalation). (CVE-2007-4573).
816 Description: O_DIRECT bypasses client statistics.
817 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
818 read_ahead_stats were updated. copied the stats section from
819 osc_send_oap_rpc() into async_internal().
823 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
827 Description: change order of libsysio includes
828 Details : '#include sysio.h' should always come before '#include xtio.h'
832 Description: handle "serious error: objid * already exists" more gracefully
833 Details : If LAST_ID value on disk is smaller than the objects existing in
834 the O/0/d* directories, it indicates disk corruption and causes an
835 LBUG(). If the object is 0-length, then we should use the existing
836 object. This will help to avoid a full fsck in most cases.
838 Severity : enhancement
840 Description: adapt the lustre_config script to support the upgrade case
841 Details : Add "-u" option for lustre_config script to support upgrading 1.4
842 server targets to 1.6 in parallel.
846 Description: To avoid grant space > avaible space when the disk is almost
847 full. Without this patch you might see the error "grant XXXX >
848 available" or some LBUG about grant, when the disk is almost
850 Details : In filter_check_grant, for non_grant cache write, we should
851 check the left space by if (*left > ungranted + bytes), instead
852 of (*left > ungranted), because only we are sure the left space
853 is enough for another "bytes", then the ungrant space should be
854 increase. In client, we should update cl_avail_grant only there
855 is OBD_MD_FLGRANT in the reply.
859 Description: Update RHEL 4 kernel to fix local root privilege escalation.
860 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
861 described in CVE-2007-4573. This problem could allow untrusted
862 local users to gain root access.
865 Frequency : when using O_DIRECT and quotas
867 Description: Incorrect file ownership on O_DIRECT output files
868 Details : block usage reported by 'lfs quota' does not take into account
869 files that have been written with O_DIRECT.
874 Description: touch file failed when fs is not full
875 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
876 otherwise we can get ENOSP while fs is not full.
880 Description: parallel lock callbacks
881 Details : Instead of sending blocking and completion callbacks as separated
882 requests, adding them to a set and sending in parallel.
885 Frequency : only for Cray XT3
886 Bugzilla : 12829/13455
887 Description: Changing primary group doesn't change the group lustre assigns to
889 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
890 overridden with the primary group provided by the group upcall,
891 whereas the supplied fsgid can be trusted if it is in the list of
892 supplementary groups returned by the group upcall.
894 Severity : enhancement
896 Description: Allow masking D_WARNING, D_ERROR messages from console
897 Details : Console messages can now be disabled via lnet.printk.
901 Description: User code with malformed file open parameter crashes client node
902 Details : Before packing join_file req, all the related reference should be
903 checked carefully in case some malformed flags cause fake
904 join_file req on client.
908 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
909 Details : ldlm_completion_ast() assumes that a lock is granted when the req
910 mode is equal to the granted mode. However, it should also check
911 that LDLM_FL_CP_REQD is not set.
915 Description: Heavy nfs access might result in deadlocks
916 Details : After ELC code landed, it is now improper to enqueue any mds
917 locks under och_sem, because enqueue might want to decide to
918 cancel open locks for same inode we are holding och_sem for.
922 Description: Client eviction while running blogbench
923 Details : A lot of unlink operations with concurrent I/O can lead to a
924 deadlock causing evictions. To address the problem, the number of
925 oustanding OST_DESTROY requests is now throttled to
926 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
927 callbacks are processed in priority.
931 Description: enable ACLs on MDS by default
932 Details : ACLs must be enabled on MDS by default.
935 Frequency : PPC/PPC64 only
937 Description: conflicts between asm-ppc64/types.h and lustre_types.h
938 Details : fix duplicated definitions between asm-ppc64/types.h and
939 lustre_types.h on PPC.
942 Frequency : PPC/PPC64 only
944 Description: asm-ppc/segment.h does not exist
945 Details : fix compile issue on PPC.
949 Description: better handle error messages in extents code
952 Frequency : RHEL4 only
954 Description: mkfs is very slow on IA64/RHEL4
955 Details : A performance regression has been discovered in the MPT Fusion
956 driver between versions 3.02.73rh and 3.02.99.00rh. As a
957 consequence, we have downgraded the MPT Fusion driver in the RHEL4
958 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
960 Severity : enhancement
962 Description: SNMP support enhancement
963 Details : Adding total number of sampled request for an MDS node in snmp
966 Severity : enhancement
968 Description: Optimize ldlm waiting list processing for PR extent locks
969 Details : When processing waiting list for read extent lock and meeting read
970 lock that is same or wider to it that is not contended, skip
971 processing rest of the list and immediatelly return current
972 status of conflictness, since we are guaranteed there are no
973 conflicting locks in the rest of the list.
977 Description: Time out and refuse to reconnect
978 Details : When the failover node is the primary node, it is possible
979 to have two identical connections in imp_conn_list. We must
980 compare not conn's pointers but NIDs, otherwise we can defeat
981 connection throttling.
985 Description: port llog fixes from b1_6 into HEAD
986 Details : Port llog reference couting and some llog cleanups from b1_6
987 (bug 10800) into HEAD, for protect from panic and access to already
988 free llog structures.
992 Description: Detect stride IO mode in read-ahead
993 Details : When a client does stride read, read-ahead should detect that and
994 read-ahead pages according to the detected stride pattern.
998 Description: data checksumming impacts single node performance
999 Details : add support for several checksum algorithm. Currently, only CRC32
1000 and Adler-32 are supported. The checksum type can be changed on
1001 the fly via /proc/fs/lustre/osc/*/checksum_type.
1005 Description: use adler32 for page checksums
1006 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1011 Description: build for x2 fails
1012 Details : fix compile issue on Cray systems.
1016 Description: Properly match for duplicate locks
1017 Details : Due to different lock order from skiplists code, we need to
1018 traverse entire list for now
1021 Frequency : only on PPC/SLES10
1023 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1024 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1025 be defined. Add a hack in mkfs_lustre.c to work around this compile
1030 Description: LASSERT on MDS when client holding flock lock dies
1031 Details : ldlm pool logic depends on number of granted locks equal to
1032 number of released locks which is not true for flock locks, so
1033 just exclude such locks from consideration.
1037 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1038 Details : Use fsfilt_sync() for both the whole filesystem sync and
1039 individual file sync to eliminate dangerous inode locking
1040 with I_LOCK that can lead to a deadlock.
1044 Description: performance in 1.6.3
1045 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1046 sure that 1MB requests are not fragmented by the block layer.
1048 Severity : enhancement
1050 Description: organize the server-side client stats on per-nid basis
1051 Details : Change the structure of stats under obdfilter and mds to
1059 The "uuid"s file would list the uuids of _active_ exports.
1060 And the clear entry is to clear all stats and stale nids.
1062 Severity : enhancement
1064 Description: eliminate client locks in face of contention
1065 Details : file contention detection and lockless i/o implementation
1066 for contended files.
1070 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1073 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1075 Description: MDS may lose file striping (and hence file data) in some cases
1076 Details : If there are additional extended attributes stored on the MDS,
1077 in particular ACLs, SELinux, or user attributes (if user_xattr
1078 is specified for the client mount options) then there is a risk
1079 of attribute loss. Additionally, the Lustre file striping
1080 needs to be larger than default (e.g. striped over all OSTs),
1081 and an additional attribute must be stored initially in the
1082 inode and then increase in size enough to be moved to the
1083 external attribute block (e.g. ACL growing in size) for file
1088 Description: skiplist implementation simplification
1089 Details : skiplists are used to group compatible locks on granted list
1090 that was implemented as tracking first and last lock of each lock group
1091 the patch changes that to using doubly linked lists
1095 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1096 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1097 because LASSERT on both the data supplied by a client, and the data
1098 on disk is dangerous and incorrect.
1100 Severity : enhancement
1102 Description: Slow trucate/writes to huge files at high offsets.
1103 Details : Directly associate cached pages to lock that protect those pages,
1104 this allows us to quickly find what pages to write and remove
1105 once lock callback is received.
1107 --------------------------------------------------------------------------------
1109 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1111 * Support for kernels:
1112 2.6.5-7.283 (SLES 9),
1113 2.6.9-55.EL (RHEL 4),
1114 2.6.16.46-0.14 (SLES 10),
1115 2.6.18.8 vanilla (kernel.org)
1116 * Client support for unpatched kernels:
1117 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1118 2.6.16 - 2.6.22 vanilla (kernel.org)
1119 * Due to recently discovered recovery problems, we do not recommend
1120 using patchless RHEL 4 clients with this or any earlier release.
1121 * Recommended e2fsprogs version: 1.39.cfs8
1122 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1123 * Starting with this release, the ldiskfs backing filesystem required
1124 by Lustre is now in its own package, lustre-ldiskfs. This package
1125 should be installed. It is versioned separately from Lustre and
1126 may be released separately in future.
1130 Description: block reactivating mgc import until all deactivates complete
1131 Details : Fix race when failing back MDT/MGS to itself (testing)
1134 Frequency : at statup only
1136 Description: mds_lov_synchronize race leads to various problems
1137 Details : simultaneous MDT->OST connections at startup can cause the
1138 sync to abort, leaving the OSC in a bad state.
1140 Severity : enhancement
1142 Description: add optional extra BUILD_VERSION info
1143 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1144 to override the lustre version.
1147 Frequency : 2.6.18 servers only
1149 Description: ll_kern_mount() doesn't release the module reference
1150 Details : The ldiskfs module reference count never drops down to 0
1151 because ll_kern_mount() doesn't release the module reference.
1156 Description: server LBUG when using old ost_num_threads parameter
1157 Details : Accept the old ost_num_threads parameter but warn that it
1158 is deprecated, and fix an off-by-one error that caused an LBUG.
1163 Description: Transient SCSI error results in persistent IO issue
1164 Details : iobuf->dr_error is not reinitialized to 0 between two
1168 Frequency : sometimes when underlying device returns I/O errors
1170 Description: OSTs not going read-only during write failures
1171 Details : OSTs are not remounted read-only when the journal commit threads
1172 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1173 instead of the ext3 wrappers.
1177 Description: poor connect scaling with increasing client count
1178 Details : Don't run filter_grant_sanity_check for more than 100 exports
1179 to improve scaling for large numbers of clients.
1182 Frequency : SLES10 only
1184 Description: sanity-quota.sh quotacheck failed: rc = -22
1185 Details : Quotas cannot be enabled on SLES10.
1188 Frequency : liblustre clients only
1190 Description: getdirentries does not give error when run on compute nodes
1191 Details : getdirentries does not fail when the size specified as an argument
1192 is too small to contain at least one entry
1194 Severity : enhancement
1196 Description: Add LNET router traceability for debug purposes
1197 Details : If a checksum failure occurs with a router as part of the
1198 IO path, the NID of the last router that forwarded the bulk data
1199 is printed so it can be identified.
1204 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1205 Details : Due to a race condition, liblustre clients were occasionally
1206 evicted incorrectly.
1208 Severity : enhancement
1210 Description: lfs setstripe use optional parameters instead of postional
1213 Severity : enhancement
1215 Description: Nanosecond timestamp support for ldiskfs
1216 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1217 resolution timestamps. There is not yet support for this at
1218 the Lustre filesystem level.
1221 Frequency : during server recovery
1223 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1224 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1228 Frequency : nfs export on patchless client
1230 Description: connectathon hang when test nfs export over patchless client
1231 Details : Disconnected dentry cannot be found with lookup, so we do not need
1232 to unhash it or make it invalid
1235 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1236 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1237 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1242 Description: open req refcounting wrong on reconnect
1243 Details : If reconnect happened between getting open reply from server and
1244 call to mdc_set_replay_data in ll_file_open, we will schedule
1245 replay for unreferenced request that we are about to free.
1246 Subsequent close will crash in variety of ways.
1247 Check that request is still eligible for replay in
1248 mdc_set_replay_data().
1253 Description: disable writes to filesystem when reading health_check file
1254 Details : the default for reading the health_check proc file has changed
1255 to NOT do a journal transaction and write to disk, because this
1256 can cause reads of the /proc file to hang and block HA state
1257 checking on a healthy but otherwise heavily loaded system. It
1258 is possible to return to the previous behaviour during configure
1259 with --enable-health-write.
1261 Severity : enhancement
1263 Description: 64-bit inode version
1264 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1265 to the inode. This will be required for version-based recovery.
1270 Description: MDS fails to start if a duplicate client export is detected
1271 Details : in some rare cases it was possible for a client to connect to
1272 an MDS multiple times. Upon recovery the MDS would detect this
1273 and fail during startup. Handle this more gracefully.
1275 Severity : enhancement
1277 Description: Add -o localflock option to simulate old noflock
1279 Details : This will achieve local-only flock/fcntl locks
1285 Description: log_commit_thread vs filter_destroy race leads to crash
1286 Details : Take import reference before releasing llog record semaphore
1291 Description: Wrong request locking in request set processing
1292 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1293 imp_delayed_list, in this place should be used imp_lock.
1296 Frequency : when reconnection
1298 Description: Grant Leak when osc reconnect to OST
1299 Details : When osc reconnect ost, OST(filter) should check whether it
1300 should grant more space to client by comparing fed_grant and
1301 cl_avail_grant, and return the granted space to client instead
1302 of "new granted" space, because client will call osc_init_grant
1303 to update the client grant space info.
1306 Frequency : when client reconnect to OST
1308 Description: Grant Leak when osc do resend and replay bulk write
1309 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1310 bulk write request, because the grant info while be sync between
1311 OSC and OST when reconnect, and we should ignore the grant info
1312 these of resend/replay write req.
1317 Description: Grant space more than avaiable left space sometimes.
1318 Details : When then OST is about to be full, if two bulk writing from
1319 different clients came to OST. Accord the avaliable space of the
1320 OST, the first req should be permitted, and the second one
1321 should be denied by ENOSPC. But if the seconde arrived before
1322 the first one is commited. The OST might wrongly permit second
1323 writing, which will cause grant space > avaiable space.
1326 Frequency : when client is evicted
1328 Description: Grant might be wrongly erased when osc is evicted by OST
1329 Details : when the import is evicted by server, it will fork another
1330 thread ptlrpc_invalidate_import_thread to invalidate the
1331 import, where the grant will be set to 0. While the original
1332 thread will update the grant it got when connecting. So if
1333 the former happened latter, the grant will be wrongly errased
1334 because of this race.
1339 Description: Checking Stale with correct fid
1340 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1341 to check whether it is stale, because sometimes, we want the
1342 enqueue happened anyway, and op_data.fid2 will not be initialized.
1344 Severity : enhancement
1346 Description: update patchless client
1347 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1350 Frequency : only with 2.4 kernel
1352 Description: random memory corruption
1353 Details : size of struct ll_inode_info is to big for union inode.u and this
1354 can be cause of random memory corruption.
1359 Description: Memory leak in recovery
1360 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1361 It should also check obd_fail before fsfilt_start, otherwise if
1362 fsfilt_start return -EROFS,(failover mds during mds recovery).
1363 then the req will return with repmsg->transno = 0 and rc = EROFS.
1364 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1365 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1366 be freed no matter whether fsfilt_commit success or not.
1369 Frequency : only with huge count clients
1371 Description: Prevents from taking the superblock lock in llap_from_page for
1373 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1374 for llap_from_page call in ll_removepage prevents from taking the
1375 superblock lock for a soon died page.
1380 Description: Not check open intent error before release open handle
1381 Details : in some rare cases, the open intent error is not checked before
1382 release open handle, which may cause
1383 ASSERTION(open_req->rq_transno != 0), because it tries to release
1384 the failed open handle.
1389 Description: Set cat log bitmap only after create log success.
1390 Details : in some rare cases, the cat log bitmap is set too early. and it
1391 should be set only after create log success.
1395 Description: Accessing a block bevice can re-enable I/O when Lustre is
1396 tearing down a device.
1397 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1402 Description: service threads may hog cpus when there are a lot of requests
1403 Details : Insert cond_resched to give other threads a chance to use some CPU
1408 Description: the cat log was not initialized in recovery
1409 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1410 unlink log on mds will not be initialized until mds post
1411 recovery. And also in mds post recovery, the unlink log will
1412 initialization will be done asynchronausly, so there will be race
1413 between add unlink log and unlink log initialization.
1417 Description: brw_stats were being printed incorrectly
1418 Details : brw_stats were being printed as log2 but all of them were not
1419 recorded as log2. Also remove some code duplication arising from
1420 filter_tally_{read,write}.
1424 Frequency : rare, only in recovery.
1425 Description: ASSERTION(req->rq_type != LI_POISON) failed
1426 Details : imp_lock should be held while iterating over imp_sending_list for
1427 prevent destroy request after get timeout in ptlrpc_queue_wait.
1431 Description: replay-single.sh test 52 fails
1432 Details : A lock's skiplist need to be cleanup when it being unlinked
1433 from its resource list.
1437 Description: Short directio read returns full requested size rather than
1439 Details : Direct I/O operations should return actual amount of bytes
1440 transferred rather than requested size.
1442 Severity : enhancement
1444 Description: metadata RPC reduction (e.g. for rm performance)
1445 Details : decrease the amount of synchronous RPC between clients and servers
1446 by canceling conflicing lock before the operation on the client
1447 and packing thier handles into the main operation RPC to server.
1449 Severity : enhancement
1451 Description: Async OSC create to avoid the blocking unnecessarily.
1452 Details : If a OST has no remain object, system will block on the creating
1453 when need to create a new object on this OST. Now, ways use
1454 pre-created objects when available, instead of blocking on an
1455 empty osc while others are not empty. If we must block, we block
1456 for the shortest possible period of time.
1458 Severity : enhancement
1460 Description: refine locking for avoid write wrong info into lov_objid file
1461 Details : fix possible races with add new target and write/update data in
1464 --------------------------------------------------------------------------------
1466 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1471 Frequency : on some architectures
1473 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1474 Details : Uninitialized flags sometimes cause configuration commands to
1478 Frequency : patchless clients only
1480 Description: missing __iget() symbol export
1481 Details : The __iget() symbol export is missing. To avoid the need for
1482 this on patchless clients the deathrow inode reaper is turned
1483 off, and we depend on the VM to clean up old inodes. This
1484 dependency was during via the fix for bug 12181.
1486 --------------------------------------------------------------------------------
1488 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1490 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1491 INTEROPERATE with older versions automatically. In many cases a
1492 special upgrade step is needed. Please read the
1493 user documentation before upgrading any part of a live system.
1494 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1495 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1496 * WARNING: Lustre configuration and startup changes are required with
1497 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1503 Severity : enhancement
1505 Description: MountConf
1506 Details : Lustre configuration is now managed via mkfs and mount
1507 commands instead of lmc and lconf. New obd types (MGS, MGC)
1508 are added for dynamic configuration management. See
1509 https://mail.clusterfs.com/wikis/lustre/MountConf for
1512 Severity : enhancement
1514 Description: dynamic OST addition
1515 Details : OSTs can now be added to a live filesystem
1517 Severity : enhancement
1519 Description: startup order invariance
1520 Details : MDTs and OSTs can be started in any order. Clients only
1521 require the MDT to complete startup.
1523 Severity : enhancement
1525 Description: parallel, asynchronous orphan cleanup
1526 Details : orphan cleanup is now performed in separate threads for each
1527 OST, allowing parallel non-blocking operation.
1529 Severity : enhancement
1531 Description: optimized stripe assignment
1532 Details : stripe assignments are now made based on ost space available,
1533 ost previous usage, and OSS previous usage, in order to try
1534 to optimize storage space and networking resources.
1536 Severity : enhancement
1538 Description: Permanently set tunables
1539 Details : All writable /proc/fs/lustre tunables can now be permanently
1540 set on a per-server basis, at mkfs time or on a live system.
1542 Severity : enhancement
1544 Description: Lustre message v2
1545 Details : Add lustre message format v2.
1547 Severity : enhancement
1549 Description: client OST exclusion list
1550 Details : Clients can be started with a list of OSTs that should be
1551 declared "inactive" for known non-responsive OSTs.
1555 Description: SPEC SFS validation failure on NFS v2 over lustre.
1556 Details : Changes the blocksize for regular files to be 2x RPC size,
1557 and not depend on stripe size.
1559 Severity : enhancement
1561 Description: Multiple MD RPCs in flight.
1562 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1563 To avoid overly-overloading MDS, introduce a limit on number of
1564 MDS RPCs in flight for a single client and add /proc controls
1565 to adjust this limit.
1567 Severity : enhancement
1569 Description: client read/write statistics
1570 Details : Add client read/write call usage stats for performance
1571 analysis of user processes.
1572 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1573 file access. extents_stats shows chunk size distribution.
1574 extents_stats_per_process show chunk size distribution per
1577 Severity : enhancement
1579 Description: mds statistics
1580 Details : Add detailed mds operations statistics in
1581 /proc/fs/lustre/mds/*/stats.
1585 Description: Failure of copying files with lustre special EAs.
1586 Details : Client side always return success for setxattr call for lustre
1587 special xattr (currently only "trusted.lov").
1592 Description: Refcount LNET uuids
1593 Details : The global LNET uuid list grew linearly with every startup;
1594 refcount repeated list entries instead of always adding to
1597 Severity : enhancement
1599 Description: Dynamic service threads
1600 Details : Within a small range, start extra service threads
1601 automatically when the request queue builds up.
1604 Frequency : mixed-endian client/server environments
1606 Description: mixed-endian crashes
1607 Details : The new msg_v2 system had some failures in mixed-endian
1610 Severity : enhancement
1612 Description: Easy OST removal
1613 Details : OSTs can be permanently deactivated with e.g. 'lctl
1614 conf_param lustre-OST0001.osc.active=0'
1616 Severity : enhancement
1618 Description: MGS proc entries
1619 Details : Added basic proc entries for the MGS showing what filesystems
1622 Severity : enhancement
1624 Description: provide MGS failover
1625 Details : Added config lock reacquisition after MGS server failover.
1627 Severity : enhancement
1629 Description: add Linux 2.4 support
1630 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1634 Description: a large application tries to do I/O to the same resource and dies
1635 in the middle of it.
1636 Details : Check the req->rq_arrival time after the call to
1637 ost_brw_lock_get(), but before we do anything about
1638 processing it & sending the BULK transfer request. This
1639 should help move old stale pending locks off the queue as
1640 quickly as obd_timeout.
1643 Frequency : when an incorrect nid is specified during startup
1645 Description: ptlrpc connect to non-existant node causes kernel crash
1646 Details : LNET can't be re-entered from an event callback, which
1647 happened when we expire a message after the export has been
1648 cleaned up. Instead, hand the zombie cleanup off to another
1651 Severity : enhancement
1653 Description: plain/inodebits lock performance improvement
1654 Details : Grouping plain/inodebits in granted list by their request modes
1655 and bits policy, thus improving the performance of search through
1659 Frequency : only if OST filesystem is corrupted
1661 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1662 Details : for a short time RPCs with bulk IO are in the replay list,
1663 but replay of bulk IOs is unimplemented. If the OST filesystem
1664 is corrupted due to disk cache incoherency and then replay is
1665 started it is possible to trip an assertion. Avoid putting
1666 committed RPCs into the replay list at all to avoid this issue.
1669 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1671 Description: System hang on startup
1672 Details : This bug allowed the liblustre (e.g. catamount) client to
1673 return to the app before handling all startup RPCs. This
1674 could leave the node unresponsive to lustre network traffic
1675 and manifested as a server ptllnd timeout.
1677 Severity : enhancement
1679 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1680 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1681 boolean to control whether to print peer debug info when a
1682 client's RPC times out.
1685 Frequency : only for kernels with patches from Lustre below 1.4.3
1687 Description: Remove old rdonly API
1688 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1691 Frequency : only for devices with external journals
1693 Description: Set external device read-only also
1694 Details : During a commanded failover stop, we set the disk device
1695 read-only while the server shuts down. We now also set any
1696 external journal device read-only at the same time.
1699 Frequency : when upgrading from 1.4 while trying to change parameters
1701 Description: The wrong (new) MDC name was used when setting parameters for
1702 upgraded MDT's. Also allows changing of OSC (and MDC)
1703 parameters if --writeconf is specified at tunefs upgrade time.
1706 Frequency : when setting specific ost indicies
1708 Description: QOS code breaks on skipped indicies
1709 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1710 created with --index need not be sequential.
1712 Severity : enhancement
1714 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1715 Details : The uninit_groups feature works in conjunction with the kernel
1716 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1717 up the pass1 processing of e2fsck. This is a read-only feature
1718 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1719 mount filesystems that have had this feature enabled.
1721 Severity : enhancement
1723 Description: Improve multi-block allocation algorithm to avoid fragmentation
1724 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1725 allocation locality and avoid filesystem fragmentation.
1727 ------------------------------------------------------------------------------
1729 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1731 * Support for kernels:
1732 2.4.21-47.0.1.EL (RHEL 3)
1733 2.6.5-7.283 (SLES 9)
1734 2.6.9-42.0.10.EL (RHEL 4)
1735 2.6.12.6 vanilla (kernel.org)
1736 2.6.16.27-0.9 (SLES 10)
1737 * Recommended e2fsprogs version: 1.39.cfs5
1739 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1743 Frequency : occasional, depends on client load and configuration
1744 Bugzilla : 12181, 12203
1745 Description: data loss for recently-modified files
1747 Details : In some cases it is possible that recently written or created
1748 files may not be written to disk in a timely manner (this should
1749 normally be within 30s unless client IO load is very high).
1750 The problem appears as zero-length files or files that are a
1751 multiple of 1MB in size after a client crash or client eviction
1752 that are missing data at the end of the file.
1754 This problem is more likely to be hit on clients where files are
1755 repeatedly created and unlinked in the same directory, clients
1756 have a large amount of RAM, have many CPUs, the filesystem has
1757 many OSTs, the clients are rebooted frequently, and/or the files
1758 are not accessed by other nodes after being written.
1760 The presence of the problem can be detected by looking at
1761 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1762 smaller than the second (nr_unused) then dirty files will not
1763 be flushed automatically to disk. "sync; sleep 10" should be
1764 run several times on the node before unmounting it to update
1765 Lustre (this is also safe to run on nodes without this problem).
1767 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1768 that can cause this same problem, so customers using that kernel
1769 also need to update the kernel in addition to Lustre. In order
1770 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1772 It is normal that files written just before a client crash (less
1773 than 30s) may not yet have been flushed to disk, even for local
1777 Frequency : frequent on thin XT3 nodes
1779 Description: UUID collision on thin XT3 Linux nodes
1780 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1781 randomly, since we relied on an insufficiently-seeded PRNG.
1786 Description: OSS hangs after "All ost request buffers busy"
1787 Details : A deadlock between quota and journal operations caused OSS
1788 hangs after printing "All ost request buffers busy."
1791 Frequency : always on liblustre builds
1793 Description: Cleanup compiler warnings on liblustre
1796 Frequency : always on liblustre builds on XT3
1798 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1802 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1803 > 32000 subdirectories
1804 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1805 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1806 subdirectory count crosses 32000. This will aid e2fsck to
1807 correctly handle more than 32000 subdirectories.
1810 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1812 Description: System hang on startup
1813 Details : This bug allowed the liblustre (e.g. catamount) client to
1814 return to the app before handling all startup RPCs. This
1815 could leave the node unresponsive to lustre network traffic
1816 and manifested as a server ptllnd timeout.
1818 Severity : enhancement
1820 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1821 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1822 boolean to control whether to print peer debug info when a
1823 client's RPC times out.
1828 Description: make O_SYNC working on 2.6 kernels
1829 Details : 2.6 kernels use different method for mark pages for write,
1830 so need add a code to lustre for O_SYNC work.
1835 Description: Failure to close file and release space on NFS
1836 Details : Put inode details into lock acquired in ll_intent_file_open.
1837 Use mdc_intent_lock in ll_intent_open to properly
1838 detect all kind of errors unhandled by mdc_enqueue.
1843 Description: proc file read during shutdown sometimes raced obd removal,
1845 Details : Add lock to prevent obd access after proc file removal.
1848 Frequency : Only for files larger than 4GB on 32-bit clients.
1850 Description: improperly doing page alignment of locks
1851 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1852 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1857 Description: RPCs being resent when they shouldn't be
1858 Details : Some RPCs that should not be resent are being resent. This
1859 can cause inconsistencies in the RPC state machine. Do not
1860 resend such requests.
1863 Frequency : rare, only with NFS export
1865 Description: Crash on NFS re-export node
1866 Details : under very unusual load conditions an assertion is hit in
1867 ll_intent_file_open()
1870 Frequency : only if OST filesystem is corrupted
1872 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1873 Details : for a short time RPCs with bulk IO are in the replay list,
1874 but replay of bulk IOs is unimplemented. If the OST filesystem
1875 is corrupted due to disk cache incoherency and then replay is
1876 started it is possible to trip an assertion. Avoid putting
1877 committed RPCs into the replay list at all to avoid this issue.
1882 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1883 Details : Large single O_DIRECT read and write calls can fail to allocate
1884 a sufficiently large buffer to process the request. In case of
1885 allocation failure the allocation is retried with a smaller
1886 buffer and broken into smaller requests.
1888 Severity : enhancement
1890 Description: Add -o localflock option to simulate old noflock behaviour.
1891 Details : This will achieve local-only flock/fcntl locks coherentness.
1896 Description: versioning check is incomplete
1897 Details : Checking the version difference of client vs. server, report
1898 error if the gap is too big.
1903 Description: add support PG_writeback bit
1904 Details : add support for PG_writeback bit for Lustre, for more carefull
1905 work with page cache in 2.6 kernel. This also fix some deadlocks
1906 and remove hack for work O_SYNC with 2.6 kernel.
1908 Severity : enhancement
1910 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1911 Details : The uninit_groups feature works in conjunction with the kernel
1912 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1913 up the pass1 processing of e2fsck. This is a read-only feature
1914 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1915 mount filesystems that have had this feature enabled.
1917 Severity : enhancement
1919 Description: Improve multi-block allocation algorithm to avoid fragmentation
1920 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1921 allocation locality and avoid filesystem fragmentation.
1923 ------------------------------------------------------------------------------
1925 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1927 * Support for kernels:
1928 2.6.9-42.0.3.EL (RHEL 4)
1929 2.6.5-7.276 (SLES 9)
1930 2.4.21-47.0.1.EL (RHEL 3)
1931 2.6.12.6 vanilla (kernel.org)
1932 2.6.16.21-0.8 (SLES10)
1933 * Recommended e2fsprogs version: 1.39.cfs2-0
1935 * The backwards-compatible /proc/sys/portals symlink has been removed
1936 in this release. Before upgrading, please ensure that you change
1937 any configuration scripts or /etc/sysctl.conf files that access
1938 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1939 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1940 in advance of the upgrade on any system running Lustre 1.4.6 or
1941 newer, since /proc/sys/lnet was added in that version.
1942 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1946 Frequency : only when quota is used
1948 Description: avoid scanning export list for quota master
1949 Details : Change the algorithms to avoid scanning export list in order
1950 to improve the efficiency.
1953 Frequency : MDS failover only, very rarely
1955 Description: "went back in time" messages on mds failover
1956 Details : The greatest transno may be lost when the current operation
1957 finishes with an error (transno==0) and the client's last_rcvd
1958 record is over-written. Save the greatest transno in the
1959 mds_last_transno for this case.
1962 Frequency : always for specific kernels and striping counts
1964 Description: client may get "Matching packet too big" without ACL support
1965 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
1966 when trying to access files in certain configurations. The
1967 clients should in fact be denied when mounting because they do
1968 not understand ACLs.
1971 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
1973 Description: many clients connecting with IO in progress causes connect timeouts
1974 Details : Avoid synchronous journal commits to avoid delays caused by many
1975 clients connecting/disconnecting when bulk IO is in progress.
1976 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
1977 OST_IO_PORTAL to avoid delays behind potentially many pending
1981 Frequency : occasionally with multiple writers to a single file
1983 Description: shared writes to file may result in wrong size reported by stat()
1984 Details : Allow growing of kms when extent lock is cancelled
1987 Frequency : always with random mmap IO to multi-striped file
1989 Description: mmap write might be lost if we are writing to a 'hole' in stripe
1990 Details : Only if the hole is at the end of OST object so that kms is too
1991 small. Fix is to increase kms accordingly in ll_nopage.
1994 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
1996 Description: writes to a missing object would leak memory on the OST
1997 Details : If there is an inconsistency between the MDS and OST filesystems,
1998 such that the MDS references an object that doesn't exist, writes
1999 to that object will leak memory due to incorrect cleanup in the
2000 error handling path, eventually running out of memory on the OST.
2005 Description: Creating too long symlink causes lustre errors
2006 Details : Check symlink and name lengths before sending requests to MDS.
2009 Frequency : only if flock is enabled (not on by default)
2011 Description: posix locks not released on fd closure on 2.6.9+
2012 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2013 kernels, this caused such locks not to be released on fd close and
2014 then assertions on fs unmount about still used locks.
2017 Frequency : MDS failover only, very rarely
2019 Description: clients may get ASSERTION(granted_lock != NULL)
2020 Details : When request was taking a long time, and a client was resending
2021 a getattr by name lock request. The were multiple lock requests
2022 with the same client lock handle and
2023 mds_getattr_name->fixup_handle_for_resent_request found one of the
2024 lock handles but later failed with ASSERTION(granted_lock != NULL).
2029 Description: handle->h_buffer_credits > 0, assertion failure
2030 Details : h_buffer_credits is zero after truncate, causing assertion
2031 failure. This patch extends the transaction or creates a new
2035 Frequency : NFS re-export or patchless client
2036 Bugzilla : 11179, 10796
2037 Description: Crash on NFS re-export node (__d_move)
2038 Details : We do not want to hash the dentry if we don't have a lock.
2039 But if this dentry is later used in d_move, we'd hit uninitialised
2040 list head d_hash, so we just do this to init d_hash field but
2041 leave dentry unhashed.
2044 Frequency : NFS re-export or patchless client
2046 Description: NFS exports has problem with symbolic link
2047 Details : lustre client didn't properly install dentry when re-exported
2048 to NFS or running patchless client.
2051 Frequency : NFS re-export or patchless client
2053 Description: Various nfs/patchless fixes.
2054 Details : fixes reuse disconected alias for lookup process - this fixes
2055 warning "find_exported_dentry: npd != pd",
2056 fix permission error with open files at nfs.
2057 fix apply umask when do revalidate.
2060 Frequency : occasional
2062 Description: Crash on NFS re-export node
2063 Details : calling clear_page() on the wrong pointer triggered oops in
2064 generic_mapping_read().
2067 Frequency : rarely, using O_DIRECT IO
2069 Description: unaligned directio crashes client with LASSERT
2070 Details : check for unaligned buffers before trying any requests.
2073 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2075 Description: stale data returned from RAID cache
2076 Details : If only a small amount of IO is done to the RAID device before
2077 reading it again it is possible to get stale data from the RAID
2078 cache instead of reading it from disk.
2081 Frequency : always for sles10 kernel
2083 Description: sles10 support
2084 Details : ll_follow_link: compile fixes and using of nd_set_link
2085 under newer kernels.
2088 Frequency : depends on arch, kernel and compiler version, always on sles10
2091 Description: recursive or deep enough symlinks cause stack overflow
2092 Details : getting rid of large stack-allocated variable in
2096 Frequency : depends on hardware
2098 Description: lustre write performance loss in the SLES10 kernel
2099 Details : the performance loss is caused by using of write barriers in the
2100 ext3 code. The SLES10 kernel turns barrier support on by
2101 default. The fix is to undo that change for ldiskfs.
2103 ------------------------------------------------------------------------------
2105 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2107 * Support for kernels:
2108 2.6.9-42.0.3EL (RHEL 4)
2109 2.6.5-7.276 (SLES 9)
2110 2.4.21-47.0.1.EL (RHEL 3)
2111 2.6.12.6 vanilla (kernel.org)
2115 Frequency : quota enabled and large files being deleted
2117 Description: releasing more than 4GB of quota at once hangs OST
2118 Details : If a user deletes more than 4GB of files on a single OST it
2119 will cause the OST to spin in an infinite loop. Release
2120 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2125 Description: statfs data retrieved from /proc may be stale or zero
2126 Details : When reading per-device statfs data from /proc, in the
2127 {kbytes,files}_{total,free,avail} files, it may appear
2128 as zero or be out of date.
2131 Frequency : systems with MD RAID1 external journal devices
2133 Description: lconf's call to blkid is confused by RAID1 journal devices
2134 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2135 of returning all block devices that match the journal UUID.
2138 Frequency : always, for aggregate stripe size over 4GB
2140 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2141 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2142 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2145 Frequency : always if "lfs find" used on a local file/directory
2147 Description: "lfs find" segfaults if used on a local file/directory
2148 Details : The case where a directory component was not specified wasn't
2149 handled correctly. Handle this properly.
2152 Frequency : always on ppc64
2154 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2155 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2156 perform endianness conversion.
2159 Frequency : rarely (truncate to non-zero file size after write under load)
2160 Bugzilla : 10730, 10687
2161 Description: Files padded with zeros to next 4K multiple
2162 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2163 it is possible that files that are truncated to a non-zero size
2164 immediately after being written are filled with zero bytes beyond
2165 the truncated size. No file data is lost.
2167 Severity : enhancement
2169 Description: Allow recovery/failover for liblustre clients.
2170 Details : liblustre clients were unaware of failover configurations until
2173 Severity : enhancement
2175 Description: user file locks should fail when not mounting with flock option
2176 Details : Set up an error-returning stub in ll_file_operations.lock field
2177 to prevent incorrect behaviour when client is mounted without
2178 flock option. Also, set up properly f_op->flock field for
2182 Frequency : always on ia64
2184 Description: "lfs df" loops on printing out MDS statfs information
2185 Details : The obd_ioctl_data was not initialized and in some systems
2186 this caused a failure during the ioctl that did not return
2187 an error. Initialize the struct and return an error on failure.
2190 Frequency : SLES 9 only
2192 Description: Error of copying files with lustre special EAs as root
2193 Details : Client side always return success for setxattr call for lustre
2194 special xattr (currently only "trusted.lov").
2197 Frequency : rarely on clusters with both ia64+i386 clients
2199 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2200 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2201 clients are doing concurrent writes to the same file it is
2202 possible that the ia64 clients may LASSERT because the OST
2203 extent locks are not PAGE_SIZE aligned. Ensure that grown
2204 locks are always aligned on the request boundary.
2207 Frequency : specific use, occasional
2209 Description: Overwriting in use executable truncates on-disk binary image
2210 Details : If one node attempts to overwrite an executable in use by
2211 another node, we now correctly return ETXTBSY instead of
2212 truncating the file.
2214 Severity : enhancement
2216 Description: Async OSC create to avoid the blocking unnecessarily.
2217 Details : If a OST has no remain object, system will block on the creating
2218 when need to create a new object on this OST. Now, ways use
2219 pre-created objects when available, instead of blocking on an
2220 empty osc while others are not empty. If we must block, we block
2221 for the shortest possible period of time.
2226 Description: chmod on Lustre root is propagated to other clients
2227 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2228 invalid by the follow_mount time.
2231 Frequency : liblustre clients only
2233 Description: Race in 'instant cancel' lock handling could lead to such locks
2234 never to be granted in case of SMP MDS
2235 Details : Do not destroy not yet granted but cbpending locks in
2239 Frequency : replay/resend of open
2241 Description: non null lock assetion failure in mds_intent_policy
2242 Details : Trying to replay/resend lockless open requests resulted in
2243 mds_open() returning 0 with no lock. Now it sets a flag if
2244 a lock is going to be returned.
2246 Severity : enhancement
2248 Description: Checksum enhancements
2249 Details : New checksum enhancements allow for resending RPCs that failed
2252 Severity : enhancement
2254 Description: Tunables on number of dirty pages in cacche
2255 Details : Allow to set limit on number of dirty pages cached.
2260 Description: client crash on unmount - lock still has references
2261 Details : In some error handling cases it was possible to leak a lock
2262 reference on a client while accessing a file. This was not
2263 harmful to the client during operation, but would cause the
2264 client to crash when the filesystem is unmounted.
2267 Frequency : specific case, rare
2269 Description: ETXTBSY on mds though file not in use by client
2270 Details : ETXTBSY is no longer incorrectly returned when attempting to
2271 chmod or chown a directory that the user previously tried to
2272 execute or a currently-executing binary.
2275 Frequency : extremely rare except on liblustre-based clients
2277 Description: Lustre space not freed when files are deleted
2278 Details : Clean up open-unlinked files after client eviction. Previously
2279 the unlink was skipped and the files remained as orphans.
2284 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2285 Details : In some cases it was possible to send a blocking callback to
2286 a client doing a glimpse, even though that client didn't get
2287 a lock granted. When the glimpse lock is cancelled on the OST
2288 the freed lock is left on the waiting list and corrupted the list.
2291 Frequency : all core dumps
2293 Description: Broke core dumps to lustre
2294 Details : Negative dentry may be unhashed if parent does not have UPDATE
2295 lock, but some callers, e.g. do_coredump, expect dentry to be
2296 hashed after successful create, hash it in ll_create_it.
2298 ------------------------------------------------------------------------------
2300 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2302 * Support for kernels:
2303 2.6.9-42.0.2.EL (RHEL 4)
2304 2.6.5-7.276 (SLES 9)
2305 2.4.21-40.EL (RHEL 3)
2306 2.6.12.6 vanilla (kernel.org)
2310 Frequency : always on RHEL 3
2312 Description: Number of open files grows over time
2313 Details : The number of open files grows over time, whether or not
2314 Lustre is started. This was due to a filp leak introduced
2315 by one of our kernel patches.
2317 ------------------------------------------------------------------------------
2319 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2321 * Support for kernels:
2322 2.6.9-42.EL (RHEL 4)
2323 2.6.5-7.276 (SLES 9)
2324 2.4.21-40.EL (RHEL 3)
2325 2.6.12.6 vanilla (kernel.org)
2330 Bugzilla : 5719, 9635, 9792, 9684
2331 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2332 Details : If a server is under heavy load and cannot reply to new
2333 connection requests before the client resends the (re)connect,
2334 the connection handling code can behave badly if two service
2335 threads are concurrently handing separate (re)connections from
2336 the same client. Add better locking to the connection handling
2337 code, and ensure that only a single connection will be processed
2338 for a given client UUID, even if the lock is dropped.
2340 Severity : enhancement
2342 Description: add TCP zero-copy support to kernel
2343 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2344 sends if the hardware supports scatter-gather and checksumming.
2345 This allows socklnd to do client-write and server-read more
2346 efficiently and reduce CPU utilization from skbuf copying.
2349 Frequency : only if NFS exporting from client
2351 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2352 Details : If setting attributes on a file created under NFS that had
2353 never been opened it would be possible to oops the client
2354 if the file had no objects.
2357 Frequency : always for liblustre
2359 Description: liblustre client does MDS+OSTs setattr RPC for each write
2360 Details : When doing a write from a liblustre client, the client
2361 incorrectly issued an RPC to the MDS and each OST the file was
2362 striped over in order to update the timestamps. When writing
2363 with small chunks and many clients this could overwhelm the MDS
2364 with RPCs. In all cases it would slow down the write because
2365 these RPCs are unnecessary.
2367 Severity : enhancement
2369 Description: allow number of MDS service threads to be changed at module load
2370 Details : It is now possible to change the number of MDS service threads
2371 running. Adding "options mds mds_num_threads={N}" to the MDS's
2372 /etc/modprobe.conf will set the number of threads for the next
2373 time Lustre is restarted (assuming the "mds" module is also
2374 reloaded at that time). The default number of threads will
2375 stay the same, 32 for most systems.
2380 Description: OST crash if filesystem is unformatted or corrupt
2381 Details : If an OST is started on a device that has never been formatted
2382 or if the filesystem is corrupt and cannot even mount then the
2383 error handling cleanup routines would dereference a NULL pointer.
2388 Description: NULL pointer deref in llap_from_page.
2389 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2390 incorrect) mapping if the page was truncated/reclaimed while it was
2391 searched for. Check for this condition and skip such pages when
2392 doing readahead. Introduce extra check to llap_from_page() to
2393 verify page->mapping->host is non-NULL (so page is not anonymous).
2396 Frequency : Sometimes when using sys_sendfile
2398 Description: "page not covered by a lock" warnings from ll_readpage
2399 Details : sendfile called ll_readpage without right page locks present.
2400 Now we introduced ll_file_sendfile that does necessary locking
2401 around call to generic_file_sendfile() much like we do in
2405 Frequency : with certain MDS communication failures at client mount time
2407 Description: NULL pointer deref after failed client mount
2408 Details : a client connection request may delayed by the network layer
2409 and not be sent until after the PTLRPC layer has timed out the
2410 request. If the client fails the mount immediately it will try
2411 to clean up before the network times out the request. Add a
2412 reference from the request import to the obd device and delay
2413 the cleanup until the network drops the request.
2416 Frequency : occasionally during client (re)connect
2418 Description: assertion failure during client (re)connect
2419 Details : processing a client connection request may be delayed by the
2420 client or server longer than the client connect timeout. This
2421 causes the client to resend the connection request. If the
2422 original connection request is replied in this interval, the
2423 client may trip an assertion failure in ptlrpc_connect_interpret()
2424 which thought it would be the only running connect process.
2427 Frequency : only with obd_echo servers and clients that are rebooted
2429 Description: kernel BUG accessing uninitialized data structure
2430 Details : When running an obd_echo server it did not start the ping_evictor
2431 thread, and when a client was evicted an uninitialized data
2432 structure was accessed. Start the ping_evictor in the RPC
2433 service startup instead of the OBD startup.
2435 Severity : enhancement
2436 Bugzilla : 10193 (patchless)
2437 Description: Remove dependency on various unexported kernel interfaces.
2438 Details : No longer need reparent_to_init, exit_mm, exit_files,
2439 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2442 Frequency : rare (only users of deprecated and unsupported LDAP config)
2444 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2445 Details : LDAP apparently contains 'lustreName' attributes instead of
2446 'name'. A simple remapping of the name is sufficient.
2449 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2451 Description: waiting_locks_callback trips kernel BUG if client is evicted
2452 Details : Running with the dump_on_timeout debug flag turned on makes
2453 it possible that the waiting_locks_callback() can try to dump
2454 the Lustre kernel debug logs from an interrupt handler. Defer
2455 this log dumping to the expired_lock_main() thread.
2457 Severity : enhancement
2459 Description: Support NFS exporting on 2.6 kernels.
2460 Details : Implement non-rawops metadata methods for NFS server to use without
2461 changing NFS server code.
2464 Frequency : very rare (synthetic metadata workload only)
2466 Description: two racing renames might cause an MDS thread to deadlock
2467 Details : Running the "racer" program may cause one MDS thread to rename
2468 a file from being the source of a rename to being the target of
2469 a rename at exactly the same time that another thread is doing
2470 so, and the second thread has already enqueued these locks after
2471 doing a lookup of the target and is trying to relock them in
2472 order. Ensure that we don't try to re-lock the same resource.
2475 Frequency : only very large systems with liblustre clients
2477 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2478 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2479 all OSTs in parallel. This allows the eviction of stale liblustre
2480 clients to proceed much faster than if they were done in series,
2481 and also offers similar improvements for other set_info RPCs.
2486 Description: excessive CPU usage during initial read phase on client
2487 Details : During the initial read phase on a client, it would agressively
2488 retry readahead on the file, consuming too much CPU and impacting
2489 performance (since 1.4.5.8). Improve the readahead algorithm
2490 to avoid this, and also improve some other common cases (read
2491 of small files in particular, where "small" is files smaller than
2492 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2497 Description: MDS crash when receiving packet with unknown intent.
2498 Details : Do not LBUG in unknown intent case, just return -EFAULT
2500 Severity : enhancement
2501 Bugzilla : 9293, 9385
2502 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2503 Details : Do not serialize getattr (non-intent version) and statfs.
2506 Frequency : occasional, when OST network is overloaded/intermittent
2508 Description: client evicted by OST after bulk IO timeout
2509 Details : If a client sends a bulk IO request (read or write) the OST
2510 may evict the client if it is unresposive to its data GET/PUT
2511 request. This is incorrect if the network is overloaded (takes
2512 too long to transfer the RPC data) or dropped the OST GET/PUT
2513 request. There is no need to evict the client at all, since
2514 the pinger and/or lock callbacks will handle this, and the
2515 client can restart the bulk request.
2518 Frequency : Always when mmapping file with no objects
2520 Description: client crashes when mmapping file with no objects
2521 Details : Check that we actually have objects in a file before doing any
2522 operations on objects in ll_vm_open, ll_vm_close and
2528 Description: Request leak when working with deleted CWD
2529 Details : Introduce advanced request refcount tracking for requests
2530 referenced from lustre intent.
2532 Severity : Enhancement
2534 Description: Cache open file handles on client.
2535 Details : MDS now will return special lock along with openhandle, if
2536 requested and client is allowed to hold openhandle, even if unused,
2537 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2538 closing files for every read/write openration.
2540 Severity : Enhancement
2542 Description: Cache open negative dentries on client when possible.
2543 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2544 negative dentries on lock revocation.
2549 Description: Remounting a client read-only wasn't possible with a zconf mount
2550 Details : It wasn't possible to remount a client read-only with llmount.
2552 Severity : enhancement
2553 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2554 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2555 MPICH is included to add Lustre-specific ADIO interfaces.
2556 This is based closely on the UFS ADIO layer and only differs
2557 in file creation, in order to allow the OST striping to be set.
2558 This is user-contributed code and not supported by CFS.
2563 Description: extended inode attributes (immutable, append-only) work improperly
2564 when 2.4 and 2.6 kernels are used on client/server or vice versa
2565 Details : Introduce kernel-independent values for these flags.
2567 Severity : enhancement
2570 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2571 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2572 granularity. Now, Fractional MB granularity is be supported,
2573 this is very useful for low memory system.
2575 Severity : enhancement
2577 Description: Getattr by fid
2578 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2579 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2580 have locks and avoids getattr rpc requests for it on every stat.
2583 Frequency : Always, for filesystems larger than 2TB
2585 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2586 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2587 filesystems larger than 2TB to be mounted with mballoc enabled.
2590 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2592 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2593 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2594 filesystems larger than 2TB is mounted then the kernel will
2595 silently corrupt the start of the filesystem. CONFIG_LBD is
2596 enabled for all CFS-supported kernels, but the possibility of
2597 this happening with a modified kernel config exists.
2599 Severity : enhancement
2601 Description: add client O_DIRECT support for 2.6 kernels
2602 Details : It is now possible to do O_DIRECT reads and writes to files
2603 in the Lustre client mountpoint on 2.6 kernel clients.
2605 Severity : enhancement
2607 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2608 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2609 parallel, not waiting for response from every OST before sending
2610 a rpc to the next OST.
2615 Description: setattr vs write race when updating file timestamps
2616 Details : Client processes that update a file timestamp into the past
2617 right after writing to the file (e.g. tar) it is possible that
2618 the updated file modification time can be reset to the current
2619 time due to a race between processing the setattr and write RPC.
2621 Severity : enhancement
2623 Description: Bring 'lfs find' closer in line with regular Linux find.
2624 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2625 -print0 options and obtains all the needed info through the lustre
2628 Severity : enhancement
2630 Description: support up to 1024 configured devices on one node
2631 Details : change obd_dev array from statically allocated to dynamically
2632 allocated structs as they are first used to reduce memory usage
2637 Description: Flush dirty partially truncated pages during truncate
2638 Details : Immediatelly flush partially truncated pages in filter_setattr,
2639 this way we completely avoid having any pages in page cache on OST
2640 and can retire ugly workarounds during writes to flush such pages.
2645 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2646 Details : protect inode from truncation within vfs_unlink() context
2647 just take a reference before calling vfs_unlink() and release it
2648 when parent's i_sem is free.
2653 Description: last_id value checked outside lock on OST caused LASSERT failure
2654 Details : If there were multiple MDS->OST object precreate requests in
2655 flight, it was possible that the OST's last object id was checked
2656 outside a lock and incorrectly tripped an assertion. Move checks
2657 inside locks, and discard old precreate requests.
2660 Frequency : always, if extents are used on OSTs
2662 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2663 Details : OSTs using the extents format would not zero the high 16 bits of
2664 the index physical block number. This is not a problem for any
2665 OST filesystems smaller than 16TB, and no kernels support ext3
2666 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2667 new/modified files) and can be fixed for existing filesystems
2668 with e2fsprogs-1.39-cfs1.
2673 Description: import connection selection may be incorrect if timer wraps
2674 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2675 import connections to be ignored if the 32-bit jiffies counter
2676 wraps. Use a 64-bit jiffies counter.
2679 Frequency : very large clusters immediately after boot
2681 Description: LNET request buffers exhausted under heavy short-term load
2682 Details : If a large number of client requests are generated on a service
2683 that has previously never seen so many requests it is possible
2684 that the request buffer growth cannot keep up with the spike in
2685 demand. Instead of dropping incoming requests, they are held in
2686 the LND until the RPC service can accept more requests.
2689 Frequency : Sometimes during replay
2691 Description: Assertion failure in ll_local_open after replay.
2692 Details : If replay happened on an open request reply before we were able
2693 to set replay handler, reply will become not swabbed tripping the
2694 assertion in ll_local_open. Now we set the handler right after
2695 recognising of open request
2698 Frequency : very rare
2700 Description: kernel reports "badness in vsnprintf"
2701 Details : Reading from the "recovery_status" /proc file in small chunks
2702 may cause a negative length in lprocfs_obd_rd_recovery_status()
2703 call to vsnprintf() (which is otherwise harmless). Exit early
2704 if there is no more space in the output buffer.
2706 Severity : enhancement
2708 Description: clear OBD RPC statistics by writing to them
2709 Details : It is now possible to clear the OBD RPC statistics by writing
2710 to the "stats" file.
2715 Description: Client mtime is not the same on different clients after utimes
2716 Details : In some cases, the client was using the utimes() syscall on
2717 a file cached on another node. The clients now validate the
2718 ctime from the MDS + OSTs to determine which one is right.
2723 Description: Inability to activate failout mode
2724 Details : lconf script incorrectly assumed that in pythong string's numeric
2725 value is used in comparisons.
2728 Frequency : always with multiple stripes per file
2730 Description: Inefficient object allocation for mutli-stripe files
2731 Details : When selecting which OSTs to stripe files over, for files with
2732 a stripe count that divides evenly into the number of OSTs,
2733 the MDS is always picking the same starting OST for each file.
2734 Return the OST selection heuristic to the original design.
2739 Description: mount failures may take full timeout to return an error
2740 Details : Under some heavy load conditions it is possible that a
2741 failed mount can wait for the full obd_timeout interval,
2742 possibly several minutes, before reporting an error.
2743 Instead return an error as soon as the status is known.
2745 Frequency : quota enabled and large files being deleted
2747 Description: releasing more than 4GB of quota at once hangs OST
2748 Details : If a user deletes more than 4GB of files on a single OST it
2749 will cause the OST to spin in an infinite loop. Release
2750 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2755 Description: statfs data retrieved from /proc may be stale or zero
2756 Details : When reading per-device statfs data from /proc, in the
2757 {kbytes,files}_{total,free,avail} files, it may appear
2758 as zero or be out of date.
2761 Frequency : systems with MD RAID1 external journal devices
2763 Description: lconf's call to blkid is confused by RAID1 journal devices
2764 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2765 of returning all block devices that match the journal UUID.
2768 Frequency : always, for aggregate stripe size over 4GB
2770 Description: assertion fails when trying to use 4GB stripe size
2771 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2772 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2775 Frequency : always on ppc64
2777 Description: the first write on an ext3 filesystem with mballoc got stuck
2778 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2779 perform endianness conversion.
2781 ------------------------------------------------------------------------------
2783 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2785 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2786 INTEROPERATE with older versions automatically. Please read the
2787 user documentation before upgrading any part of a live system.
2788 * WARNING: Lustre networking configuration changes are required with
2789 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2792 * Support for kernels:
2793 2.6.9-22.0.2.EL (RHEL 4)
2794 2.6.5-7.244 (SLES 9)
2795 2.6.12.6 vanilla (kernel.org)
2798 Severity : enhancement
2799 Bugzilla : 7981/8208
2800 Description: Introduced Lustre Networking (LNET)
2801 Details : LNET is new networking infrastructure for Lustre, it includes
2802 a reorganized network configuration mode (see the user
2803 documentation for full details) as well as support for routing
2804 between different network fabrics. Lustre Networking Devices
2805 (LNDs) for the supported network fabrics have also been
2806 created for this new infrastructure.
2808 Severity : enhancement
2809 Description: Introduced Access control lists
2810 Details : clients can set ACLs on files and directories in order to have
2811 more fine-grained permissions than the standard Unix UGO+RWX.
2812 The MDS must be started with the "-o acl" mount option.
2814 Severity : enhancement
2815 Description: Introduced filesystem quotas
2816 Details : Administrators may now establish per-user quotas on the
2819 Severity : enhancement
2821 Description: Configuration change for the XT3
2822 The PTLLND is now used to run Lustre over Portals on the XT3
2823 The configure option(s) --with-cray-portals are no longer used.
2824 Rather --with-portals=<path-to-portals-includes> is used to
2825 enable building on the XT3. In addition to enable XT3 specific
2826 features the option --enable-cray-xt3 must be used.
2831 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2832 Details : A race between allocating a new llcd and re-getting the llcd_lock
2833 allowed another thread to grab newly-allocated llcd.
2835 Severity : enhancement
2837 Description: 2.6 OST async journal commit and locking fix to improve performance
2838 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2839 2.6 kernels are now async as they already were in 2.4 kernels so
2840 that they can commit concurrently with the network bulk transfer.
2841 For block-allocated files the filter allocation semaphore is held
2842 to avoid filesystem fragmentation during allocation. BKL lock
2843 removed for 2.6 xattr operations where it is no longer needed.
2848 Description: lconf incorrectly determined whether two IP networks could talk
2849 Details : In some more complicated routing and multiple-network
2850 configurations, lconf will avoid trying to make a network
2851 connection to a disjoint part of the IP space. It was doing the
2852 math incorrectly for one set of cases.
2857 Description: Fix for potential infinite loop processing records in an llog.
2858 Details : If an llog record is corrupted/zeroed, it is possible to loop
2859 forever in llog_process(). Validate the llog record length
2860 and skip the remainder of the block on error.
2863 Frequency : occasional (liblustre only)
2865 Description: liblustre could not open files whose last component is a symlink
2866 Details : sysio_path_walk() would incorrectly pass the open intent to
2867 intermediate path components.
2870 Frequency : rare (liblustre only with non-standard tuning)
2871 Bugzilla : 7201 (7350)
2872 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2873 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2874 after it has been released, only do this for referenced locks
2876 Severity : enhancement
2878 Description: specifying an (invalid) directory default stripe_size of -1
2879 would reset the directory default striping
2880 Details : stripe_size -1 was used internally to signal directory stripe
2881 removal, now use "all default" to signal dir stripe removal
2882 as a directory striping of "all default" is not useful
2885 Frequency : common for large clusters running liblustre clients
2887 Description: doing an ls when liblustre clients are running is slow
2888 Details : sending a glimpse AST to a liblustre client waits for every AST
2889 to time out, as liblustre clients will not respond. Since they
2890 cannot cache data we refresh the OST lock LVB from disk instead.
2892 Severity : enhancement
2894 Description: doing an ls at the same time as file IO can be slow
2895 Details : enqueue and other "small" requests can be blocked behind many
2896 large IO requests. Create a new OST IO portal for non-IO
2897 requests so they can be processed faster.
2900 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2902 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2903 created file with mode 000
2904 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2905 allow an ftruncate() as a non-root user to a file with mode 000.
2906 Lustre now respects this flag to disable mode checks when
2907 truncating a file owned by the user
2910 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2913 Description: Revoking locks from clients that went dead or catatonic might take
2915 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2916 cancellation of such locks instant on servers without waiting for
2917 any reply from clients. Clients drops these locks when cancel
2918 notification from server is received without replying.
2921 Frequency : liblustre-only, when liblustre client dies or becomes busy
2923 Description: Doing ls on Linux clients can take a long time with active
2925 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2926 granting such locks to it in the first place if possible. Locks
2927 are taken by proxy on the OST during the read or write and
2928 dropped immediately afterward. Add connect flags handling, do
2929 not grant locks to liblustre clients for glimpse ASTs.
2931 Severity : enhancement
2933 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2934 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2935 sequential reads and behaves badly with applications doing random
2936 reads. Improve it by reading ahead at least read region, and
2937 avoiding excessive large RPC for small reads.
2939 Severity : enhancement
2941 Description: Creating more than 1000 files for a single job may cause a load
2942 imbalance on the OSTs if there are also a large number of OSTs.
2943 Details : qos_prep_create() uses an OST index reseed value that is an
2944 even multiple of the number of available OSTs so that if the
2945 reseed happens in the middle of the object allocation it will
2946 still utilize the OSTs as uniformly as possible.
2951 Description: OST or MDS may oops in ping_evictor_main()
2952 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
2953 but doesn't restart at beginning of obd_exports_timed list
2956 Severity : enhancement
2958 Description: improve by-nid export eviction on the MDS and OST
2959 Details : allow multiple exports with the same NID to be evicted at one
2960 time without re-searching the exports list.
2963 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
2965 Description: MDS may oops in groups_free()
2966 Details : in rare race conditions a newly allocated group_info struct is
2967 freed again, and this can be NULL. The 2.4 compatibility code
2968 for groups_free() checked for a NULL pointer, but 2.6 did not.
2971 Frequency : common for liblustre clients doing little filesystem IO
2972 Bugzilla : 9352, 7313
2973 Description: server may evict liblustre clients accessing contended locks
2974 Details : if a client is granted a lock or receives a completion AST
2975 with a blocking AST already set it would not reply to the AST
2976 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
2979 Frequency : lfs setstripe, only systems with more than 160 OSTs
2981 Description: unable to set striping with a starting offset beyond OST 160
2982 Details : llapi_create_file() incorrectly limited the starting stripe
2983 index to the maximum single-file stripe count.
2986 Frequency : LDAP users only
2988 Description: lconf did not handle in-kernel recovery with LDAP properly
2989 Details : lconf/LustreDB get_refs() is searching the wrong namespace
2991 Severity : enhancement
2993 Description: bind OST threads to NUMA nodes to improve performance
2994 Details : all OST threads are uniformly bound to CPUs on a single NUMA
2995 node and do their allocations there to localize memory access
2997 Severity : enhancement
2999 Description: llmount can determine client NID directly from Myrinet (GM)
3000 Details : the client NID code from gmnalnid was moved directly into
3001 llmount, removing the need to use this or specifying the
3002 client NID explicitly when mounting GM clients with zeroconf
3005 Frequency : if client is started with down MDS
3007 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3008 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3009 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3010 ptlrpc_import_delayed_req() because we didn't check if the
3011 request was interrupted, and we also didn't break out of the
3012 event loop if there was no timeout
3017 Description: data loss during non-page-aligned writes to a single file from
3018 both multiple nodes and multiple threads on one node at same time
3019 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3020 inconsistency led to false short-reads, that were cached and later
3021 used by ->prepare_write() to fill in partially written page,
3022 leading to data loss.
3025 Frequency : always, if lconf --abort_recovery used
3027 Description: lconf --abort_recovery fails with 'Operation not supported'
3028 Details : lconf was attempting to abort recovery on the MDT device and not
3031 Severity : enhancement
3033 Description: remove cleanup logs
3034 Details : replace lconf-generated cleanup logs with lustre internal
3035 cleanup routines. Eliminates the need for client-cleanup and
3038 Severity : enhancement
3040 Description: add support for EAs (user and system) on lustre filesystems
3041 Details : it is now possible to store extended attributes in the Lustre
3042 client filesystem, and with the user_xattr mount option it
3043 is possible to allow users to store EAs on their files also
3045 Severity : enhancement
3047 Description: Add possibility (config option) to show minimal available OST free
3049 Details : When compiled with --enable-mindf configure option, statfs(2)
3050 (and so, df) will return least minimal free space available from
3051 all OSTs as amount of free space on FS, instead of summary of
3052 free spaces of all OSTs.
3054 Severity : enhancement
3056 Description: do not expand extent locks acquired on OST-side
3057 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3058 by server: they are not cached anyway.
3061 Frequency : when mmap is used/binaries executed from Lustre
3063 Description: Unmmap pages before throwing them away from read cache.
3064 Details : llap_shrink cache now attempts to unmap pages before discarding
3065 them (if unmapping failed - do not discard). SLES9 kernel has
3066 extra checks that trigger if this unmapping is not done first.
3071 Description: lconf didn't resolve symlinks before checking to see whether a
3072 given mountpoint was already in use
3075 Frequency : when migrating failover services
3076 Bugzilla : 6395, 9514
3077 Description: When migrating a subset of services from a node (e.g. failback
3078 from a failover service node) the remaining services would
3079 time out and evict clients.
3080 Details : lconf --force (implied by --failover) sets the global obd_timeout
3081 to 5 seconds in order to quickly disconnect, but this caused
3082 other RPCs to time out too quickly. Do not change the global
3083 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3085 Severity : enhancement
3086 Frequency : if MDS is started with down OST
3087 Bugzilla : 9439,5706
3088 Description: Allow startup/shutdown of an MDS without depending on the
3089 availability of the OSTs.
3090 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3091 Add appropriate locking and lov-osc refcounts for safe
3092 cleaning. Add osc abort_inflight calls in case the
3093 synchronize never started.
3096 Frequency : occasional (Cray XT3 only)
3098 Description: root not authorized to access files in CRAY_PORTALS environment
3099 Details : The client process capabilities were not honoured on the MDS in
3100 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3101 been accessed by an authorized user then root was able to access
3102 the file on the local client also. The root user capabilities
3103 are now allowed on the MDS, as this environment has secure UID.
3106 Frequency : occasional
3108 Description: ldiskfs "too long searching" message happens too often
3109 Details : A debugging message (otherwise harmless) prints too often on
3110 the OST console. This has been reduced to only happen when
3111 there are fragmentation problems on the filesystem.
3116 Description: Division by zero in statfs when all OSCs are inactive
3117 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3118 lov_statfs divides by value returned by lov_get_stripecnt().
3122 Bugzilla : 9489, 3273
3123 Description: First write from each client to each OST was only 4kB in size,
3124 to initialize client writeback cache, which caused sub-optimal
3125 RPCs and poor layout on disk for the first writen file.
3126 Details : Clients now request an initial cache grant at (re)connect time
3127 and so that they can start streaming writes to the cache right
3128 away and always do full-sized RPCs if there is enough data.
3129 If the OST is rebooted the client also re-establishes its grant
3130 so that client cached writes will be honoured under the grant.
3135 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3136 Details : Now I/O RPCs go to different portal number and (presumably) fast
3137 lock requests (and glimses) and other RPCs get their own service
3138 threads pool that should be able to service those RPCs
3141 Severity : enhancement
3143 Description: Ability to exchange lustre version between client and servers and
3144 issue warnings at client side if client is too old. Also for
3145 liblustre clients there is ability to refuse connection of too old
3147 Details : New 'version' field is added to connect data structure that is
3148 filled with version info. That info is later checked by server and
3152 Frequency : rare, liblustre only.
3153 Bugzilla : 9296, 9581
3154 Description: Two simultaneous writes from liblustre at offset within same page
3155 might proceed at the same time overwriting eachother with stale
3157 Details : I/O lock withing llu_file_prwv was released too early, before data
3158 actually was hitting the wire. Extended lock-holding time until
3159 server acknowledges receiving data.
3162 Frequency : extremely rare. Never observed in practice.
3164 Description: avoid generating lustre_handle cookie of 0.
3165 Details : class_handle_hash() generates handle cookies by incrementing
3166 global counter, and can hit 0 occasionaly (this is unlikely, but
3167 not impossible, because initial value of cookie counter is
3168 selected randonly). Value of 0 is used as a sentinel meaning
3169 "unassigned handle" --- avoid it. Also coalesce two critical
3170 sections in this function into one.
3172 Severity : enhancement
3174 Description: allow liblustre clients to delegate truncate locking to OST
3175 Details : To avoid overhead of locking, liblustre client instructs OST to
3176 take extent lock in ost_punch() on client's behalf. New connection
3177 flag is added to handle backward compatibility.
3179 Severity : enhancement
3180 Bugzilla : 4928, 7341, 9758
3181 Description: allow number of OST service threads to be specified
3182 Details : a module parameter allows the number of OST service threads
3183 to be specified via "options ost ost_num_threads={N}" in the
3184 OSS's /etc/modules.conf or /etc/modprobe.conf.
3188 Bugzilla : 6146, 9635, 9895
3189 Description: servers crash with bad pointer in target_handle_connect()
3190 Details : In rare cases when a client is reconnecting it was possible that
3191 the connection request was the last reference for that export.
3192 We would temporarily drop the export reference and get a new
3193 one, but this may have been the last reference and the export
3194 was just destroyed. Get new reference before dropping old one.
3196 Severity : enhancement
3197 Frequency : if client is started with failover MDS
3199 Description: Allow multiple MDS hostnames in the mount command
3200 Details : Try to read the configuration from all specified MDS
3201 hostnames during a client mount in case the "primary"
3204 Severity : enhancement
3206 Description: Stop sending data to evicted clients as soon as possible.
3207 Details : Check if the client we are about to send or are sending data to
3208 was evicted already. (Check is done every second of waiting,
3209 for which l_wait_event interface was extended to allow checking
3210 of exit condition at specified intervals).
3213 Frequency : rare, normally only when NFS exporting is done from client
3215 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3217 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3220 Frequency : always, if srand() is called before liblustre initialization
3222 Description: Liblustre uses system PRNG disturbing its usage by user application
3223 Details : Introduce internal to lustre fast and high-quality PRNG for
3224 lustre usage and make liblustre and some other places in generic
3225 lustre code to use it.
3227 Severity : enhancement
3228 Bugzilla : 9477, 9557, 9870
3229 Description: Verify that the MDS configuration logs are updated when xml is
3230 Details : Check if the .xml configuration logs are newer than the config
3231 logs stored on the MDS and report an error if this is the case.
3232 Request --write-conf, or allow starting with --old_conf.
3234 Severity : enhancement
3236 Description: Handle symlinks in the path when checking if Lustre is mounted.
3237 Details : Resolve intermediate symlinks when checking if a client has
3238 mounted a filesystem to avoid duplicate client mounts.
3243 Description: lconf can hit an error exception but still return success.
3244 Details : The lconf command catches the Command error exception at the top
3245 level script context and will exit with the associated exit
3246 status, but doesn't ensure that this exit status is non-zero.
3251 Description: failure of ptlrpc thread startup can cause oops
3252 Details : Starting a ptlrpc service thread can fail if there are a large
3253 number of threads or the server memory is very fragmented.
3254 Handle this without oopsing.
3257 Frequency : always, only if liblustre and non-default acceptor port was used
3259 Description: liblustre cannot connect to servers with non-default acceptor port
3260 Details : tcpnal_set_default_params() was not called and was therefore
3261 ignoring the environment varaible TCPNAL_PORT, as well as other
3262 TCPNAL_ environment variables
3267 Description: two objects could be created on the same OST for a single file
3268 Details : If an OST is down, in some cases it was possible to create two
3269 objects on a single OST for a single file. No problems other
3270 than potential performance impact and spurious error messages.
3274 Bugzilla : 5681, 9562
3275 Description: Client may oops in ll_unhash_aliases
3276 Details : Client dcache may become inconsistent in race condition.
3277 In some cases "getcwd" can fail if the current directory is
3283 Description: Inode refcounting problems in NFS export code
3284 Details : link_raw functions used to call d_instantiate without obtaining
3285 extra inode reference first.
3289 Bugzilla : 9942, 9903
3290 Description: Referencing freed requests leading to crash, memleaks with NFS.
3291 Details : We used to require that call to ll_revalidate_it was always
3292 followed by ll_lookup_it. Also with revalidate_special() it is
3293 possible to call ll_revalidate_it() twice for the same dentry
3294 even if first occurence returned success. This fix changes semantic
3295 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3296 reference on a request referred from the intent.
3297 ll_intent_release() then releases such a request.
3300 Frequency : rare, normally benchmark loads only
3302 Description: unlinked inodes were kept in memory on the client
3303 Details : If a client is repeatedly creating and unlinking files it
3304 can accumulate a lot of stale inodes in the inode slab cache.
3305 If there is no other client load running this can cause the
3306 client node to run out of memory. Instead flush old inodes
3307 from client cache that have the same inode number as a new inode.
3310 Frequency : SLES9 2.6.5 kernel and long filenames only
3311 Bugzilla : 9969, 10379
3312 Description: utime reports stale NFS file handle
3313 Details : SLES9 uses out-of-dentry names in some cases, which confused
3314 the lustre dentry revalidation. Change it to always use the
3318 Frequency : rare, unless heavy write-truncate concurrency is continuous
3319 Bugzilla : 4180, 6984, 7171, 9963, 9331
3320 Description: OST becomes very slow and/or deadlocked during object unlink
3321 Details : filter_destroy() was holding onto the parent directory lock
3322 while truncating+unlinking objects. For very large objects this
3323 may block other threads for a long time and slow overall OST
3324 responsiveness. It may also be possible to get a lock ordering
3325 deadlock in this case, or run out of journal credits because of
3326 the combined truncate+unlink. Solution is to do object truncate
3327 first in one transaction without parent lock, and then do the
3328 final unlink in a new transaction with the parent lock. This
3329 reduces the lock hold time dramatically.
3332 Frequency : rare, 2.4 kernels only
3334 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3335 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3336 before unmounting their filesystem in order to avoid deadlock.
3337 The kernel_locked() function in 2.4 kernels only checks whether
3338 the kernel lock is held, not whether it is this process that is
3339 holding it as 2.6 kernels do.
3344 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3345 Details : The client ptlrpc code may be trying to reconnect to a down
3346 server before a previous connection attempt has timed out.
3347 Increase the reconnect interval to be longer than the connection
3348 timeout interval to avoid sending duplicate connections to
3352 Frequency : echo_client brw_test command
3354 Description: fix echo_client to work with OST preallocated code
3355 Details : OST preallocation code (5137) didn't take echo_client IO path
3356 into account: echo_client calls filter methods outside of any
3357 OST thread and, hence, there is no per-thread preallocated
3358 pages and buffers to use. Solution: hijack pga pages for IO. As
3359 a byproduct, this avoids unnecessary data copying.
3363 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3364 Description: Client can oops in mdc_commit_close() after open replay
3365 Details : It was possible for the MDS to return an open request with no
3366 transaction number in mds_finish_transno() if the client was
3367 evicted, but without actually returning an error. Clients
3368 would later try to replay that open and may trip an assertion
3369 Simplify the client close codepath, and always return an error
3370 from the MDS in case the open is not successful.
3373 Frequency : rare, 2.6 OSTs only
3375 Description: OST may deadlock under high load on fragmented files
3376 Details : If there was a heavy load and highly-fragmented OST filesystems
3377 it was possible to have all the OST threads deadlock waiting on
3378 allocation of biovecs, because the biovecs were not released
3379 until the entire RPC IO was completed. Instead, release biovecs
3380 as soon as they are complete to ensure forward IO progress.
3382 Severity : enhancement
3384 Description: Support for specifying external journal device at mount
3385 Details : If an OST or MDS device is formatted with an external journal
3386 device, this device major/minor is stored in the ext3 superblock
3387 and may not be valid for failover. Allow detecting and
3388 specifying the external journal at mount time.
3393 Description: Mounting an MDS with pending unlinked files may cause oops
3394 Details : target_finish_recovery() calls mds_postrecov() which returned
3395 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3396 considers this an error and immediately begins cleaning up the
3397 lov, just after starting the mds_lov process
3399 Severity : enhancement
3401 Description: Implement 'lfs df' to report actual free space on per-OST basis
3402 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3403 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3404 usage of objects; '-h' to report in human readable format.
3406 ------------------------------------------------------------------------------
3408 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3415 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3416 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3417 at startup during orphan unlinking. Instead, skip buddy update
3418 before setup, as it will be regenerated after recovery is complete.
3423 Description: If an OST is inactive, its locks might reference stale inodes.
3424 Details : lov_change_cbdata() must iterate over all namespaces, even if
3425 they are inactive to clear inode references from the lock.
3427 Severity : enhancement
3428 Frequency : occasional, if non-standard max_dirty_mb used
3430 Description: Client will block write RPCs if not enough grant
3431 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3432 then the client will block writes while waiting for another RPC
3433 to complete instead of consuming its dirty limit. With change
3434 we get improved performance when max_dirty_mb is small.
3436 Severity : enhancement
3437 Bugzilla : 3389, 6253
3438 Description: Add support for supplementary groups on the MDS.
3439 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3440 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3441 lookups for user supplementary groups into a cache.
3445 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3446 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3447 for liblustre O_EXCL is enforced by MDS.
3451 Description: readdir never returns NULL in liblustre.
3452 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3453 dentry in struct dirent.
3457 Frequency : liblustre only, depends on application IO pattern
3458 Description: liblustre clients evicted if not contacting servers
3459 Details : Don't put liblustre clients into the ping_evictor list, so
3460 they will not be evicted by the pinger ever.
3462 Severity : enhancement
3464 Description: Add ability to evict clients by NID from MDS.
3465 Details : By echoing "nid:$NID" string into
3466 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3467 $NID would be instantly evicted from this MDS and from all active
3468 OSTs connected to it.
3472 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3473 Details : lookup_it_finish() used to query file size from OSTs that was not
3478 Description: service threads change working directory to that of init
3479 Details : Starting lustre service threads may pin the working directory
3480 of the parent thread, making that filesystem busy. Threads
3481 now change to the working directory of init to avoid this.
3485 Frequency : during shutdown only
3486 Description: shutdown with a failed MDS or OST can cause unmount to hang
3487 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3492 Frequency : 2.6 only
3493 Description: chmod/chown may include an extra supplementary group
3494 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3495 supplementary group and if none is specified this is used.
3498 Bugzilla : 5479 (6816)
3500 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3501 Details : If lookup is in progress on a file that is unlinked we might try
3502 to revalidate the inode and fail in revalidate after lookup is
3503 complete and ll_file_open() enqueues the open again but
3504 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3507 Frequency : always, if lconf --abort_recovery used
3509 Description: lconf --abort_recovery fails with 'Operation not supported'
3510 Details : lconf was attempting to abort recovery on the MDT device and not
3513 ------------------------------------------------------------------------------
3515 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3520 Frequency : rare (only unsupported configurations with a node running as an
3522 Bugzilla : 6514, 5137
3523 Description: Mounting a Lustre file system on a node running as an OST could
3525 Details : OSTs now preallocates memory needed to write out data at
3526 startup, instead of when needed, to avoid having to
3527 allocate memory in possibly low memory situations.
3528 Specifically, if the file system is mounted on on OST,
3529 memory pressure could force it to try to write out data,
3530 which it needed to allocate memory to do. Due to the low
3531 memory, it would be unable to do so and the node would
3532 become unresponsive.
3534 Severity : enhancement
3536 Description: Addition of lconf --service command line option
3537 Details : lconf now accepts a '--service <arg>' option, which is
3538 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3540 Severity : enhancement
3542 Description: Failover mode is now the default for OSTs.
3543 Details : By default, OSTs will now run in failover mode. To return to
3544 the old behaviour, add '--failout' to the lmc line for OSTs.
3546 Severity : enhancement
3548 Description: Health checks are now provided for MDS and OSTs
3549 Details : Additional detailed health check information on MSD and OSTs
3550 is now provided through the procfs health_check value.
3553 Frequency : occasional, depends on IO load
3555 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3556 after numerous create/delete cycles
3557 Details : The ext3 inode allocation policy would not allocate new inodes
3558 very well on the OSTs because there are no new directories
3559 being created. Instead we look for groups with free space if
3560 the parent directories are nearly full.
3565 Description: Network or server problems during mount may cause partially
3566 mounted clients instead of returning an error.
3567 Details : The config llog parsing code may overwrite the error return
3568 code during mount error handling, returning success instead
3574 Description: MDS can fail to allocate large reply buffers
3575 Details : After long uptimes the MDS can fail to allocate large reply
3576 buffers (e.g. zconf client mount config records) due to memory
3577 fragmentation or consumption by the buffer cache. Preallocate
3578 some large reply buffers so that these replies can be sent even
3579 under memory pressure.
3583 Frequency : rare (liblustre)
3584 Description: fsx running with liblustre complained that using truncate() to
3585 extend the file doesn't work. This patch corrects that issue.
3586 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3587 ATTR_SIZE and lsm use in llu_setattr_raw.
3591 Frequency : rare, only 2.6 kernels
3592 Description: Unusual file access patterns on the MDS may result in inode
3593 data being lost in very rare circumstances.
3594 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3595 code in the 2.6 kernel caused the inode and/or EA data not to
3596 be read from disk, causing single-file corruption.
3600 Frequency : rare, only 2.6 filesystems using extents
3601 Description: Heavy concurrent write and delete load may cause data corruption.
3602 Details : It was possible under high-load situations to have an extent
3603 metadata block in the block device cache from a just-unlinked
3604 file overwrite a newly-allocated data block. We now unmap any
3605 metadata buffers that alias just-allocated data blocks.
3609 Frequency : filesystems with default stripe_count larger than 77
3610 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3611 Details : lconf specifies an inode size of 4096 bytes when the default
3612 stripe_count is larger than 77. This conflicts with the default
3613 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3614 this case to avoid pinning too much memory for large EAs.
3616 ------------------------------------------------------------------------------
3618 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3623 Frequency : rare (extremely heavy IO load with hundreds of clients)
3625 Description: Client is evicted, gets IO error writing to file
3626 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3627 caused clients to be evicted for AST timeouts. The fixes in
3628 bug 5192 mean we no longer need to have such short AST timeouts
3629 so ldlm_timeout has been increased.
3632 Frequency : occasional during --force or --failover shutdown under load
3633 Bugzilla : 5949, 4834
3634 Description: Server oops/LBUG if stopped with --force or --failover under load
3635 Details : a collection of import/export refcount and cleanup ordering
3636 issues fixed for safer force cleanup
3639 Frequency : only filesystems larger than 120 OSTs
3640 Bugzilla : 5990, 6223
3641 Description: lfs getstripe would oops on a very large filesystem
3642 Details : lov_getconfig used kfree on vmalloc'd memory
3645 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3646 Bugzilla : 6242, 6243
3647 Description: reading from files that had been truncated to a non-zero size
3648 but never opened returned no data
3649 Details : ll_file_read() reads zeros from no-object files to EOF
3654 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3655 Details : There's little harm in aborting MDS/OSS recovery and letting it
3656 try again, so I removed the LASSERT and return an error instead.
3658 Severity : enhancement
3660 Description: New debugging infrastructure for tracking down data corruption
3661 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3662 (b) cover more of the client-side code path, and (c) try to narrow
3663 down where problems occurred
3667 Bugzilla : 3819, 4364, 4397, 6313
3668 Description: Racing close and eviction MDS could cause assertion in mds_close
3669 Details : It was possible to get multiple mfd references during close and
3670 client eviction, leading to one thread referencing a freed mfd.
3672 Severity: : enhancement
3673 Bugzilla : 3262, 6359
3674 Description: Attempts to reconnect to servers are now more aggressive.
3675 Details : This builds on the enhanced upcall-less recovery that was added
3676 in 1.4.2. When trying to reconnect to servers, clients will
3677 now try each server in the failover group every 10 seconds. By
3678 default, clients would previously try one server every 25 seconds.
3683 Description: After recovery, certain operations trigger a failed
3684 assertion on a client.
3685 Details : Failing over an mds, using lconf -d --failover, while a
3686 client was doing a readdir() call would cause the client to
3687 LBUG after recovery completed and the readdir() was resent.
3689 Severity : enhancement
3691 Description: Default groups are now added by lconf
3692 Details : You can now run lconf --group <servicename> without having to
3693 manually add groups with lmc.
3696 Frequency : occasional
3698 Description: Nodes with an elan id of 0 trigger a failed assertion
3701 Frequency : always when accessing e.g. tty/console device nodes
3703 Description: tty and some other devices nodes cannot be used on lustre
3704 Details : file's private_data field is used by device data and lustre
3705 values in there got lost. New field was added to struct file to
3706 store fs-specific private data.
3709 Frequency : when exporting Lustre via NFS
3711 Description: NFSD failed occasionally when looking up a path component
3712 Details : NFSD is looking up ".." which was broken in ext3 directories
3713 that had grown large enough to become hashed.
3716 Frequency : Clusters with multiple interfaces not on the same subnet
3718 Description: Nodes will repeatedly try to reconnect to an interface which it
3719 cannot reach and report an error to the log.
3720 Details : Extra peer list entries will be created by lconf with some peers
3721 unreachable. lconf now validates the peer before adding it.
3724 Frequency : Only if a default stripe is set on the filesystem root.
3726 Description: Setting a default stripe on the filesystem root prevented the
3727 filesystem from being remounted.
3728 Details : The client was sending extra request flags in the root getattr
3729 request and did not allocate a reply buffer for the dir EA.
3732 Frequency : occasional, higher if lots of files are accessed by one client
3733 Bugzilla : 6159, 6097
3734 Description: Client trips assertion regarding lsm mismatch/magic
3735 Details : While revalidating inodes the VFS looks up inodes with ifind()
3736 and in rare cases can find an inode that is being freed.
3737 The ll_test_inode() code will free the lsm during ifind()
3738 when it finds an existing inode and then the VFS later attaches
3739 this free lsm to a new inode.
3743 Bugzilla : 6422, 7030
3744 Description: MDS deadlock between mkdir and client eviction
3745 Details : Creating a new file via mkdir or mknod (starting a transaction
3746 and getting the ns lock) can deadlock with client eviction
3747 (gets ns lock and trying to finish a synchronous transaction).
3750 Frequency : occasional
3751 Description: While starting a server, the fsfilt_ext3 module could not be
3753 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3754 kernels. Previously, lconf would still use the ext3 name
3755 when trying to load modules. Now, it will correctly use
3756 ext3 on 2.4 and ldiskfs on 2.6.
3758 Severity : enhancement
3759 Description: The default stripe count has been changed to 1
3760 Details : The interpretation of the default stripe count (0, to lfs
3761 or lmc) has been changed to mean striping across a single
3762 OST, rather than all available. For general usage we have
3763 found a stripe count of 1 or 2 works best.
3765 Severity : enhancement
3766 Description: Add support for compiling against Cray portals.
3767 Details : Conditional compiling for some areas that are different
3771 Frequency : occasional
3772 Bugzilla : 6409, 6834
3773 Description: Creating files with an explicit stripe count may lead to
3774 a failed assertion on the MDS
3775 Details : If some OSTs are full or unavailable, creating files may
3776 trigger a failed assertion on the MDS. Now, Lustre will
3777 try to use other servers or return an error to the
3781 Frequency : occasional
3783 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3784 caused lock timeout during readahead (since 1.4.2).
3785 Details : Processes doing readahead might match a lock that hasn't been
3786 granted yet if there are overlapping and conflicting lock
3787 requests. The readahead process waits on ungranted lock
3788 (original lock is CBPENDING), while OST waits for that process
3789 to cancel CBPENDING read lock and eventually evicts client.
3791 Severity : enhancement
3793 Description: Initial enabling of flock support for clients
3794 Details : Implements fcntl advisory locking and file status functions.
3795 This feature is provided as an optional mount flag (default
3796 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3797 locking are implemented yet, and those that are are not guaranteed
3798 to be completely correct in production environments.
3799 mount -t lustre -o [flock|noflock] ...
3802 Frequency : occasional
3804 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3805 assertion in the ext3 JBD (journaling) layer.
3806 Details : The b_committed_data struct is protected by the big kernel lock
3807 in 2.4 kernels, serializing journal_commit_transaction() and
3808 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3809 access to this struct. In 2.6 kernels there is finer grained
3810 locking to improve SMP performance of the JBD layer.
3814 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3819 Description: merge and cleanup kernel patches.
3822 -----------------------------------------------------------------------------
3824 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3826 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3827 versions of Lustre. Please update all servers and clients to
3828 version 1.4.2 or later at the same time. You must also run
3829 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3830 to update the configuration logs.
3832 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3833 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3834 - swab llog records of type '0' so we get proper header size/idx (5861)
3835 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3836 - fix rename of one directory over another leaking an inode (5953)
3837 - avoid SetPageDirty on 2.6 (5981)
3838 - don't re-add just-being-destroyed locks to the waiting list (5653)
3839 - when creating new directories, inherit the parent's custom
3840 striping settings if present parent (3048)
3841 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3842 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3843 - don't hold client locks on temporary worklist from l_lru (5666)
3844 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3845 - automatically evict dead clients (5921)
3846 - Update file size properly in create+truncate+fstat case (6196)
3847 - Do not unhash mountpoint dentries, do not allow removal of
3849 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3850 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3851 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3852 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3853 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3854 - Correctly handle reads of files with no objects (6243)
3855 - lctl recover will also mark a device active if deactivate used (5933)
3857 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3858 - allow --write-conf on an MDS with different nettype than client (5619)
3859 - don't write config llogs to MDS for mounts not from that MDS (5617)
3860 - lconf should create multiple TCP connections from a client (5201)
3861 - init scripts are now turned off by default; run chkconfig --on
3862 lustre and chkconfig --on lustrefs to use them
3863 - upcalls are no longer needed for clients to recover to failover
3865 - add --abort-recovery option to lconf to abort recovery on device
3867 - add support for an arbitrary number of OSTs (3026)
3868 - Quota support protocol changes.
3869 - forward compatibility changes to wire structs (6007)
3870 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3871 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3872 - improved "lustre status" script
3873 - initialize blocksize for non-regular files (6062)
3874 - added --disable-server and --disable-client configure options (5782)
3875 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3876 - Vanilla 2.4.29 support
3877 - increase maximum number of obd devices to 520 (6242)
3878 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3879 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3880 SLES 9 2.6.5 kernels
3881 - limit stripes per file to 160 (the maximum EA size) (6093)
3883 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3886 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3887 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3888 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3889 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3890 - small ext3 extents cleanups and fixes (5733)
3891 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3892 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3893 - avoid needless client->OST connect, fix handle mismatch (5317)
3894 - fix DLM error path that led to out-of-sync client, long delays (5779)
3895 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3896 - fix several locking issues related to i_size (5492,5624,5654,5672)
3897 - don't move pending lock onto export if it is already evicted (5683)
3898 - fix kernel oops when creating .foo in unlinked directory (5548)
3899 - fix deadlock in obdfilter statistics vs. object create (5811)
3900 - use time_{before,after} to avoid timer jiffies wrap (5882)
3901 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3902 - Do not leak request if server was not able to process it (5154)
3903 - If mds_open unable to find parent dir, make that negative lookup(5154)
3904 - don't create new directories with extent-mapping (5909, 5936)
3906 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3907 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3908 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3909 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3910 - support for RHEL4 kernel on i686 (5773)
3911 - provide error messages when incompatible logs are encountered (5898)
3913 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3914 * version 1.4.0.10 (1.4.1 release candidate 1)
3916 - don't keep a lock reference when lock is not granted (4238)
3917 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3918 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3919 - drop import inflight refcount on signal_completed_replay error (5255)
3920 - unlock page after async write error during send (3677)
3921 - handle missing objects in filter_preprw_read properly (5265)
3922 - no transno return for symlink open, don't save no-trasno open (3440)
3923 - don't try to complete elan receive that already failed (4012)
3924 - free RPC server reply state on error (5406)
3925 - clean up thread from ptlrpc_start_thread() on error (5160)
3926 - readahead could read extra page into cache that wasn't ejected (5388)
3927 - prevent races in class_attach/setup/cleanup/detach (5260)
3928 - don't dereference de->d_inode after l_dput of de (5458)
3929 - use "int" for stripe value returned from lock_to_stripe (5544)
3930 - mballoc allocation and error-checking fixes in 2.6 (5504)
3931 - block device patches to fix I/O request sizes in 2.6 (5482)
3932 - look up hostnames for IB nals (5602)
3933 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3934 - don't start multiple acceptors for the same port (5277)
3935 - fix incorrect LASSERT in mds_getattr_name (5635)
3936 - export a proc file for general "ping" checking (5628)
3937 - fix "lfs check" to not block when the MDS is down (5628)
3939 - service request history (4965)
3940 - put {ll,lov,osc}_async_page structs in a single slab (4699)
3941 - create an "evict_client" /proc entry on OSTs, like the MDS has
3942 - fix mount usage message, return errors per mount(8) (5168)
3943 - change grep [] to grep "[]" in tests so they work in more UMLs
3944 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
3945 - remove ancient LOV_MAGIC_V0 EA support (5047)
3946 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
3947 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
3948 - allow readahead allocations to fail when low on memory (5383)
3949 - mmap locking landed again, after considerable improvement (2828)
3950 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
3952 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
3955 - send OST transaction number in read/write reply to free req (4966)
3956 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3957 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3958 - the watchdog thread now runs as interruptible (5246)
3959 - flock/lockf fixes (but it's still disabled, pending 5135)
3960 - don't use EXT3 constants in llite code (5094)
3961 - memory shortage at startup could cause assertion (5176)
3963 - reorganization of lov code
3964 - single portals codebase
3966 - add extents/mballoc support (5025)
3967 - direct I/O reads in the obdfilter (4048)
3968 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
3970 tbd Cluster File Systems, Inc. <info@clusterfs.com>
3973 - send OST transaction number in read/write reply to free req (4966)
3974 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
3975 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
3976 - the watchdog thread now runs as interruptible (5246)
3977 - handle missing objects in filter_preprw_read properly (5265)
3978 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3979 - drop import inflight refcount on signal_completed_replay error (5255)
3980 - unlock page after async write error during send (3677)
3981 - return original error code on reconstructed replies (3761)
3982 - no transno return for symlink open, don't save no-trasno open (3440)
3984 - add pid to ldlm debugging output (4922)
3985 - bump the watchdog timeouts -- we can't handle 30sec yet
3986 - extra debugging for orphan dentry/inode bug (5259)
3988 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
3991 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
3992 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
3993 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
3994 - fix NULL dereference in /proc/sys/portals/routes (4827)
3995 - allow failed mdc_close() operations to be interrupted (4561)
3996 - stop precreate on OST before MDS would time out on it (4778)
3997 - don't send partial-page writes before EOF from client (4410)
3998 - discard client grant for sub-page writes on large-page clients (4520)
3999 - don't free dentries not owned by NFS code, check generation (4806)
4000 - fix lsm leak if mds_create_objects() fails (4801)
4001 - limit debug_daemon file size, always print CERROR messages (4789)
4002 - use transno after validating reply (3892)
4003 - process timed out requests if import state changes (3754)
4004 - update mtime on OST during writes, return in glimpse (4829)
4005 - add mkfsoptions to LDAP (4679)
4006 - use ->max_readahead method instead of zapping global ra (5039)
4007 - don't interrupt __l_wait_event() during strace
4009 - add software watchdogs to catch hung threads quickly (4941)
4010 - make lustrefs init script start after nfs is mounted
4011 - fix CWARN/ERROR duplication (4930)
4012 - return async write errors to application if possible (2248)
4013 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4014 - print NAL number in %x format (4645)
4015 - update barely-supported suse-2.4.21-171 series (4842)
4016 - support for sles 9 %post scripts
4017 - support for building 2.6 kernel-source packages
4018 - support for sles km_* packages
4020 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4023 - ignore -ENOENT errors in osc_destroy (3639)
4024 - notify osc create thread that OSC is being cleaned up (4600)
4025 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4026 - reconstruct ost_handle() like mds_handle() (4657)
4027 - create a new thread to do import eviction to avoid deadlock (3969)
4028 - let lconf resolve symlinked-to devices (4629)
4029 - don't unlink "objects" from directory with default EA (4554)
4030 - hold socknal file ref over connect in case target is down (4394)
4031 - allow more than 32000 subdirectories in a single directory (3244)
4032 - fix blocks count for O_DIRECT writes (3751)
4033 - OST returns ENOSPC from object create when no space left (4539)
4034 - don't send truncate RPC if file size isn't changing (4410)
4035 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4036 - bind to privileged port in socknal and tcpnal (3689)
4038 - rate limit CERROR/CWARN console message to avoid overload (4519)
4039 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4040 - basic mmap support (3918)
4041 - kernel patch series update from b1_4 (4711)
4043 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4046 - avoid crash during MDS cleanup with OST shut down (2775)
4047 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4048 - don't use bad inodes on the MDS (3744)
4049 - dynamic object preallocation to improve recovery speed (4236)
4050 - don't hold spinlock over lock dumping or change debug flags (4401)
4051 - don't zero obd_dev when it is force cleaned (3651)
4052 - print grants to console if they go negative (4431)
4053 - "lctl deactivate" will stop automatic recovery attempts (3406)
4054 - look for existing locks in ldlm_handle_enqueue() (3764)
4055 - don't resolve lock handle twice in recovery avoiding race (4401)
4056 - revalidate should check working dir is a directory (4134)
4058 - don't always mark "slow" obdfilter messages as errors (4418)
4060 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4063 - don't close LustreDB during write_conf until it is done (3860)
4064 - fix typo in lconf for_each_profile (3821)
4065 - allow dumping logs from multiple threads at one time (3820)
4066 - don't allow multiple threads in OSC recovery (3812)
4067 - fix debug_size parameters (3864)
4068 - fix mds_postrecov to initialize import for llog ctxt (3121)
4069 - replace config semaphore with spinlock (3306)
4070 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4071 - don't allow enqueue to complete on a destroyed export (3822)
4072 - down write_lock before checking llog header bitmap (3825)
4073 - recover from lock replay timeout (3764)
4074 - up llog sem before sending rpc (3652)
4075 - reduce ns lock hold times when setting kms (3267)
4076 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4077 - fix NULL deref and obd_dev leak on setup error (3312)
4078 - replace some LBUG about llog ops with error handling (3841)
4079 - don't match INVALID dentries from d_lookup and spin (3784)
4080 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4081 - fix invalid assertion in ptlrpc_set_wait (3880)
4083 - add libwrap support for the TCP acceptor (3996)
4084 - add /proc/sys/portals/routes for non-root route listing (3994)
4085 - allow setting MDS UUID in .xml (2580)
4086 - print the stack of a process that LBUGs (4228)
4088 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4091 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4092 - make sure to unset replay flag from failed open requests (3440)
4093 - if default stripe count is 0, use OST count for inode size (3636)
4094 - update parent mtime/ctime on client for create/unlink (2611)
4095 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4096 - free recovery state on server during a forced cleanup (3571)
4097 - unregister_reply for resent reqs (3063)
4098 - loop back devices mounting and status check on 2.6 (3563)
4099 - fix resource-creation race that can provoke i_size == 0 (3513)
4100 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4101 - more debugging for page-accounting assertion (3746)
4102 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4103 - avoid deadlock after precreation failure (3758)
4104 - fix race and lock order deadlock in orphan handling (3450, 3750)
4105 - add validity checks when grabbing inodes from l_ast_data (3599)
4107 - add /proc/.../recovery_status to obdfilter (3428)
4108 - lightweight CDEBUG infrastructure, debug daemon (3668)
4109 - change default OSC RPC parameters to be better on small clusters
4110 - turn off OST read cache for files smaller than 32MB
4111 - install man pages and include them in rpms (3100)
4112 - add new init script for (un)mounting lustre filesystems (2593)
4113 - run chkconfig in %post for init scripts (3701)
4114 - drop scimac NAL (unmaintained)
4116 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4119 - clean kiobufs before and after use (3485)
4120 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4121 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4122 - change init script to fail more clearly if not run as root (1528)
4123 - allow clients to reconnect during replay (1742)
4124 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4125 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4126 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4127 - fix cond_resched (3554)
4128 - only evict unfinished clients after recovery (3515)
4129 - allow bulk resend, prevent data loss (3570)
4130 - dynamic ptlrpc request buffer allocation (2102)
4131 - don't allow unlinking open directory if it isn't empty (2904)
4132 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4133 - remove extraneous obd dereference causing LASSERT failure (3334)
4134 - don't use get_cycles() when creating temp. files on the mds (3156)
4135 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4136 - handle EEXIST for set-stripe, set proper directory name (3336)
4138 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4139 - fix ksocknal_fmb_callback() error messages (2918)
4141 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4144 - don't copy lvb into (possibly NULL) reply on error (2983)
4145 - don't deref dentry after dput, don't free lvb on error (2922)
4146 - use the kms to determine writeback rpc length (2947)
4147 - increment oti_logcookies when osc is inactive (2948)
4148 - update client's i_blocks count via lvb messages (2543)
4149 - handle intent open/close of special files properly (1557)
4150 - mount MDS with errors=remount-ro, like obdfilter (2009)
4151 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4152 - don't use cancelling-locks' kms values (2947)
4153 - use highest lock extent for kms, not last one (2925)
4154 - don't dereference ERR_PTR() dentry in error handling path (3107)
4155 - fix thread race in portals_debug_dumplog() (3122)
4156 - create lprocfs device entries at setup instead of at attach (1519)
4157 - common AST error handler, don't evict client on completion race (3145)
4158 - zero nameidata in detach_mnt in 2.6 (3118)
4159 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4160 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4161 - initialize RPC timeout timer earlier for 2.6 (3219)
4162 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4163 - print nal/nid for unknown nid (3258)
4164 - additional checks for oscc recovery before doing precreate (3284)
4165 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4166 - don't crash in mdc_close for bad permissions on open (3285)
4167 - zero i_rdev for non-device files (3147)
4168 - clear page->private before handing to FS, better assertion (3119)
4169 - tune the read pipeline (3236)
4170 - fix incorrect decref of invalidated dentry (2350)
4171 - provide read-ahead stats and refine rpc in flight stats (3328)
4172 - don't hold journal transaction open across create RPC (3313)
4173 - update atime on MDS at close time (3265)
4174 - close LDAP connection when recovering to avoid server load (3315)
4175 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4176 - don't leak open file on MDS after open resend (3325)
4177 - serialize filter_precreate and filter_destroy_precreated (3329)
4178 - loop device shouldn't call sync_dev() for nul device (3092)
4179 - clear page cache after eviction (2766)
4180 - resynchronize MDS->OST in background (2824)
4181 - refuse to mount the same filesystem twice on same mountpoint (3394)
4182 - allow llmount to create routes for mounting behind routers (3320)
4183 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4184 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4185 - fix rare problem with rename on htree directories (3417)
4187 - allow default OST striping configuration per directory (1414)
4188 - fix compilation for qswnal for 2.6 kernels (3125)
4189 - increase maximum number of MDS request buffers for large systems
4190 - change liblustreapi to be useful for external progs like lfsck (3098)
4191 - increase local configuration timeout for slow disks (3353)
4192 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4194 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4197 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4198 - don't overwrite extent policy data in reply if lock was blocked (2901)
4199 - drop filter export grants atomically with removal from device (2663)
4200 - del obd_self_export from work_list in class_disconnect_exports (2908)
4201 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4202 - swab reply message in mdc_close, other PPC fixes (2464)
4203 - fix destroying of named logs (2325)
4204 - overwrite old logs when running lconf --write_conf (2264)
4205 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4206 - fix race in target_handle_connect (2898)
4207 - mds_reint_create() should take same inode create lock (2926)
4208 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4209 - don't close files for self_export to avoid uninitialized obd (2936)
4210 - allow MDS with the same name as client node (2939)
4211 - hold dentry reference for closed log files for unlink (2325)
4212 - reserve space for all logs during transactions (2059)
4213 - don't evict page beyond end of stripe extent (2925)
4214 - don't oops on a deleted current working directory (2399)
4215 - handle hard links to targets without a parent properly (2517)
4216 - don't dereference NULL lock when racing during eviction (2867)
4217 - don't grow lock extents when lots of conflicting locks (2919)
4219 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4222 - account for cache space usage on clients to avoid data loss (974)
4223 - lfsck support in lustre kernel code (2349)
4224 - reduce journal credits needed for BRW writes (2370)
4225 - orphan handling to avoid losing space on client/server crashes
4226 - ptlrpcd can be blocked, stopping ALL progress (2477)
4227 - use lock value blocks to assist in proper KMS, faster stat (1021)
4228 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4229 - recovery for initial connections (2355)
4230 - fixes for mds_cleanup_orphans (1934)
4231 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4232 - block all file creations until orphan recovery completes (1901)
4233 - client remove rq_connection from request struct (2423)
4234 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4235 - recovery timer race (2670)
4236 - mdc_close recovey bug (2532)
4237 - ptlrpc cleanup bug (2710)
4238 - mds timeout on local locks (2588)
4239 - namespace lock held during RPCs (2431)
4240 - handle interrupted sync write properly (2503)
4241 - don't try to handle a message that hasn't been replied to (2699)
4242 - client assert failure during cleanup after abort recovery (2701)
4243 - leak mdc device after failed mount (2712)
4244 - ptlrpc_check_set allows timedout requests to complete (2714)
4245 - wait for inflight reqs when ptlrpcd finishes (2710)
4246 - make sure unregistered services are removed from the srv_list
4247 - reset bulk XID's when resending them (caught by 1138 test)
4248 - unregister_bulk after timeout
4249 - fix lconf error (2694)
4250 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4251 - readahead locks pages, leaves pending causing memory pressure (2673)
4252 - increase OST request buffers to 4096 on large machines (2729)
4253 - fix up permission of existing directories in simple_mkdir (2661)
4254 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4255 - don't assign transno to errored transactions (2742)
4256 - don't delete objects on OST if given a bogus objid from MDS (2751)
4257 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4258 - if rq_no_resend, then timeout request after recovery (2432)
4259 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4260 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4261 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4262 - MDS recovery shouldn't delete objects that it has given out (2730)
4263 - if enqueue arrives after completion, don't clobber LVB (2819)
4264 - don't unlock pages twice when trigger_group_io returns error (2814)
4265 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4266 - don't write pages to disk if there was an error (1450)
4267 - don't ping imports that have recovery disabled (2676)
4268 - take buffered bytes into account when balancing socknal conn (2817)
4269 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4270 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4271 - remove little-endian swabbing of llog records (1987)
4272 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4273 - retry reposting request buffers if they fail (1191)
4274 - grow extent at grant time to avoid granting a revoked lock (2809)
4275 - lock revoke doesn't evict page if covered by a second lock (2765)
4276 - disable VM readahead to avoid reading outside lock extents (2805)
4278 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4279 - updated kernel patches for hp-2.4.20 kernel (2681)
4281 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4284 - fix truncated write corruption (2366)
4285 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4287 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4288 - ensure len(uuid) < 37 in lmc (1171)
4289 - fix ia64 OOPS in llog_test (2255)
4290 - zero end of page at obdfilter for partial page writes (2648)
4291 - don't leave stale dentries around after renames (bug 2428)
4292 - fix timeouts when evicting a client with a single lock held (2642)
4293 - set deadline for the initial HELLO message to drain (2634)
4294 - print out dotted-quad IP addresses in the socknal (2302)
4296 - additional debugging for MDS client eviction problem (2443)
4297 - fix mkfsoptions support for osts (2603, 2604)
4299 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4302 - add series for the vanilla 2.6.0 kernel
4303 - add series for the vanilla 2.4.24 kernel
4304 - add series for a cray x86/64 UL kernel drop
4305 - fix xattr patches for the vanilla 2.4.19 series
4307 - generate true UUIDs in lmc (1171)
4308 - have portals stack dumping break in UML (2466)
4309 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4310 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4311 - wake callback waiting threads on client eviction (2460)
4312 - Add --ptldebug and --subsystem to lmc (1719)
4313 - update assertion to allow safe interrupt allocation
4314 - set rq_no_resend for cancel requests (2432)
4315 - recalculate ptlrpcd timeout after resend (2494)
4316 - call vfs_rmdir when removing pending directories (2368)
4317 - fix renaming a file to itself (2429)
4318 - lmc creates a default one-stripe lov (2454)
4319 - expand procfs space to handle large clusters (2326)
4320 - increase UML stack to avoid overflow
4321 - update lconf's list of debug and subsystem masks
4322 - fix lfs find --obd (2510)
4323 - /proc tunable for disabling filter read caching (2591)
4324 - stop rpm packages from altering slapd.conf (2301)
4325 - disable nagle in the socknal under 0conf (2578)
4326 - choose mds inode size based on stripe count (2572)
4327 - fix kernel-source rpm problems (2516)
4329 - add --disable-doc to avoid pdf generation (2421)
4330 - update documentation, tests, type-os, comments
4331 - avoid format warnings on ia64
4332 - remove the TOE NAL
4333 - tiny code cleanups by removing unused fields
4335 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4338 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4339 - remove the most blatant lies from BUILDING file (2371)
4340 - change default debug level to reasonable production setting
4341 - reduce client side cache size to reduce cache flush time
4342 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4343 - make TCP zerocopy and pinger support enabled by default (2476)
4344 - sync writes completed after process exits caused crashes (2319)
4345 - maintain correct mount count on the MDS (2356)
4346 - backout 1557, because 2316 wasn't really fixed
4347 - better file I/O statistics gathering in /proc
4348 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4349 - another kernel patch to fix zero-copy TCP function export
4350 - don't take duplicate lock when processing re-sent getattr (2420)
4351 - lctl uses obd_self_export instead of creating new conn (2353)
4352 - MDS/OST recovery case which requires object creation asserted (2425)
4353 - move lfs from /usr/sbin to /usr/bin in packages
4354 - fix race between mds_client_add and mds_client_free (2417)
4355 - use kmalloc instead of slabs in portals (2430)
4356 - don't create duplicate records when a failover MDS is present (2442)
4357 - remove unnecessary mount age check (2332)
4358 - don't remove directory inodes from locks prematurely (2451)
4359 - don't break if MDS service name is the same as hostname (2103)
4360 - fix races in client write RPC generation when cache full (2482)
4362 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4365 - remove now-unused request->rq_obd (278)
4366 - if an allocation fails, print out how much memory we've used (1933)
4367 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4368 - add the "configurable stack size" patch to most series files (1256)
4369 - ability to write large log records, for 100+ OST configs (2306)
4370 - fix NULL deref when filter_prep fails (2314)
4371 - fix operator precedence error in filter_sync
4372 - dynamic allocation of socknal TX descriptors (2315)
4373 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4374 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4375 - ensure that CWARN messages in Portals always get to the syslog
4376 - __init/__exit are not for prototype decls (ldlm_init/exit)
4377 - x86-64 compile warning fixes
4378 - fix gateway LMC keyword conflict (2318)
4379 - fix MDS lock inversions in getattr/reint paths (1844)
4380 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4381 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4382 - DLM race condition prevented some lock evictions (2328)
4383 - ENOMEM detection and retry on socknal sends (2230)
4384 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4385 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4386 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4387 - better error messages when a client is rejected during recovery (1505)
4388 - avoid cancelling locks which were never granted, after failure (2330)
4389 - fix i_sem/journal inversion in mds_client_add (2333)
4390 - fix truncate/getattr lock cycle deadlock (2334)
4391 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4392 - fix two rare exit paths which could leak an l_lock() ref (2321)
4393 - fencepost error in MDS/OST orphan recovery (2226)
4394 - make log record alignment 8 bytes (1988)
4395 - lstripe now fails when requested offset > ost_count (2237)
4396 - ensure that all kernel series have a complete list.h (1607)
4397 - fix crashes in special-file operations (2316)
4398 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4400 - allow configurable automake binary, for testing new versions
4401 - small update to the lfs documentation
4403 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4405 * fix negative export reference count in fsfilt_sync (2312)
4407 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4408 * release candidate 0.9.1
4410 - orphans are moved into the PENDING directory for possible recovery
4411 - replayed opens now open by fid for orphan/rename safety (1042)
4412 - last close of an orphan inode generates a transno (683)
4413 - chdir() and mount() now pin the directory entry (1020)
4414 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4415 - discard very old requests without processing them (1502)
4416 - remove some common, well-understood CERRORs (1505)
4417 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4418 - clear "grant" flags in OST replies until OST grant code lands (1644)
4419 - fix read performance by not clobbering i_blksize on client (1598)
4420 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4421 - make LRU size a /proc tunable, clears locks when reduced (707)
4422 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4423 - server locks take references on exports now (1558)
4424 - build fixes for 2.4.20-rh trees (1663)
4425 - return an error from lov_create if all OSCs are inactive (1751)
4426 - fix import levels when a reconnect happens without a timeout (1597)
4427 - exit early from mds_open if we get a lookup error (1749)
4428 - partial page read at EOF wouldn't wait for disk before sending (1642)
4429 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4430 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4431 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4432 - write/truncate lock inversion (1639)
4433 - Don't auto-load obdclass, portals modules during cleanup (1495)
4434 - fix timestamps from jumping to "now" (1763)
4435 - extra journal assertions (1648)
4436 - add an extra multiunlink test (1771)
4437 - fix read_record/write_record API (1776)
4438 - fix leak of offset_extent, possible incorrect i_size later (1772)
4439 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4440 - Debugging for the kqswnal_get_idle_tx problems (1820)
4441 - Allow recovery to be attempted multiple times (1536)
4442 - Write out MDS last_rcvd file after it is first created (1600)
4443 - Fix tx_descriptor leak in failed transmit situations (1827)
4444 - ext3 journaling fixes for assertion failure after IO error (1871)
4445 - class_export_put() on freed export after completion AST error (1896)
4446 - Fix revalidate looping in VFS (1322)
4447 - Don't access a freed export during MDS_REINT timeout (1521)
4448 - Add open-unlink recovery support on the MDS (1673,1764)
4449 - Return an error if no MDS data was read from last_rcvd (1946)
4450 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4451 - Don't setup a disk device that doesn't match exported UUID (317)
4452 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4453 - avoid committing NULL handle in force close
4454 - local.sh is now a one-stripe LOV configuration
4455 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4456 - fix ext3 htree duplicate directory entry corruption (1516)
4457 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4458 - update to new LOV EA format (2097)
4459 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4461 2003-06-15 Phil Schwan <phil@clusterfs.com>
4464 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4465 - per-import/export recovery handling (958, 931, 959)
4466 - multiple last-rcvd slots, for serving multiple FSes (949)
4467 - connections are again shared between multiple imp/exports (963, 964)
4468 - "umount -f" would hang if any requests needed to be sent (393, 978)
4469 - avoid pinning large req buffer by copying for queued messages (989)
4470 - add "uuid" to "lctl device" command to help upcalls (991)
4471 - "open" RPCs with transnos would confuse recovery counters (1037)
4472 - do proper endian conversion of all wire messages (288, 340, 891)
4473 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4474 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4475 - fix ptlrpc_connection leak in target_handle_connect (1174)
4476 - fix import refcounting bug in OST and MDS cleanup (1134)
4477 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4478 - fix very unlikely obd_types race condition (501)
4479 - remove osc_open hack for echo_client (1187)
4480 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4481 - a failure in read_inode2 leads to deadlock (1139)
4482 - cancel ack-locks as soon as transaction is committed (1072)
4483 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4484 - make sure to commitrw after any preprw to avoid deadlock (1162)
4485 - failing to execute a file in a lustre FS would lock inode (1203)
4486 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4487 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4488 - obd_uuid2tgt would walk off the end of the list (1255)
4489 - on IA64 the osc would give portals incorrect bulk size (1258)
4490 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4491 - fix lock inversion caused by new llite matching code (1282)
4492 - limit the number of dirty pages on a client to 10MB (1286)
4493 - timed out locks were not being corrected cancelled (1289)
4494 - fix O_DIRECT above 4GB on IA-32 (1292)
4495 * major user-visible changes
4496 - fail out/fail over policy now controlled by the upcall (993)
4498 - add OBD_PING to check server availability and failure (954)
4499 - lustre messages are now sent in sending host order (288, 340, 891)
4500 - add eadatalen to MDS getattr reply (340)
4501 - OST read replies may contain second buffer, with per-page status (593)
4503 2003-03-11 Phil Schwan <phil@clusterfs.com>
4506 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4507 - failed open()s could cause deadlock; fixed (867, 869)
4508 - stop cancelling OST locks when files are closed (481)
4509 - overlapping XID spaces caused network corruption (851, 853)
4510 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4511 - setattr_raw added, to do single-RPC, server-side setattrs
4512 - lmc/lconf syntax change for OST UUIDs
4513 - fix crashy race condition between ptlrpc_free_req and osc_close
4514 - don't use request in mdc_enqueue if we hit a timeout (889)
4515 - don't set the inode i_size for regular files from the MDS (896)
4516 - handle out of order completion AST (842)
4517 - don't LBUG if a lock request times out after receiving AST (913)
4518 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4519 - if a bad lock AST arrives, send an error instead of dropping entirely
4520 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4521 - fix leak in bulk IO when only partially completed (899, 900, 926)
4522 - fix O_DIRECT for ia64 (55)
4523 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4524 - C-z after timeout could hang a process forever; fixed (977)
4526 - client-side I/O cache (678, 924, 929, 941, 970)
4528 - READPAGE and SETATTRs which don't take server-side locks get
4531 2003-02-11 Phil Schwan <phil@clusterfs.com>
4534 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4535 - Fix fsfilt_extN_readpage() to read a full page of directory
4536 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4537 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4538 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4539 - Most intent operations were removed, in favour of a new RPC mode
4540 that does a single RPC to the server and bypasses most of the VFS
4541 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4542 - Aggressively cancel local locks on DLM servers
4543 - mds_reint_unlink sends EA to the client if it's the last nlink.
4544 client uses that EA to unlink OST objects.
4545 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4546 - recursive symlinks were fixed (439)
4547 - fixed NULL deref in DEBUG_REQ
4548 - filter_update_lastobjid no longer calls sync, which annoyed extN
4549 - fixed multi-client small-writes to a single file problem (445)
4550 - fixed mtime updates during file writes (607)
4551 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4552 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4553 - fixed Linux OST size reporting problem (444, 656)
4554 - OST now updates object mtime with writes or setattr (607, 619)
4555 - client verifies file size before zeroing page past EOF (445)
4556 - OST now writes last allocated objid to disk with allocation (108)
4557 - LOV on echo now works (409)
4559 - mds_reint_unlink sends a new buffer, with the EA included. this
4560 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4561 set if a regular file was being unlinked, and it was the last link
4562 - use PtlGet from the target for bulk writes (315)
4563 - OST now updates object mtime with writes or setattr (607, 619)
4564 - LDLM now has a grant-time callback to revalidate locked items, if
4566 - Many MDS operations were reorganized to combat race conditions
4568 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4569 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4571 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4574 - Fully reactivate OST imports after reconnection (512, others)
4575 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4576 - More graceful error handling for truncating on dead OST (515)
4577 - Don't error out unless we're actually accessing dead stripes (474)
4578 - Fix garbage sizes when stripes are missing (410)
4579 - LRU counters were broken, causing constant lock purge (433, 432)
4580 - garbage on read from stripes with failed OSTs (441)
4581 - mark OSCs as active before reconnecting during recovery (438)
4582 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4583 - lfind did not preserve OST order in output (443)
4584 - symlinks cause hung clients, incorrect data (439)
4585 - stop dereferencing request after dropping refcount (457)
4586 - don't LASSERT(spin_is_locked) on non-SMP (455)
4587 - fixes for many rename() bugs
4588 - fstat didn't correctly synchronize attributes (399)
4589 - server must handle lock cancellation during blocking AST prep (487)
4590 - bulk descriptors were free()d too soon (511)
4591 - fix paths in lconf, which would load incorrect modules (451, 507)
4592 - fix confusing lconf 'host not found' error message (386)
4593 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4594 - fix race condition in mdc_blocking_ast() for inode access (526)
4595 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4596 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4597 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4598 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4599 - fix lstripe in several ways (kernel side) (527)
4600 - fix request leak in ldlm_cli_enqueue (262)
4601 - incorrect OSC was marked inactive after OST failure
4602 - call mds_fs_cleanup before unmounting filesystem (524)
4603 - fix races between taking ns_lock and ldlm_lock_change_resource
4604 - fix races updating LOV export open file list
4605 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4606 - fix recovery NULL deref in ldlm_cli_cancel_unused
4607 - fix some DLM races by using new hash table for lock handles (419)
4608 - permit the client to specify desired inodes, at replay
4609 - duplicate requests when we queue them for replay reintegration
4610 - fix last_rcvd offset calculation
4611 - sync after each recovered transaction, so we always make progress
4612 - never, not always, ERESTART requests without transnos
4613 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4615 - skip replay if the MDS says that the client is already connected
4616 - don't check for a recovery-enabled export to match lctl's UUID
4617 - don't INC_USE_COUNT for phantom exports
4618 - don't crash when cleaning up phantom exports (567)
4619 - don't double-finish or set replay data for errored mdc_open requests
4620 - abort requests when they time out, so we don't get old replies
4621 - send/receive replies for AST messages again
4622 - if the client says that it doesn't have the lock, cancel it on the
4624 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4625 mark it as destroyed, it will all work out when decref is called
4626 - fix module use counts (22, 581)
4628 - ASTs now expect a reply (server cancels lock on error reply)
4630 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4633 - fix many simultaneous client startup (392)
4634 - fix dentry->d_it clobbering
4635 - credentials weren't being shipped for readdir/getattr operations
4636 - remove invalid assertions triggered during some concurrent MD
4638 - proper Lustre versions added (336, 389)
4639 - fix memory leak for create error case (398)
4640 - fix LOV locking bug that would get cli/srv out of sync
4641 - fix echo client over LOV (409)
4642 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4643 - fix double-O_EXCL intent crash (424)
4644 - avoid sending multiple lock CANCELs (352)
4646 - MDS can do multi-client recovery (modulo bugs in new code)
4648 - many updates, edits, cleanups
4650 2002-11-18 Phil Schwan <phil@clusterfs.com>
4653 - fix null d_it dereference (346)
4654 - fix full OST/dbench hang (333)
4655 - fix permission problem with file removal (286)
4656 - fix removal of OSCs from LOV when they fail
4657 - fix NULL deref during bulk timeout (214)
4658 - fix problems related to multiple filesystems on one MDS (241)
4659 - fixed serious subtle metadata locking bugs
4660 - free locks on clients when inodes are removed due to memory
4662 - fix inode pointer in lock data (285)
4663 - partial support for multiple MDS on a single host (241)
4664 - data locks weren't cancelled at clear_inode time (290, 311)
4665 - intent locks could lead to unbounded lock growth (205)
4666 - added a maximum lock count, an LRU list, and a flusher
4667 - fix multiple rename (365)
4668 - properly abstracted the echo client
4669 - OSC locked 1 byte too many; fixed
4670 - rewrote brw callback code:
4671 - fixed recovery bugs related to LOVs (306)
4672 - fixed too-many-pages-in-one-write crash (191)
4673 - fixed (again) crash in sync_io_timeout (214)
4674 - probably fixed callback-related race (385)
4676 - Add capability to MDS protocol
4677 - LDLM cancellations and callbacks on different portals
4679 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4682 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4683 - defer open object destruction to close time (601981, 138)
4684 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4685 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4686 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4687 - don't call obd_setattr in ll_file_release for destroyed objects
4688 * protocol change to lustre_msg: move |version| and add |flags|
4689 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4690 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4691 * added replay of create, unlink, link and rename operations during
4692 MDS failover; recovery should be much more robust now
4693 * remove failed OSCs from LOVs (only lov_create uses this so far)
4694 * the lustre-HOWTO was brought (more) up to date (582544)
4696 2002-10-23 Phil Schwan <phil@clusterfs.com>
4699 - in-use dentries weren't being reused properly (617851)
4700 - prevent multiple LDLM setup (599178)
4701 - fix LOV size calculations for truncate (617853)
4702 - fix client handling of MDS intent errors (POSIX)
4703 - fix permission bug in lovstripe.c test (624321)
4704 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4705 - truncate past end of file could corrupt data
4706 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4707 - a race in recovery could return ETIMEDOUT to apps (623947)
4708 - building outside the source directory was fixed
4709 * the lustre-HOWTO was brought (more) up to date (582544)
4710 * major progress was made on recovery functionality
4712 2002-10-10 Phil Schwan <phil@clusterfs.com>
4715 - recovery deadlock fix
4716 - rm -rf causes LBUG fix (617817)
4717 - file open by multiple tasks fix (618962)
4718 - directory permissions bugs (602707 and 620007)
4719 - journal_stop fixed with locking (611313)
4720 - O_APPEND failures resolved (618273, perhaps 614459)
4721 - lconf PATH fix (619770)
4722 - IA64 build fix (621450)
4723 - RPC buffer sizes scale with amount of memory
4725 2002-10-01 Phil Schwan <phil@clusterfs.com>
4728 - locks would be cancelled without throwing away data pages,
4729 resulting in inconsistent data (605627)
4730 - inode attributes were not always being refreshed (605627, 612449)
4731 - lconf now continues to cleanup after lctl reports an error
4732 - MDS now enforces user permissions (602707)
4733 - lprocfs cleanup fixed, but not yet enabled (614157)
4734 - fixed infinite server hang, should a client not respond to an AST
4735 - avoid going into recovery if user calls readlink() with a buffer
4736 that's too small (613941)
4737 - AST RPCs no longer require replies (614867) -- this may be changed
4738 - don't crash server if client sends an IOV that's too big (611336)
4739 - fixed lock conversion deadlock (611892)
4740 - fixed the following of symlinks (614622)
4741 * recovery: the server can remove locks from a client that dies, other
4742 clients can make progress
4743 * more extN patch fixes
4744 * compile-time configurable ptlrpc buffer allocations
4746 - collaborative read cache document
4747 - Lustre Lite Performance CDR document-in-progress
4749 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4752 - fix typo in patch-2.4.18
4754 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4757 - clear ptlrpc request each time in handle_incoming_request()
4758 - unlink of files now destroys the object on the OST
4760 2002-09-19 Peter Braam <braam@clusterfs.com>
4762 * add hard link support
4763 * change obdfile creation method
4764 * kernel patch changed
4766 2002-09-19 Peter Braam <braam@clusterfs.com>
4769 - stack overflow bug in extN fixed
4771 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4773 * documentation updates
4774 - add man pages for config tools
4775 - update tests/README to describe testing with new config tools
4776 - finish metadata API descriptions
4777 * bug fixes and cleanups
4778 - statfs workaround for 16TB limit
4779 - LOV stripe allocation improved, can stripe on subset of OSTs
4780 - LOV file size/IO offset was wrong for files > 4GB in size
4781 - object EA data was being dropped, caused files to be unreadable
4782 - memory overflow with non-LOV OST caused memory corruption
4783 - fixed regression tests to work with new config tools, obdfilter
4784 - fixed bug when directory size became larger than 1 block
4785 - fixed bug (for single client case) when PWD was deleted
4786 - invalidate local directory pages when doing intent-based ops
4787 - avoid LDLM oops when lock callback contained bad data
4789 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4791 * documentation updates
4792 * bug fixes and cleanups
4793 - configuration tools
4796 - 64-bit compile warnings
4797 - 64-bit internal statfs data
4799 * test_brw on persistent OST devices
4801 * lprocfs (disabled)
4803 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4805 * documentation updates
4806 * bug fixes and cleanups
4807 * configuration tools
4809 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4812 * many small fixes to 0_5_4
4813 * io/network handling
4814 * thinkos in MDS operations
4816 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4819 * crucial basic fixes to 0.5.3
4820 * IOR, Iozone work over Elan
4823 2002-08-07 Phil Schwan <phil@clusterfs.com>
4824 * version 0_5_3, our first alpha
4825 * we use the new Portals iovs
4826 * documentation updates
4827 * bug fixes and cleanups
4828 * small changes in the DLM wire protocol
4830 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4831 * version 0_5_1 with some initial stability,
4832 * locking on MD and file I/O.
4833 * documentation updates
4834 * several bug fixes since 0.5.0
4835 * small changes in wire protocol
4837 2002-07-18 Phil Schwan <phil@clusterfs.com>
4839 * delivered as Lustre Light Alpha
4840 * fixed a crash after handling invalid MDS requests
4841 * fixed directory pages for architectures with non-4k pages sizes
4843 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4844 * release version v0_4_4
4845 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4847 * Fixes a number of interruption problems with OST operations.
4848 * Update documentation for portals header changes
4849 * Move all wire protocol structs/defines to lustre_idl.h
4850 * Fixes symlink length bug.
4851 * Add tcpdump to repository.
4853 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4854 * release version v0_4_3
4855 * Fixes statfs for inodes on extN.
4856 * Fixes bug in runtests which would delete /etc/hosts.
4857 * Use 64-bit object IDs wherever possible (not into VFS though)
4858 Remove ost_get_info, which is unused by lustre, and out of date.
4860 2002-07-03 Peter Braam <braam@clusterfs.com>
4861 * release version v0_4_2 Fixes a lookup error (type not passed)
4862 * move forward to head of Portals
4863 * move forward to latest Lustre kernel
4865 2002-06-25 Peter Braam <braam@clusterfs.com>
4866 * release version v0_4_1. Hopefully stable on single node use.