1 tbd Sun Microsystems, Inc.
4 2.6.16.54-0.2.5 (SLES 10),
5 2.6.18-53.1.21.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.40.7-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: add recount protection for osc callbacks, so avoid panic on shutdown
21 Description: sanity test 65a fails if stripecount of -1 is set
22 Details : handle -1 striping on filesystem in ll_dirstripe_verify
27 Description: ASSERTION(CheckWriteback(page,cmd)) failed
28 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
31 Severity : enhancement
33 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
37 Description: do not process already freed flock
38 Details : flock can possibly be freed by another thread before it reaches
39 to ldlm_flock_completion_ast.
43 Description: LBUG during stress test
44 Details : Need properly lock accesses the flock deadlock detection list.
48 Description: oops in page fault handler
49 Details : kernel page fault handler can return two special 'pages' in error case, don't
50 try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
54 Description: timeout with invalidate import.
55 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which should be
56 handled by ptlrpcd. This produce long age waiting and -ETIMEOUT
57 ptlrpc_invalidate_import and as result LASSERT.
59 Severity : enhancement
61 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
65 Description: enable MGS and MDT services start separately
66 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
67 a co-located MGS without starting the MGS, which is a complement
68 to 'nosvc' mount option.
72 Frequency : after recovery
73 Description: precreate to many object's after del orphan.
74 Details : del orphan st in oscc last_id == next_id and this triger growing
75 count of precreated objects. Set flag LOW to skip increase count
76 of precreated objects.
80 Frequency : rare, on clear nid stats
81 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
82 Details : when clean nid stats sometimes try destroy live entry,
83 and this produce panic in free.
87 Description: Stack overflow during MDS log replay
88 ease stack pressure by using a thread dealing llog_process.
92 Description: wait until IO finished before start new when do lock cancel.
93 Details : VM protocol want old IO finished before start new, in this case
94 need wait until PG_writeback is cleared until check dirty flag and
95 call writepages in lock cancel callback.
97 Severity : enhancement
99 Description: using special macro for print time and cleanup in includes.
103 Description: mds_mfd_close() ASSERTION(rc == 0)
104 Details : In mds_mfd_close(), we need protect inode's writecount change
105 within its orphan write semaphore to prevent possible races.
109 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
113 Frequency : rare, on shutdown ost
114 Description: don't hit live lock with umount ost.
115 Details : shrink_dcache_parent can be in long loop with destroy dentries,
116 use shrink_dcache_sb instead.
120 Description: don't panic with use echo client
121 Details : echo client pass NULL as client nid pointer and this produce null
126 Description: fix build on ppc32
127 Details : compile code with -m64 flag produce wrong object file for ppc32.
131 Description: add message levels for liblustreapi
135 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
136 Details : Move the 'good_osts' check before the 'total_bavail' check. This
137 will result in an -EAGAIN and in the exit call path we call
138 alloc_rr() which will with increasing aggressiveness attempt to
139 aquire precreated objects on the minimum number of required OSCs.
143 Description: Use old size assignment to avoid deadlock
144 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
145 the scheduling while holding a spinlock. We do not need locking
146 for size in ll_update_inode() because size is only updated from
147 the MDS for directories or files without objects, so there is no
148 other place to do the update, and concurrent access to such inodes
149 are protected by the inode lock.
153 Description: resolve "_IOWR redefined" build error on SLES10
157 Description: dump the memory debugging after all modules are unloaded to
158 suppress false negative in conf_sanity test 39
160 Severity : enhancement
162 Description: build kernel-ib packages for OFED 1.3 in our release cycle
167 Description: fix SLES kernel versioning
168 Details : the kernel version for our SLES 10 kernel did not include a "-"
169 before the "smp" at the end. while this was not a problem in
170 general, it did mean that software trying to use the kernel
171 version to try to detect a vendor specific kernel would fail.
172 this was most evident by the OFED build scripts.
176 Description: Don't update lov_desc members until making sure they are valid
177 Details : When updating lov_desc members via proc fs, need fix their
178 validities before doing the real update.
182 Description: don't put request into delay list while invalidate in flight.
183 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
184 import in flight. this produce timeout for invalidate and sometimes
185 can cause stale data.
189 Frequency : on ppc only
190 Description: not convert ost objects for directory because it's not exist.
191 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
195 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
197 Severity : enhancement
199 Description: Update to support 2.6.22.14 vanilla kernel.
203 Frequency : rare, on recovery
204 Description: read procfs can produce deadlock in some situation
205 Details : Holding lprocfs lock which send rpc can produce block for destroy
206 obd objects and this also block reconnect with -EALREADY. This isn't
207 fix all lprocfs bugs - but make it rare.
209 Severity : enhancement
211 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
214 Frequency : frequent on X2 node
216 Description: mdc_set_open_replay_data LBUG
217 Details : Set replay data for requests that are eligible for replay.
221 Description: lustre_mgs: operation 101 on unconnected MGS
222 Details : When MGC is disconnected from MGS long enough, MGS will evict the
223 MGC, and late on MGC cannot successfully connect to MGS and a lot
224 of the error messages complaining that MGS is not connected.
228 Frequency : on network error
229 Description: panic with double free request if network error
230 Details : mdc_finish_enqueue is finish request if any network error ocuring,
231 but it's true only for synchronus enqueue, for async enqueue
232 (via ptlrpcd) this incorrect and ptlrpcd want finish request
235 Severity : enhancement
237 Description: client-side metadata stat-ahead during readdir(directory readahead)
238 Details : perform client-side metadata stat-ahead when the client detects
239 readdir and sequential stat of dir entries therein
242 Frequency : on start mds
244 Description: Implement get_info(last_id) in obdfilter.
247 Frequency : occasional
249 Description: Correctly check stale fid, not start epoch if ost not support SOM
250 Details : open with flag O_CREATE need set old fid in op_fid3 because op_fid2
251 overwrited with new generated fid, but mds can anwer with one of these
252 two fids and both is not stale. setattr incorectly start epoch and
253 assume will be called done_writeting, but without SOM done_writing
257 Frequency : rare, depends on device drivers and load
259 Description: MDS or OSS nodes crash due to stack overflow
260 Details : Code changes in 1.8.0 increased the stack usage of some functions.
261 In some cases, in conjunction with device drivers that use a lot
262 of stack the MDS (or possibly OSS) service threads could overflow
263 the stack. One change which was identified to consume additional
264 stack has been reworked to avoid the extra stack usage.
267 Frequency : occasional
269 Description: Do not fail import if osc_interpret_create gets -EAGAIN
270 Details : If osc_interpret_create got -EAGAIN it immediately exits and
271 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
272 oscc_has_objects and see OSC has no objests and call
273 oscc_internal_create to resend create request.
275 Severity : enhancement
277 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
279 Severity : enhancement
281 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
284 Frequency : very rare
286 Description: Fix replay if there is an un-replied request and open
287 Details : In some cases, older replay request will revert the
288 mcd->mcd_last_xid on MDS which is used to record the client's
291 Severity : enhancement
293 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
295 Severity : enhancement
297 Description: Add rhel5 support to HEAD.
299 Serverity : enhancement
301 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
306 Description: Don't allow skipping OSTs if index has been specified.
307 Details : Don't allow skipping OSTs if index has been specified, make locking
308 in internal create lots better.
312 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
313 Details : access to bitfield in structure is always rounded to long
314 and this produce problem with not atomic change any bit.
318 Description: Lustre make rpms failed.
319 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
320 builds Lustre rpms from tarball.
323 Frequency : rare on shutdown ost
325 Description: If llog cancel was not send before clean_exports phase, this can
326 produce deadlock in llog code.
327 Details : If llog thread has last reference to obd and call class_import_put
328 this produce deadlock because llog_cleanup_commit_master wait when
329 last llog_commit_thread exited, but this never success because was
330 called from llog_commit_thread.
334 Description: allow userland application know is lost one of stripes.
335 Details : fill lvb_blocks with error code on ost and return it to
336 application if error flag found.
340 Description: NULL lov_tgts causing MDS oops
341 Details : more safe checks for NULL lov_tgts for avoid oops.
343 Severity : enhancement
345 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
349 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
350 Details : If an OST is down the MDS will hang indefinitely in
351 obd_statfs_rqset() waiting for the statfs data. While for
352 MDS QOS usage of statfs, it should not stuck in waiting.
354 Severity : enhancement
356 Description: remote_acl support
357 Details : Support ACL-based permission check for remote user.
358 Support setfacl/getfacl for remote user with the utils
359 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
360 the system "{s,g}etfacl" utils.
362 Severity : enhancement
364 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
366 Severity : enhancement
368 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
372 Description: make mgs_setparam() handle fsname containing dash
373 Details : fsname containing a dash does not work with lctl conf_param
375 Severity : enhancement
377 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
379 Severity : enhancement
381 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
385 Frequency : rare, at shutdown
386 Description: access already free / zero obd_namespace.
387 Details : if client_disconnect_export was called without force flag set,
388 and exist connect request in flight, this can produce access to
389 NULL pointer (or already free pointer) when connect_interpret
390 store ocd flags in obd_namespace.
394 Frequency : only at startup
395 Description: not alloc memory with spinlock held.
396 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
397 if any spinlock held.
399 Severity : enhancement
401 Description: make lustre randomly fail allocating memory
402 Details : Make lustre randomly failed allocating memory for testing purpose.
404 Severity : enhancement
406 Description: lost problems with lov objid file
407 Details : Fixes some scability and access to not inited memory problems
408 in work with lov objdid file.
413 Description: lfs find does not continue on file error
414 Details : Continue other files processing when a file/dir is absent.
418 Description: Inconsistent usage of lustre_pack_reply()
419 Details : Standardize the usage of lustre_pack_reply() such that it
420 always generate a CERROR on failure.
423 Frequency : occasional
425 Description: MDS hang or stay in waiting lock
426 Details : If client receive lock with CBPENDING flag ldlm need send lock
427 cancel as separate rpc, to avoid situation when cancel request
428 can't processed due all i/o threads stay in wait lock.
432 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
436 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
440 Description: Update to SLES9 kernel-2.6.5-7.287.3.
444 Description: Update to RHEL4 latest kernel.
446 Severity : enhancement
448 Description: Build SLES10 patchless client fails
449 Details : The configure was broken by run ./configure with
450 --with-linux-obj=.... argument for patchless client. When the
451 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
452 can't search header adequately. Use absolute path such as
453 -I($LINUX)/include instead.
457 Description: interrupt oig_wait produce painc on resend.
458 Details : brw_redo_request can be used for resend requests from ptlrpcd and
459 private set, and this produce situation when rq_ptlrpcd_data not
460 copyed to new allocated request and triggered LBUG on assert
461 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
462 ptlrpcd set if request is changed and can be safety changed to use
467 Description: LASSERT_{REQ,REP}SWAB macros are buggy
468 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
469 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
470 these macros and replace them with their difinitions instead.
474 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
475 Details : Update which_patch & target file for SLES10 latest kernel.
478 Description: Remove client patches from SLES 10 kernel.
479 Details : This causes SLES 10 clients to behave as patchless clients
480 even on a Lustre-patched (server) kernel.
482 Severity : enhancement
484 Description: self-adjustable client's lru lists
485 Details : use adaptive algorithm for managing client cached locks lru
486 lists according to current server load, other client's work
487 pattern, memory activities, etc. Both, server and client
488 side namespaces provide number of proc tunables for controlling
491 Severity : enhancement
493 Description: light-weight GSS support
494 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
495 and reduce performance overhead.
497 Severity : enhancement
499 Description: Linux keyring support
500 Details : Support using service of Linux keyring for Lustre GSS internal
501 context refresh/cache mechanism.
505 Description: Fix errors in lfs documentation
506 Details : Fixes man pages
510 Description: don't use GFP_* in generic Lustre code.
511 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
515 Description: obdclass is limited by single OBD_ALLOC(idarray)
516 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
520 Description: warnings with build patchless client with vanila 2.6.19 and up
521 Details : change old ctl_table style and replace ctl_table/ctl_table_header
522 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
526 Frequency : Only for SLES
527 Description: sanity_quota fail test_1
528 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
529 file, which makes TSTID[2] inunique.
533 Frequency : Only for SLES9
534 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
535 Details : Update target/ChangeLog/which_patch .
539 Frequency : Only for test_10 in sanity-quota.sh
540 Description: a bug in quota test code
541 Details : checking a test flag in a wrong place
545 Frequency : Only for test_13 in sanity-quota.sh
546 Description: a bug in quota test script
547 Details : shouldn't check overfull allocated quota
551 Description: jbd statistics
552 Details : Port older jbd statistics patch for sles10
554 Severity : enhancement
556 Description: Add printing inode info into message about error in writepage.
561 Description: reply_lock_interpret crash due to race with it and lock cancel.
562 Details : Do not replay locks that are being cancelled. Do not reference
563 locks by their address during replay, just by their handle.
567 Frequency : When flocks are used.
568 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
569 Details : Flock locks might destroy just granted lock if it could be merged
570 with another existing flock, this is done in completion handler,
571 so teach ldlm_cli_enquque_fini that this is a valid case for
577 Description: Oops in read and write path when failing to allocate lock.
578 Details : Check if lock allocation failed and return error back.
582 Description: lstripe command fails for valid OST index
583 Details : The stripe offset is compared to lov->desc.ld_tgt_count
584 instead of lov->desc.ld_active_tgt_count.
588 Description: sanity.sh failed test 103
589 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
594 Description: df doesn't work properly if diskfs blocksize != 4K
595 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
599 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
600 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
604 Description: obd_health_check_timeout too short
605 Details : set obd_health_check_timeout as 1.5x of obd_timeout
609 Description: llapi_file_create() does not allow some changes
610 Details : add llapi_file_open() that allows specifying the mode and
611 open flags, and also returns an open file handle.
615 Description: merge and cleanup kernel patches.
616 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
620 Description: Add journal checksum support.(Kernel part)
621 Details : The journal checksum feature adds two new flags i.e
622 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
623 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
624 indicates that the commit block contains the checksum for
625 the blocks described by the descriptor blocks. Now commit
626 record can be sent to disk without waiting for descriptor
627 blocks to be written to disk. This behavior is controlled
628 using JBD2_FEATURE_ASYNC_COMMIT flag.
632 Description: OSS needs mutliple precreate threads
633 Details : Add ability to start more than one create thread per OSS.
637 Description: Sanity.sh test_65k failed: setstripe should have succeeded
638 Details : execute lfs setstripe on client
642 Description: mds_obd_create error creating tmp object
643 Details : When the user sets quota on root, llog will be affected and can't
644 create files and write files.
647 Frequency : Always on ia64 patchless client, and possibly others.
649 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
650 Details : This allows the patchless client to be loaded on architectures
655 Description: RedHat Update kernel for RHEL5
656 Details : Add the kernel config file for RHEL5.
660 Description: RedHat Update kernel for RHEL5
661 Details : Modify the kernel config file more closer RHEL5.
665 Description: Build failure against Centos5 (RHEL5)
666 Details : Define PAGE_SIZE when it isn't present.
670 Description: test_7 on sanity-quota.sh can't be used on separate machines
671 Details : add facet to handle it
675 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
676 Details : with NFS, the anon dentry's parent was set to itself in
677 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
678 corresponding dentry other than use rec->ur_name.
680 Severity : enhancement
682 Description: lfs setstripe enhancement
683 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
687 Description: enable data checksumming by default
688 Details : enable checksum by default, allow --disable-checksum
689 configure option and "-o nochecksum" mount option. Checksums
690 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
694 Description: lustre support for RHEL5
695 Details : Add support for RHEL5.
699 Description: Client eviction due to failover config
700 Details : after a connection loss, the lustre client should attempt to
701 reconnect to the last active server first before trying the
702 other potential connections.
706 Description: when mds and osts use different quota unit(32bit and 64bit),
707 quota will be released repeatly.
708 Details : void sending multiple quota reqs to mds, which will keep the
709 status between the reqs.
713 Description: osts not allocated evenly to files
714 Details : change the condition to increase offset_idx
717 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
719 Description: Data corruption for OSTs that are formatted larger than 2TB
721 Details : When generating the bio request for lustre file writes the
722 sector number would overflow a temporary variable before being
723 used for the IO. The data reads correctly from Lustre (which
724 will overflow in a similar manner) but other file data or
725 filesystem metadata may be corrupted in some cases.
729 Description: Tune the kernel for good SCSI performance.
730 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
731 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
736 Description: rewrite ext2-derived code in obdclass/uuid.c
737 Details : rewrite inherited code (uuid parsing code from ext2 utils)
738 from scratch preserving functionality.
742 Description: "lfs find -obd UUID" prints directories
743 Details : "lfs find -obd UUID" will return all directory names instead
744 of just file names. It is incorrect because the directories
745 do not reside on the OSTs.
749 Description: use i_size_read and i_size_write in 2.6 port
750 Details : replace inode->i_size access with i_size_read/write()
753 Frequency : when removing large files
755 Description: scheduling issue during removal of large Lustre files
756 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
757 It causes scheduling issues when removing large files (17TB in the
761 Frequency : only with liblustre clients on XT3
763 Description: evictions taking too long
764 Details : allow llrd to evict clients directly on OSTs
767 Frequency : only on ppc
769 Description: /proc/fs/lustre/devices broken on ppc
770 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
771 We should make sure the type of 'index' is loff_t and then cast
772 explicitly as needed below. Do not assign an explicitly cast
776 Frequency : only for rhel5
778 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
779 Details : Modify the target file & which_kernel.
781 Severity : enhancement
783 Description: omit set fsid for export NFS
784 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
785 as FS_REQUIRES_DEV for avoid problems with generate fsid.
789 Frequency : Always, for kernels after 2.6.16
790 Description: Fix warning idr_remove called for id=.. which is not allocated.
791 Details : Last kernels save old s_dev before kill super and not allow
792 to restore from callback - restore it before call kill_anon_super.
796 Description: buffer overruns could theoretically occur
797 Details : llapi_semantic_traverse() modifies the "path" argument by
798 appending values to the end of the origin string, and a
799 overrun may occur. Adding buffer overrun check in liblustreapi.
803 Description: Fix error on 'ls .' at the top of the Lustre mount.
804 Details : Don't revalidate dentry if it is a root dentry.
808 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
809 Details : Modify vm-tunables-rhel4.patch.
813 Description: Kernel config for 2.6.18-vanilla.
814 Details : Modify targets/2.6-vanilla.target.in.
815 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
816 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
817 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
818 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
822 Description: improve handling recoverable errors
823 Details : if request processig with error which can be recoverable on server
824 request should be resend, otherwise page released from cache and
829 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
830 Details : Modify target file & which_patch.
831 A flaw was found in the IA32 system call emulation provided
832 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
833 value could be stored in the %RAX register, which could trigger an
834 out-of-bounds system call table access. An untrusted local user
835 could exploit this flaw to run code in the kernel
836 (ie a root privilege escalation). (CVE-2007-4573).
840 Description: O_DIRECT bypasses client statistics.
841 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
842 read_ahead_stats were updated. copied the stats section from
843 osc_send_oap_rpc() into async_internal().
847 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
851 Description: change order of libsysio includes
852 Details : '#include sysio.h' should always come before '#include xtio.h'
856 Description: handle "serious error: objid * already exists" more gracefully
857 Details : If LAST_ID value on disk is smaller than the objects existing in
858 the O/0/d* directories, it indicates disk corruption and causes an
859 LBUG(). If the object is 0-length, then we should use the existing
860 object. This will help to avoid a full fsck in most cases.
862 Severity : enhancement
864 Description: adapt the lustre_config script to support the upgrade case
865 Details : Add "-u" option for lustre_config script to support upgrading 1.4
866 server targets to 1.6 in parallel.
870 Description: To avoid grant space > avaible space when the disk is almost
871 full. Without this patch you might see the error "grant XXXX >
872 available" or some LBUG about grant, when the disk is almost
874 Details : In filter_check_grant, for non_grant cache write, we should
875 check the left space by if (*left > ungranted + bytes), instead
876 of (*left > ungranted), because only we are sure the left space
877 is enough for another "bytes", then the ungrant space should be
878 increase. In client, we should update cl_avail_grant only there
879 is OBD_MD_FLGRANT in the reply.
883 Description: Update RHEL 4 kernel to fix local root privilege escalation.
884 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
885 described in CVE-2007-4573. This problem could allow untrusted
886 local users to gain root access.
889 Frequency : when using O_DIRECT and quotas
891 Description: Incorrect file ownership on O_DIRECT output files
892 Details : block usage reported by 'lfs quota' does not take into account
893 files that have been written with O_DIRECT.
898 Description: touch file failed when fs is not full
899 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
900 otherwise we can get ENOSP while fs is not full.
904 Description: parallel lock callbacks
905 Details : Instead of sending blocking and completion callbacks as separated
906 requests, adding them to a set and sending in parallel.
909 Frequency : only for Cray XT3
910 Bugzilla : 12829/13455
911 Description: Changing primary group doesn't change the group lustre assigns to
913 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
914 overridden with the primary group provided by the group upcall,
915 whereas the supplied fsgid can be trusted if it is in the list of
916 supplementary groups returned by the group upcall.
918 Severity : enhancement
920 Description: Allow masking D_WARNING, D_ERROR messages from console
921 Details : Console messages can now be disabled via lnet.printk.
925 Description: User code with malformed file open parameter crashes client node
926 Details : Before packing join_file req, all the related reference should be
927 checked carefully in case some malformed flags cause fake
928 join_file req on client.
932 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
933 Details : ldlm_completion_ast() assumes that a lock is granted when the req
934 mode is equal to the granted mode. However, it should also check
935 that LDLM_FL_CP_REQD is not set.
939 Description: Heavy nfs access might result in deadlocks
940 Details : After ELC code landed, it is now improper to enqueue any mds
941 locks under och_sem, because enqueue might want to decide to
942 cancel open locks for same inode we are holding och_sem for.
946 Description: Client eviction while running blogbench
947 Details : A lot of unlink operations with concurrent I/O can lead to a
948 deadlock causing evictions. To address the problem, the number of
949 oustanding OST_DESTROY requests is now throttled to
950 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
951 callbacks are processed in priority.
955 Description: enable ACLs on MDS by default
956 Details : ACLs must be enabled on MDS by default.
959 Frequency : PPC/PPC64 only
961 Description: conflicts between asm-ppc64/types.h and lustre_types.h
962 Details : fix duplicated definitions between asm-ppc64/types.h and
963 lustre_types.h on PPC.
966 Frequency : PPC/PPC64 only
968 Description: asm-ppc/segment.h does not exist
969 Details : fix compile issue on PPC.
973 Description: better handle error messages in extents code
976 Frequency : RHEL4 only
978 Description: mkfs is very slow on IA64/RHEL4
979 Details : A performance regression has been discovered in the MPT Fusion
980 driver between versions 3.02.73rh and 3.02.99.00rh. As a
981 consequence, we have downgraded the MPT Fusion driver in the RHEL4
982 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
984 Severity : enhancement
986 Description: SNMP support enhancement
987 Details : Adding total number of sampled request for an MDS node in snmp
990 Severity : enhancement
992 Description: Optimize ldlm waiting list processing for PR extent locks
993 Details : When processing waiting list for read extent lock and meeting read
994 lock that is same or wider to it that is not contended, skip
995 processing rest of the list and immediatelly return current
996 status of conflictness, since we are guaranteed there are no
997 conflicting locks in the rest of the list.
1001 Description: Time out and refuse to reconnect
1002 Details : When the failover node is the primary node, it is possible
1003 to have two identical connections in imp_conn_list. We must
1004 compare not conn's pointers but NIDs, otherwise we can defeat
1005 connection throttling.
1009 Description: port llog fixes from b1_6 into HEAD
1010 Details : Port llog reference couting and some llog cleanups from b1_6
1011 (bug 10800) into HEAD, for protect from panic and access to already
1012 free llog structures.
1016 Description: Detect stride IO mode in read-ahead
1017 Details : When a client does stride read, read-ahead should detect that and
1018 read-ahead pages according to the detected stride pattern.
1022 Description: data checksumming impacts single node performance
1023 Details : add support for several checksum algorithm. Currently, only CRC32
1024 and Adler-32 are supported. The checksum type can be changed on
1025 the fly via /proc/fs/lustre/osc/*/checksum_type.
1029 Description: use adler32 for page checksums
1030 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1035 Description: build for x2 fails
1036 Details : fix compile issue on Cray systems.
1040 Description: Properly match for duplicate locks
1041 Details : Due to different lock order from skiplists code, we need to
1042 traverse entire list for now
1045 Frequency : only on PPC/SLES10
1047 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1048 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1049 be defined. Add a hack in mkfs_lustre.c to work around this compile
1054 Description: LASSERT on MDS when client holding flock lock dies
1055 Details : ldlm pool logic depends on number of granted locks equal to
1056 number of released locks which is not true for flock locks, so
1057 just exclude such locks from consideration.
1061 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1062 Details : Use fsfilt_sync() for both the whole filesystem sync and
1063 individual file sync to eliminate dangerous inode locking
1064 with I_LOCK that can lead to a deadlock.
1068 Description: performance in 1.6.3
1069 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1070 sure that 1MB requests are not fragmented by the block layer.
1072 Severity : enhancement
1074 Description: organize the server-side client stats on per-nid basis
1075 Details : Change the structure of stats under obdfilter and mds to
1083 The "uuid"s file would list the uuids of _active_ exports.
1084 And the clear entry is to clear all stats and stale nids.
1086 Severity : enhancement
1088 Description: eliminate client locks in face of contention
1089 Details : file contention detection and lockless i/o implementation
1090 for contended files.
1094 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1097 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1099 Description: MDS may lose file striping (and hence file data) in some cases
1100 Details : If there are additional extended attributes stored on the MDS,
1101 in particular ACLs, SELinux, or user attributes (if user_xattr
1102 is specified for the client mount options) then there is a risk
1103 of attribute loss. Additionally, the Lustre file striping
1104 needs to be larger than default (e.g. striped over all OSTs),
1105 and an additional attribute must be stored initially in the
1106 inode and then increase in size enough to be moved to the
1107 external attribute block (e.g. ACL growing in size) for file
1112 Description: skiplist implementation simplification
1113 Details : skiplists are used to group compatible locks on granted list
1114 that was implemented as tracking first and last lock of each lock group
1115 the patch changes that to using doubly linked lists
1119 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1120 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1121 because LASSERT on both the data supplied by a client, and the data
1122 on disk is dangerous and incorrect.
1124 Severity : enhancement
1126 Description: Slow truncate/writes to huge files at high offsets.
1127 Details : Directly associate cached pages to lock that protect those pages,
1128 this allows us to quickly find what pages to write and remove
1129 once lock callback is received.
1133 Description: more ldlm soft lockups
1134 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1135 starve other threads from the resource lock for a long time in
1136 case of long waiting queue, so change the debug level from
1137 D_OTHER to the less frequently used D_INFO.
1139 Severity : enhancement
1141 Description: add -gid, -group, -uid, -user options to lfs find
1145 Description: Hung threads in invalidate_inode_pages2_range
1146 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
1147 one is completed. As a result, some RPCs are stuck in the queue
1152 Description: filter threads hungs on waiting journal commit
1153 Details : Cleanup filter group llog code, then only filter group llog will
1154 be only created in the MDS/OST syncing process.
1158 Description: Procfs and llog threads access destoryed import sometimes.
1159 Details : Sync the import destoryed process with procfs and llog threads by
1160 the import refcount and semaphore.
1162 Severity : enhancement
1164 Description: openlock cache of b1_6 port to HEAD
1166 --------------------------------------------------------------------------------
1168 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
1170 * Support for kernels:
1171 2.6.5-7.283 (SLES 9),
1172 2.6.9-55.EL (RHEL 4),
1173 2.6.16.46-0.14 (SLES 10),
1174 2.6.18.8 vanilla (kernel.org)
1175 * Client support for unpatched kernels:
1176 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1177 2.6.16 - 2.6.22 vanilla (kernel.org)
1178 * Due to recently discovered recovery problems, we do not recommend
1179 using patchless RHEL 4 clients with this or any earlier release.
1180 * Recommended e2fsprogs version: 1.39.cfs8
1181 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1182 * Starting with this release, the ldiskfs backing filesystem required
1183 by Lustre is now in its own package, lustre-ldiskfs. This package
1184 should be installed. It is versioned separately from Lustre and
1185 may be released separately in future.
1189 Description: block reactivating mgc import until all deactivates complete
1190 Details : Fix race when failing back MDT/MGS to itself (testing)
1193 Frequency : at statup only
1195 Description: mds_lov_synchronize race leads to various problems
1196 Details : simultaneous MDT->OST connections at startup can cause the
1197 sync to abort, leaving the OSC in a bad state.
1199 Severity : enhancement
1201 Description: add optional extra BUILD_VERSION info
1202 Details : add a new environment variable (namely LUSTRE_VERS) which allows
1203 to override the lustre version.
1206 Frequency : 2.6.18 servers only
1208 Description: ll_kern_mount() doesn't release the module reference
1209 Details : The ldiskfs module reference count never drops down to 0
1210 because ll_kern_mount() doesn't release the module reference.
1215 Description: server LBUG when using old ost_num_threads parameter
1216 Details : Accept the old ost_num_threads parameter but warn that it
1217 is deprecated, and fix an off-by-one error that caused an LBUG.
1222 Description: Transient SCSI error results in persistent IO issue
1223 Details : iobuf->dr_error is not reinitialized to 0 between two
1227 Frequency : sometimes when underlying device returns I/O errors
1229 Description: OSTs not going read-only during write failures
1230 Details : OSTs are not remounted read-only when the journal commit threads
1231 get I/O errors because fsfilt_ext3 calls journal_start/stop()
1232 instead of the ext3 wrappers.
1236 Description: poor connect scaling with increasing client count
1237 Details : Don't run filter_grant_sanity_check for more than 100 exports
1238 to improve scaling for large numbers of clients.
1241 Frequency : SLES10 only
1243 Description: sanity-quota.sh quotacheck failed: rc = -22
1244 Details : Quotas cannot be enabled on SLES10.
1247 Frequency : liblustre clients only
1249 Description: getdirentries does not give error when run on compute nodes
1250 Details : getdirentries does not fail when the size specified as an argument
1251 is too small to contain at least one entry
1253 Severity : enhancement
1255 Description: Add LNET router traceability for debug purposes
1256 Details : If a checksum failure occurs with a router as part of the
1257 IO path, the NID of the last router that forwarded the bulk data
1258 is printed so it can be identified.
1263 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
1264 Details : Due to a race condition, liblustre clients were occasionally
1265 evicted incorrectly.
1267 Severity : enhancement
1269 Description: lfs setstripe use optional parameters instead of postional
1272 Severity : enhancement
1274 Description: Nanosecond timestamp support for ldiskfs
1275 Details : The on-disk ldiskfs filesystem has added support for nanosecond
1276 resolution timestamps. There is not yet support for this at
1277 the Lustre filesystem level.
1280 Frequency : during server recovery
1282 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
1283 Details : request with rq_no_resend flag not awake l_wait_event if they get a
1287 Frequency : nfs export on patchless client
1289 Description: connectathon hang when test nfs export over patchless client
1290 Details : Disconnected dentry cannot be found with lookup, so we do not need
1291 to unhash it or make it invalid
1294 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
1295 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
1296 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
1301 Description: open req refcounting wrong on reconnect
1302 Details : If reconnect happened between getting open reply from server and
1303 call to mdc_set_replay_data in ll_file_open, we will schedule
1304 replay for unreferenced request that we are about to free.
1305 Subsequent close will crash in variety of ways.
1306 Check that request is still eligible for replay in
1307 mdc_set_replay_data().
1312 Description: disable writes to filesystem when reading health_check file
1313 Details : the default for reading the health_check proc file has changed
1314 to NOT do a journal transaction and write to disk, because this
1315 can cause reads of the /proc file to hang and block HA state
1316 checking on a healthy but otherwise heavily loaded system. It
1317 is possible to return to the previous behaviour during configure
1318 with --enable-health-write.
1320 Severity : enhancement
1322 Description: 64-bit inode version
1323 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
1324 to the inode. This will be required for version-based recovery.
1329 Description: MDS fails to start if a duplicate client export is detected
1330 Details : in some rare cases it was possible for a client to connect to
1331 an MDS multiple times. Upon recovery the MDS would detect this
1332 and fail during startup. Handle this more gracefully.
1334 Severity : enhancement
1336 Description: Add -o localflock option to simulate old noflock
1338 Details : This will achieve local-only flock/fcntl locks
1344 Description: log_commit_thread vs filter_destroy race leads to crash
1345 Details : Take import reference before releasing llog record semaphore
1350 Description: Wrong request locking in request set processing
1351 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
1352 imp_delayed_list, in this place should be used imp_lock.
1355 Frequency : when reconnection
1357 Description: Grant Leak when osc reconnect to OST
1358 Details : When osc reconnect ost, OST(filter) should check whether it
1359 should grant more space to client by comparing fed_grant and
1360 cl_avail_grant, and return the granted space to client instead
1361 of "new granted" space, because client will call osc_init_grant
1362 to update the client grant space info.
1365 Frequency : when client reconnect to OST
1367 Description: Grant Leak when osc do resend and replay bulk write
1368 Details : When osc reconnect to OST, OST(filter)should clear grant info of
1369 bulk write request, because the grant info while be sync between
1370 OSC and OST when reconnect, and we should ignore the grant info
1371 these of resend/replay write req.
1376 Description: Grant space more than avaiable left space sometimes.
1377 Details : When then OST is about to be full, if two bulk writing from
1378 different clients came to OST. Accord the avaliable space of the
1379 OST, the first req should be permitted, and the second one
1380 should be denied by ENOSPC. But if the seconde arrived before
1381 the first one is commited. The OST might wrongly permit second
1382 writing, which will cause grant space > avaiable space.
1385 Frequency : when client is evicted
1387 Description: Grant might be wrongly erased when osc is evicted by OST
1388 Details : when the import is evicted by server, it will fork another
1389 thread ptlrpc_invalidate_import_thread to invalidate the
1390 import, where the grant will be set to 0. While the original
1391 thread will update the grant it got when connecting. So if
1392 the former happened latter, the grant will be wrongly errased
1393 because of this race.
1398 Description: Checking Stale with correct fid
1399 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
1400 to check whether it is stale, because sometimes, we want the
1401 enqueue happened anyway, and op_data.fid2 will not be initialized.
1403 Severity : enhancement
1405 Description: update patchless client
1406 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
1409 Frequency : only with 2.4 kernel
1411 Description: random memory corruption
1412 Details : size of struct ll_inode_info is to big for union inode.u and this
1413 can be cause of random memory corruption.
1418 Description: Memory leak in recovery
1419 Details : Lov_mds_md was not free in an error handler in mds_create_object.
1420 It should also check obd_fail before fsfilt_start, otherwise if
1421 fsfilt_start return -EROFS,(failover mds during mds recovery).
1422 then the req will return with repmsg->transno = 0 and rc = EROFS.
1423 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
1424 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
1425 be freed no matter whether fsfilt_commit success or not.
1428 Frequency : only with huge count clients
1430 Description: Prevents from taking the superblock lock in llap_from_page for
1432 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
1433 for llap_from_page call in ll_removepage prevents from taking the
1434 superblock lock for a soon died page.
1439 Description: Not check open intent error before release open handle
1440 Details : in some rare cases, the open intent error is not checked before
1441 release open handle, which may cause
1442 ASSERTION(open_req->rq_transno != 0), because it tries to release
1443 the failed open handle.
1448 Description: Set cat log bitmap only after create log success.
1449 Details : in some rare cases, the cat log bitmap is set too early. and it
1450 should be set only after create log success.
1454 Description: Accessing a block bevice can re-enable I/O when Lustre is
1455 tearing down a device.
1456 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
1461 Description: service threads may hog cpus when there are a lot of requests
1462 Details : Insert cond_resched to give other threads a chance to use some CPU
1467 Description: the cat log was not initialized in recovery
1468 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
1469 unlink log on mds will not be initialized until mds post
1470 recovery. And also in mds post recovery, the unlink log will
1471 initialization will be done asynchronausly, so there will be race
1472 between add unlink log and unlink log initialization.
1476 Description: brw_stats were being printed incorrectly
1477 Details : brw_stats were being printed as log2 but all of them were not
1478 recorded as log2. Also remove some code duplication arising from
1479 filter_tally_{read,write}.
1483 Frequency : rare, only in recovery.
1484 Description: ASSERTION(req->rq_type != LI_POISON) failed
1485 Details : imp_lock should be held while iterating over imp_sending_list for
1486 prevent destroy request after get timeout in ptlrpc_queue_wait.
1490 Description: replay-single.sh test 52 fails
1491 Details : A lock's skiplist need to be cleanup when it being unlinked
1492 from its resource list.
1496 Description: Short directio read returns full requested size rather than
1498 Details : Direct I/O operations should return actual amount of bytes
1499 transferred rather than requested size.
1501 Severity : enhancement
1503 Description: metadata RPC reduction (e.g. for rm performance)
1504 Details : decrease the amount of synchronous RPC between clients and servers
1505 by canceling conflicing lock before the operation on the client
1506 and packing thier handles into the main operation RPC to server.
1508 Severity : enhancement
1510 Description: Async OSC create to avoid the blocking unnecessarily.
1511 Details : If a OST has no remain object, system will block on the creating
1512 when need to create a new object on this OST. Now, ways use
1513 pre-created objects when available, instead of blocking on an
1514 empty osc while others are not empty. If we must block, we block
1515 for the shortest possible period of time.
1517 Severity : enhancement
1519 Description: refine locking for avoid write wrong info into lov_objid file
1520 Details : fix possible races with add new target and write/update data in
1523 --------------------------------------------------------------------------------
1525 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
1530 Frequency : on some architectures
1532 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
1533 Details : Uninitialized flags sometimes cause configuration commands to
1537 Frequency : patchless clients only
1539 Description: missing __iget() symbol export
1540 Details : The __iget() symbol export is missing. To avoid the need for
1541 this on patchless clients the deathrow inode reaper is turned
1542 off, and we depend on the VM to clean up old inodes. This
1543 dependency was during via the fix for bug 12181.
1545 --------------------------------------------------------------------------------
1547 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
1549 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
1550 INTEROPERATE with older versions automatically. In many cases a
1551 special upgrade step is needed. Please read the
1552 user documentation before upgrading any part of a live system.
1553 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
1554 version will not interoperate with 1.6 betas before beta5 (1.5.95).
1555 * WARNING: Lustre configuration and startup changes are required with
1556 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
1562 Severity : enhancement
1564 Description: MountConf
1565 Details : Lustre configuration is now managed via mkfs and mount
1566 commands instead of lmc and lconf. New obd types (MGS, MGC)
1567 are added for dynamic configuration management. See
1568 https://mail.clusterfs.com/wikis/lustre/MountConf for
1571 Severity : enhancement
1573 Description: dynamic OST addition
1574 Details : OSTs can now be added to a live filesystem
1576 Severity : enhancement
1578 Description: startup order invariance
1579 Details : MDTs and OSTs can be started in any order. Clients only
1580 require the MDT to complete startup.
1582 Severity : enhancement
1584 Description: parallel, asynchronous orphan cleanup
1585 Details : orphan cleanup is now performed in separate threads for each
1586 OST, allowing parallel non-blocking operation.
1588 Severity : enhancement
1590 Description: optimized stripe assignment
1591 Details : stripe assignments are now made based on ost space available,
1592 ost previous usage, and OSS previous usage, in order to try
1593 to optimize storage space and networking resources.
1595 Severity : enhancement
1597 Description: Permanently set tunables
1598 Details : All writable /proc/fs/lustre tunables can now be permanently
1599 set on a per-server basis, at mkfs time or on a live system.
1601 Severity : enhancement
1603 Description: Lustre message v2
1604 Details : Add lustre message format v2.
1606 Severity : enhancement
1608 Description: client OST exclusion list
1609 Details : Clients can be started with a list of OSTs that should be
1610 declared "inactive" for known non-responsive OSTs.
1614 Description: SPEC SFS validation failure on NFS v2 over lustre.
1615 Details : Changes the blocksize for regular files to be 2x RPC size,
1616 and not depend on stripe size.
1618 Severity : enhancement
1620 Description: Multiple MD RPCs in flight.
1621 Details : Further unserialise some read-only MDS RPCs - learn about intents.
1622 To avoid overly-overloading MDS, introduce a limit on number of
1623 MDS RPCs in flight for a single client and add /proc controls
1624 to adjust this limit.
1626 Severity : enhancement
1628 Description: client read/write statistics
1629 Details : Add client read/write call usage stats for performance
1630 analysis of user processes.
1631 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
1632 file access. extents_stats shows chunk size distribution.
1633 extents_stats_per_process show chunk size distribution per
1636 Severity : enhancement
1638 Description: mds statistics
1639 Details : Add detailed mds operations statistics in
1640 /proc/fs/lustre/mds/*/stats.
1644 Description: Failure of copying files with lustre special EAs.
1645 Details : Client side always return success for setxattr call for lustre
1646 special xattr (currently only "trusted.lov").
1651 Description: Refcount LNET uuids
1652 Details : The global LNET uuid list grew linearly with every startup;
1653 refcount repeated list entries instead of always adding to
1656 Severity : enhancement
1658 Description: Dynamic service threads
1659 Details : Within a small range, start extra service threads
1660 automatically when the request queue builds up.
1663 Frequency : mixed-endian client/server environments
1665 Description: mixed-endian crashes
1666 Details : The new msg_v2 system had some failures in mixed-endian
1669 Severity : enhancement
1671 Description: Easy OST removal
1672 Details : OSTs can be permanently deactivated with e.g. 'lctl
1673 conf_param lustre-OST0001.osc.active=0'
1675 Severity : enhancement
1677 Description: MGS proc entries
1678 Details : Added basic proc entries for the MGS showing what filesystems
1681 Severity : enhancement
1683 Description: provide MGS failover
1684 Details : Added config lock reacquisition after MGS server failover.
1686 Severity : enhancement
1688 Description: add Linux 2.4 support
1689 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
1693 Description: a large application tries to do I/O to the same resource and dies
1694 in the middle of it.
1695 Details : Check the req->rq_arrival time after the call to
1696 ost_brw_lock_get(), but before we do anything about
1697 processing it & sending the BULK transfer request. This
1698 should help move old stale pending locks off the queue as
1699 quickly as obd_timeout.
1702 Frequency : when an incorrect nid is specified during startup
1704 Description: ptlrpc connect to non-existant node causes kernel crash
1705 Details : LNET can't be re-entered from an event callback, which
1706 happened when we expire a message after the export has been
1707 cleaned up. Instead, hand the zombie cleanup off to another
1710 Severity : enhancement
1712 Description: plain/inodebits lock performance improvement
1713 Details : Grouping plain/inodebits in granted list by their request modes
1714 and bits policy, thus improving the performance of search through
1718 Frequency : only if OST filesystem is corrupted
1720 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1721 Details : for a short time RPCs with bulk IO are in the replay list,
1722 but replay of bulk IOs is unimplemented. If the OST filesystem
1723 is corrupted due to disk cache incoherency and then replay is
1724 started it is possible to trip an assertion. Avoid putting
1725 committed RPCs into the replay list at all to avoid this issue.
1728 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1730 Description: System hang on startup
1731 Details : This bug allowed the liblustre (e.g. catamount) client to
1732 return to the app before handling all startup RPCs. This
1733 could leave the node unresponsive to lustre network traffic
1734 and manifested as a server ptllnd timeout.
1736 Severity : enhancement
1738 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1739 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
1740 boolean to control whether to print peer debug info when a
1741 client's RPC times out.
1744 Frequency : only for kernels with patches from Lustre below 1.4.3
1746 Description: Remove old rdonly API
1747 Details : Remove old rdonly API which unused from at least lustre 1.4.3
1750 Frequency : only for devices with external journals
1752 Description: Set external device read-only also
1753 Details : During a commanded failover stop, we set the disk device
1754 read-only while the server shuts down. We now also set any
1755 external journal device read-only at the same time.
1758 Frequency : when upgrading from 1.4 while trying to change parameters
1760 Description: The wrong (new) MDC name was used when setting parameters for
1761 upgraded MDT's. Also allows changing of OSC (and MDC)
1762 parameters if --writeconf is specified at tunefs upgrade time.
1765 Frequency : when setting specific ost indicies
1767 Description: QOS code breaks on skipped indicies
1768 Details : Add checks for missing OST indicies in the QOS code, so OSTs
1769 created with --index need not be sequential.
1771 Severity : enhancement
1773 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1774 Details : The uninit_groups feature works in conjunction with the kernel
1775 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1776 up the pass1 processing of e2fsck. This is a read-only feature
1777 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1778 mount filesystems that have had this feature enabled.
1780 Severity : enhancement
1782 Description: Improve multi-block allocation algorithm to avoid fragmentation
1783 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1784 allocation locality and avoid filesystem fragmentation.
1786 ------------------------------------------------------------------------------
1788 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
1790 * Support for kernels:
1791 2.4.21-47.0.1.EL (RHEL 3)
1792 2.6.5-7.283 (SLES 9)
1793 2.6.9-42.0.10.EL (RHEL 4)
1794 2.6.12.6 vanilla (kernel.org)
1795 2.6.16.27-0.9 (SLES 10)
1796 * Recommended e2fsprogs version: 1.39.cfs5
1798 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
1802 Frequency : occasional, depends on client load and configuration
1803 Bugzilla : 12181, 12203
1804 Description: data loss for recently-modified files
1806 Details : In some cases it is possible that recently written or created
1807 files may not be written to disk in a timely manner (this should
1808 normally be within 30s unless client IO load is very high).
1809 The problem appears as zero-length files or files that are a
1810 multiple of 1MB in size after a client crash or client eviction
1811 that are missing data at the end of the file.
1813 This problem is more likely to be hit on clients where files are
1814 repeatedly created and unlinked in the same directory, clients
1815 have a large amount of RAM, have many CPUs, the filesystem has
1816 many OSTs, the clients are rebooted frequently, and/or the files
1817 are not accessed by other nodes after being written.
1819 The presence of the problem can be detected by looking at
1820 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
1821 smaller than the second (nr_unused) then dirty files will not
1822 be flushed automatically to disk. "sync; sleep 10" should be
1823 run several times on the node before unmounting it to update
1824 Lustre (this is also safe to run on nodes without this problem).
1826 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
1827 that can cause this same problem, so customers using that kernel
1828 also need to update the kernel in addition to Lustre. In order
1829 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
1831 It is normal that files written just before a client crash (less
1832 than 30s) may not yet have been flushed to disk, even for local
1836 Frequency : frequent on thin XT3 nodes
1838 Description: UUID collision on thin XT3 Linux nodes
1839 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
1840 randomly, since we relied on an insufficiently-seeded PRNG.
1845 Description: OSS hangs after "All ost request buffers busy"
1846 Details : A deadlock between quota and journal operations caused OSS
1847 hangs after printing "All ost request buffers busy."
1850 Frequency : always on liblustre builds
1852 Description: Cleanup compiler warnings on liblustre
1855 Frequency : always on liblustre builds on XT3
1857 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
1861 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
1862 > 32000 subdirectories
1863 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
1864 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
1865 subdirectory count crosses 32000. This will aid e2fsck to
1866 correctly handle more than 32000 subdirectories.
1869 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
1871 Description: System hang on startup
1872 Details : This bug allowed the liblustre (e.g. catamount) client to
1873 return to the app before handling all startup RPCs. This
1874 could leave the node unresponsive to lustre network traffic
1875 and manifested as a server ptllnd timeout.
1877 Severity : enhancement
1879 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
1880 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
1881 boolean to control whether to print peer debug info when a
1882 client's RPC times out.
1887 Description: make O_SYNC working on 2.6 kernels
1888 Details : 2.6 kernels use different method for mark pages for write,
1889 so need add a code to lustre for O_SYNC work.
1894 Description: Failure to close file and release space on NFS
1895 Details : Put inode details into lock acquired in ll_intent_file_open.
1896 Use mdc_intent_lock in ll_intent_open to properly
1897 detect all kind of errors unhandled by mdc_enqueue.
1902 Description: proc file read during shutdown sometimes raced obd removal,
1904 Details : Add lock to prevent obd access after proc file removal.
1907 Frequency : Only for files larger than 4GB on 32-bit clients.
1909 Description: improperly doing page alignment of locks
1910 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
1911 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
1916 Description: RPCs being resent when they shouldn't be
1917 Details : Some RPCs that should not be resent are being resent. This
1918 can cause inconsistencies in the RPC state machine. Do not
1919 resend such requests.
1922 Frequency : rare, only with NFS export
1924 Description: Crash on NFS re-export node
1925 Details : under very unusual load conditions an assertion is hit in
1926 ll_intent_file_open()
1929 Frequency : only if OST filesystem is corrupted
1931 Description: client incorrectly hits assertion in ptlrpc_replay_req()
1932 Details : for a short time RPCs with bulk IO are in the replay list,
1933 but replay of bulk IOs is unimplemented. If the OST filesystem
1934 is corrupted due to disk cache incoherency and then replay is
1935 started it is possible to trip an assertion. Avoid putting
1936 committed RPCs into the replay list at all to avoid this issue.
1941 Description: large O_DIRECT requests fail under memory pressure/fragmentation
1942 Details : Large single O_DIRECT read and write calls can fail to allocate
1943 a sufficiently large buffer to process the request. In case of
1944 allocation failure the allocation is retried with a smaller
1945 buffer and broken into smaller requests.
1947 Severity : enhancement
1949 Description: Add -o localflock option to simulate old noflock behaviour.
1950 Details : This will achieve local-only flock/fcntl locks coherentness.
1955 Description: versioning check is incomplete
1956 Details : Checking the version difference of client vs. server, report
1957 error if the gap is too big.
1962 Description: add support PG_writeback bit
1963 Details : add support for PG_writeback bit for Lustre, for more carefull
1964 work with page cache in 2.6 kernel. This also fix some deadlocks
1965 and remove hack for work O_SYNC with 2.6 kernel.
1967 Severity : enhancement
1969 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
1970 Details : The uninit_groups feature works in conjunction with the kernel
1971 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
1972 up the pass1 processing of e2fsck. This is a read-only feature
1973 in ldiskfs2 only, so older kernels and current ldiskfs cannot
1974 mount filesystems that have had this feature enabled.
1976 Severity : enhancement
1978 Description: Improve multi-block allocation algorithm to avoid fragmentation
1979 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
1980 allocation locality and avoid filesystem fragmentation.
1982 ------------------------------------------------------------------------------
1984 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
1986 * Support for kernels:
1987 2.6.9-42.0.3.EL (RHEL 4)
1988 2.6.5-7.276 (SLES 9)
1989 2.4.21-47.0.1.EL (RHEL 3)
1990 2.6.12.6 vanilla (kernel.org)
1991 2.6.16.21-0.8 (SLES10)
1992 * Recommended e2fsprogs version: 1.39.cfs2-0
1994 * The backwards-compatible /proc/sys/portals symlink has been removed
1995 in this release. Before upgrading, please ensure that you change
1996 any configuration scripts or /etc/sysctl.conf files that access
1997 /proc/sys/portals/* or sysctl portals.* to use the corresponding
1998 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
1999 in advance of the upgrade on any system running Lustre 1.4.6 or
2000 newer, since /proc/sys/lnet was added in that version.
2001 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2005 Frequency : only when quota is used
2007 Description: avoid scanning export list for quota master
2008 Details : Change the algorithms to avoid scanning export list in order
2009 to improve the efficiency.
2012 Frequency : MDS failover only, very rarely
2014 Description: "went back in time" messages on mds failover
2015 Details : The greatest transno may be lost when the current operation
2016 finishes with an error (transno==0) and the client's last_rcvd
2017 record is over-written. Save the greatest transno in the
2018 mds_last_transno for this case.
2021 Frequency : always for specific kernels and striping counts
2023 Description: client may get "Matching packet too big" without ACL support
2024 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
2025 when trying to access files in certain configurations. The
2026 clients should in fact be denied when mounting because they do
2027 not understand ACLs.
2030 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
2032 Description: many clients connecting with IO in progress causes connect timeouts
2033 Details : Avoid synchronous journal commits to avoid delays caused by many
2034 clients connecting/disconnecting when bulk IO is in progress.
2035 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
2036 OST_IO_PORTAL to avoid delays behind potentially many pending
2040 Frequency : occasionally with multiple writers to a single file
2042 Description: shared writes to file may result in wrong size reported by stat()
2043 Details : Allow growing of kms when extent lock is cancelled
2046 Frequency : always with random mmap IO to multi-striped file
2048 Description: mmap write might be lost if we are writing to a 'hole' in stripe
2049 Details : Only if the hole is at the end of OST object so that kms is too
2050 small. Fix is to increase kms accordingly in ll_nopage.
2053 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
2055 Description: writes to a missing object would leak memory on the OST
2056 Details : If there is an inconsistency between the MDS and OST filesystems,
2057 such that the MDS references an object that doesn't exist, writes
2058 to that object will leak memory due to incorrect cleanup in the
2059 error handling path, eventually running out of memory on the OST.
2064 Description: Creating too long symlink causes lustre errors
2065 Details : Check symlink and name lengths before sending requests to MDS.
2068 Frequency : only if flock is enabled (not on by default)
2070 Description: posix locks not released on fd closure on 2.6.9+
2071 Details : We failed to add posix locks to list of inode locks on 2.6.9+
2072 kernels, this caused such locks not to be released on fd close and
2073 then assertions on fs unmount about still used locks.
2076 Frequency : MDS failover only, very rarely
2078 Description: clients may get ASSERTION(granted_lock != NULL)
2079 Details : When request was taking a long time, and a client was resending
2080 a getattr by name lock request. The were multiple lock requests
2081 with the same client lock handle and
2082 mds_getattr_name->fixup_handle_for_resent_request found one of the
2083 lock handles but later failed with ASSERTION(granted_lock != NULL).
2088 Description: handle->h_buffer_credits > 0, assertion failure
2089 Details : h_buffer_credits is zero after truncate, causing assertion
2090 failure. This patch extends the transaction or creates a new
2094 Frequency : NFS re-export or patchless client
2095 Bugzilla : 11179, 10796
2096 Description: Crash on NFS re-export node (__d_move)
2097 Details : We do not want to hash the dentry if we don't have a lock.
2098 But if this dentry is later used in d_move, we'd hit uninitialised
2099 list head d_hash, so we just do this to init d_hash field but
2100 leave dentry unhashed.
2103 Frequency : NFS re-export or patchless client
2105 Description: NFS exports has problem with symbolic link
2106 Details : lustre client didn't properly install dentry when re-exported
2107 to NFS or running patchless client.
2110 Frequency : NFS re-export or patchless client
2112 Description: Various nfs/patchless fixes.
2113 Details : fixes reuse disconected alias for lookup process - this fixes
2114 warning "find_exported_dentry: npd != pd",
2115 fix permission error with open files at nfs.
2116 fix apply umask when do revalidate.
2119 Frequency : occasional
2121 Description: Crash on NFS re-export node
2122 Details : calling clear_page() on the wrong pointer triggered oops in
2123 generic_mapping_read().
2126 Frequency : rarely, using O_DIRECT IO
2128 Description: unaligned directio crashes client with LASSERT
2129 Details : check for unaligned buffers before trying any requests.
2132 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
2134 Description: stale data returned from RAID cache
2135 Details : If only a small amount of IO is done to the RAID device before
2136 reading it again it is possible to get stale data from the RAID
2137 cache instead of reading it from disk.
2140 Frequency : always for sles10 kernel
2142 Description: sles10 support
2143 Details : ll_follow_link: compile fixes and using of nd_set_link
2144 under newer kernels.
2147 Frequency : depends on arch, kernel and compiler version, always on sles10
2150 Description: recursive or deep enough symlinks cause stack overflow
2151 Details : getting rid of large stack-allocated variable in
2155 Frequency : depends on hardware
2157 Description: lustre write performance loss in the SLES10 kernel
2158 Details : the performance loss is caused by using of write barriers in the
2159 ext3 code. The SLES10 kernel turns barrier support on by
2160 default. The fix is to undo that change for ldiskfs.
2162 ------------------------------------------------------------------------------
2164 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
2166 * Support for kernels:
2167 2.6.9-42.0.3EL (RHEL 4)
2168 2.6.5-7.276 (SLES 9)
2169 2.4.21-47.0.1.EL (RHEL 3)
2170 2.6.12.6 vanilla (kernel.org)
2174 Frequency : quota enabled and large files being deleted
2176 Description: releasing more than 4GB of quota at once hangs OST
2177 Details : If a user deletes more than 4GB of files on a single OST it
2178 will cause the OST to spin in an infinite loop. Release
2179 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2184 Description: statfs data retrieved from /proc may be stale or zero
2185 Details : When reading per-device statfs data from /proc, in the
2186 {kbytes,files}_{total,free,avail} files, it may appear
2187 as zero or be out of date.
2190 Frequency : systems with MD RAID1 external journal devices
2192 Description: lconf's call to blkid is confused by RAID1 journal devices
2193 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2194 of returning all block devices that match the journal UUID.
2197 Frequency : always, for aggregate stripe size over 4GB
2199 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
2200 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
2201 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
2204 Frequency : always if "lfs find" used on a local file/directory
2206 Description: "lfs find" segfaults if used on a local file/directory
2207 Details : The case where a directory component was not specified wasn't
2208 handled correctly. Handle this properly.
2211 Frequency : always on ppc64
2213 Description: the write to an ext3 filesystem mounted with mballoc got stuck
2214 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2215 perform endianness conversion.
2218 Frequency : rarely (truncate to non-zero file size after write under load)
2219 Bugzilla : 10730, 10687
2220 Description: Files padded with zeros to next 4K multiple
2221 Details : With filesystems mounted using the "extents" option (2.6 kernels)
2222 it is possible that files that are truncated to a non-zero size
2223 immediately after being written are filled with zero bytes beyond
2224 the truncated size. No file data is lost.
2226 Severity : enhancement
2228 Description: Allow recovery/failover for liblustre clients.
2229 Details : liblustre clients were unaware of failover configurations until
2232 Severity : enhancement
2234 Description: user file locks should fail when not mounting with flock option
2235 Details : Set up an error-returning stub in ll_file_operations.lock field
2236 to prevent incorrect behaviour when client is mounted without
2237 flock option. Also, set up properly f_op->flock field for
2241 Frequency : always on ia64
2243 Description: "lfs df" loops on printing out MDS statfs information
2244 Details : The obd_ioctl_data was not initialized and in some systems
2245 this caused a failure during the ioctl that did not return
2246 an error. Initialize the struct and return an error on failure.
2249 Frequency : SLES 9 only
2251 Description: Error of copying files with lustre special EAs as root
2252 Details : Client side always return success for setxattr call for lustre
2253 special xattr (currently only "trusted.lov").
2256 Frequency : rarely on clusters with both ia64+i386 clients
2258 Description: ia64+i686 clients doing shared IO on the same file may LBUG
2259 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
2260 clients are doing concurrent writes to the same file it is
2261 possible that the ia64 clients may LASSERT because the OST
2262 extent locks are not PAGE_SIZE aligned. Ensure that grown
2263 locks are always aligned on the request boundary.
2266 Frequency : specific use, occasional
2268 Description: Overwriting in use executable truncates on-disk binary image
2269 Details : If one node attempts to overwrite an executable in use by
2270 another node, we now correctly return ETXTBSY instead of
2271 truncating the file.
2273 Severity : enhancement
2275 Description: Async OSC create to avoid the blocking unnecessarily.
2276 Details : If a OST has no remain object, system will block on the creating
2277 when need to create a new object on this OST. Now, ways use
2278 pre-created objects when available, instead of blocking on an
2279 empty osc while others are not empty. If we must block, we block
2280 for the shortest possible period of time.
2285 Description: chmod on Lustre root is propagated to other clients
2286 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
2287 invalid by the follow_mount time.
2290 Frequency : liblustre clients only
2292 Description: Race in 'instant cancel' lock handling could lead to such locks
2293 never to be granted in case of SMP MDS
2294 Details : Do not destroy not yet granted but cbpending locks in
2298 Frequency : replay/resend of open
2300 Description: non null lock assetion failure in mds_intent_policy
2301 Details : Trying to replay/resend lockless open requests resulted in
2302 mds_open() returning 0 with no lock. Now it sets a flag if
2303 a lock is going to be returned.
2305 Severity : enhancement
2307 Description: Checksum enhancements
2308 Details : New checksum enhancements allow for resending RPCs that failed
2311 Severity : enhancement
2313 Description: Tunables on number of dirty pages in cacche
2314 Details : Allow to set limit on number of dirty pages cached.
2319 Description: client crash on unmount - lock still has references
2320 Details : In some error handling cases it was possible to leak a lock
2321 reference on a client while accessing a file. This was not
2322 harmful to the client during operation, but would cause the
2323 client to crash when the filesystem is unmounted.
2326 Frequency : specific case, rare
2328 Description: ETXTBSY on mds though file not in use by client
2329 Details : ETXTBSY is no longer incorrectly returned when attempting to
2330 chmod or chown a directory that the user previously tried to
2331 execute or a currently-executing binary.
2334 Frequency : extremely rare except on liblustre-based clients
2336 Description: Lustre space not freed when files are deleted
2337 Details : Clean up open-unlinked files after client eviction. Previously
2338 the unlink was skipped and the files remained as orphans.
2343 Description: OST failure "would be an LBUG" in waiting_locks_callback()
2344 Details : In some cases it was possible to send a blocking callback to
2345 a client doing a glimpse, even though that client didn't get
2346 a lock granted. When the glimpse lock is cancelled on the OST
2347 the freed lock is left on the waiting list and corrupted the list.
2350 Frequency : all core dumps
2352 Description: Broke core dumps to lustre
2353 Details : Negative dentry may be unhashed if parent does not have UPDATE
2354 lock, but some callers, e.g. do_coredump, expect dentry to be
2355 hashed after successful create, hash it in ll_create_it.
2357 ------------------------------------------------------------------------------
2359 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
2361 * Support for kernels:
2362 2.6.9-42.0.2.EL (RHEL 4)
2363 2.6.5-7.276 (SLES 9)
2364 2.4.21-40.EL (RHEL 3)
2365 2.6.12.6 vanilla (kernel.org)
2369 Frequency : always on RHEL 3
2371 Description: Number of open files grows over time
2372 Details : The number of open files grows over time, whether or not
2373 Lustre is started. This was due to a filp leak introduced
2374 by one of our kernel patches.
2376 ------------------------------------------------------------------------------
2378 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2380 * Support for kernels:
2381 2.6.9-42.EL (RHEL 4)
2382 2.6.5-7.276 (SLES 9)
2383 2.4.21-40.EL (RHEL 3)
2384 2.6.12.6 vanilla (kernel.org)
2389 Bugzilla : 5719, 9635, 9792, 9684
2390 Description: OST (or MDS) trips assertions in (re)connection under heavy load
2391 Details : If a server is under heavy load and cannot reply to new
2392 connection requests before the client resends the (re)connect,
2393 the connection handling code can behave badly if two service
2394 threads are concurrently handing separate (re)connections from
2395 the same client. Add better locking to the connection handling
2396 code, and ensure that only a single connection will be processed
2397 for a given client UUID, even if the lock is dropped.
2399 Severity : enhancement
2401 Description: add TCP zero-copy support to kernel
2402 Details : Add support to the kernel TCP stack to allow zero-copy bulk
2403 sends if the hardware supports scatter-gather and checksumming.
2404 This allows socklnd to do client-write and server-read more
2405 efficiently and reduce CPU utilization from skbuf copying.
2408 Frequency : only if NFS exporting from client
2410 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
2411 Details : If setting attributes on a file created under NFS that had
2412 never been opened it would be possible to oops the client
2413 if the file had no objects.
2416 Frequency : always for liblustre
2418 Description: liblustre client does MDS+OSTs setattr RPC for each write
2419 Details : When doing a write from a liblustre client, the client
2420 incorrectly issued an RPC to the MDS and each OST the file was
2421 striped over in order to update the timestamps. When writing
2422 with small chunks and many clients this could overwhelm the MDS
2423 with RPCs. In all cases it would slow down the write because
2424 these RPCs are unnecessary.
2426 Severity : enhancement
2428 Description: allow number of MDS service threads to be changed at module load
2429 Details : It is now possible to change the number of MDS service threads
2430 running. Adding "options mds mds_num_threads={N}" to the MDS's
2431 /etc/modprobe.conf will set the number of threads for the next
2432 time Lustre is restarted (assuming the "mds" module is also
2433 reloaded at that time). The default number of threads will
2434 stay the same, 32 for most systems.
2439 Description: OST crash if filesystem is unformatted or corrupt
2440 Details : If an OST is started on a device that has never been formatted
2441 or if the filesystem is corrupt and cannot even mount then the
2442 error handling cleanup routines would dereference a NULL pointer.
2447 Description: NULL pointer deref in llap_from_page.
2448 Details : get_cache_page_nowait can return a page with NULL (or otherwise
2449 incorrect) mapping if the page was truncated/reclaimed while it was
2450 searched for. Check for this condition and skip such pages when
2451 doing readahead. Introduce extra check to llap_from_page() to
2452 verify page->mapping->host is non-NULL (so page is not anonymous).
2455 Frequency : Sometimes when using sys_sendfile
2457 Description: "page not covered by a lock" warnings from ll_readpage
2458 Details : sendfile called ll_readpage without right page locks present.
2459 Now we introduced ll_file_sendfile that does necessary locking
2460 around call to generic_file_sendfile() much like we do in
2464 Frequency : with certain MDS communication failures at client mount time
2466 Description: NULL pointer deref after failed client mount
2467 Details : a client connection request may delayed by the network layer
2468 and not be sent until after the PTLRPC layer has timed out the
2469 request. If the client fails the mount immediately it will try
2470 to clean up before the network times out the request. Add a
2471 reference from the request import to the obd device and delay
2472 the cleanup until the network drops the request.
2475 Frequency : occasionally during client (re)connect
2477 Description: assertion failure during client (re)connect
2478 Details : processing a client connection request may be delayed by the
2479 client or server longer than the client connect timeout. This
2480 causes the client to resend the connection request. If the
2481 original connection request is replied in this interval, the
2482 client may trip an assertion failure in ptlrpc_connect_interpret()
2483 which thought it would be the only running connect process.
2486 Frequency : only with obd_echo servers and clients that are rebooted
2488 Description: kernel BUG accessing uninitialized data structure
2489 Details : When running an obd_echo server it did not start the ping_evictor
2490 thread, and when a client was evicted an uninitialized data
2491 structure was accessed. Start the ping_evictor in the RPC
2492 service startup instead of the OBD startup.
2494 Severity : enhancement
2495 Bugzilla : 10193 (patchless)
2496 Description: Remove dependency on various unexported kernel interfaces.
2497 Details : No longer need reparent_to_init, exit_mm, exit_files,
2498 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
2501 Frequency : rare (only users of deprecated and unsupported LDAP config)
2503 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
2504 Details : LDAP apparently contains 'lustreName' attributes instead of
2505 'name'. A simple remapping of the name is sufficient.
2508 Frequency : rare (only with non-default dump_on_timeout debug enabled)
2510 Description: waiting_locks_callback trips kernel BUG if client is evicted
2511 Details : Running with the dump_on_timeout debug flag turned on makes
2512 it possible that the waiting_locks_callback() can try to dump
2513 the Lustre kernel debug logs from an interrupt handler. Defer
2514 this log dumping to the expired_lock_main() thread.
2516 Severity : enhancement
2518 Description: Support NFS exporting on 2.6 kernels.
2519 Details : Implement non-rawops metadata methods for NFS server to use without
2520 changing NFS server code.
2523 Frequency : very rare (synthetic metadata workload only)
2525 Description: two racing renames might cause an MDS thread to deadlock
2526 Details : Running the "racer" program may cause one MDS thread to rename
2527 a file from being the source of a rename to being the target of
2528 a rename at exactly the same time that another thread is doing
2529 so, and the second thread has already enqueued these locks after
2530 doing a lookup of the target and is trying to relock them in
2531 order. Ensure that we don't try to re-lock the same resource.
2534 Frequency : only very large systems with liblustre clients
2536 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
2537 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
2538 all OSTs in parallel. This allows the eviction of stale liblustre
2539 clients to proceed much faster than if they were done in series,
2540 and also offers similar improvements for other set_info RPCs.
2545 Description: excessive CPU usage during initial read phase on client
2546 Details : During the initial read phase on a client, it would agressively
2547 retry readahead on the file, consuming too much CPU and impacting
2548 performance (since 1.4.5.8). Improve the readahead algorithm
2549 to avoid this, and also improve some other common cases (read
2550 of small files in particular, where "small" is files smaller than
2551 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
2556 Description: MDS crash when receiving packet with unknown intent.
2557 Details : Do not LBUG in unknown intent case, just return -EFAULT
2559 Severity : enhancement
2560 Bugzilla : 9293, 9385
2561 Description: MDS RPCs are serialised on client. This is unnecessary for some.
2562 Details : Do not serialize getattr (non-intent version) and statfs.
2565 Frequency : occasional, when OST network is overloaded/intermittent
2567 Description: client evicted by OST after bulk IO timeout
2568 Details : If a client sends a bulk IO request (read or write) the OST
2569 may evict the client if it is unresposive to its data GET/PUT
2570 request. This is incorrect if the network is overloaded (takes
2571 too long to transfer the RPC data) or dropped the OST GET/PUT
2572 request. There is no need to evict the client at all, since
2573 the pinger and/or lock callbacks will handle this, and the
2574 client can restart the bulk request.
2577 Frequency : Always when mmapping file with no objects
2579 Description: client crashes when mmapping file with no objects
2580 Details : Check that we actually have objects in a file before doing any
2581 operations on objects in ll_vm_open, ll_vm_close and
2587 Description: Request leak when working with deleted CWD
2588 Details : Introduce advanced request refcount tracking for requests
2589 referenced from lustre intent.
2591 Severity : Enhancement
2593 Description: Cache open file handles on client.
2594 Details : MDS now will return special lock along with openhandle, if
2595 requested and client is allowed to hold openhandle, even if unused,
2596 until such a lock is revoked. Helps NFS a lot, since NFS is opening
2597 closing files for every read/write openration.
2599 Severity : Enhancement
2601 Description: Cache open negative dentries on client when possible.
2602 Details : Guard negative dentries with UPDATE lock on parent dir, drop
2603 negative dentries on lock revocation.
2608 Description: Remounting a client read-only wasn't possible with a zconf mount
2609 Details : It wasn't possible to remount a client read-only with llmount.
2611 Severity : enhancement
2612 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
2613 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
2614 MPICH is included to add Lustre-specific ADIO interfaces.
2615 This is based closely on the UFS ADIO layer and only differs
2616 in file creation, in order to allow the OST striping to be set.
2617 This is user-contributed code and not supported by CFS.
2622 Description: extended inode attributes (immutable, append-only) work improperly
2623 when 2.4 and 2.6 kernels are used on client/server or vice versa
2624 Details : Introduce kernel-independent values for these flags.
2626 Severity : enhancement
2629 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
2630 Details : Many of the /proc/ tunables can only be tuned at a megabyte
2631 granularity. Now, Fractional MB granularity is be supported,
2632 this is very useful for low memory system.
2634 Severity : enhancement
2636 Description: Getattr by fid
2637 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
2638 locks, avoids extra getattr rpc requests to MDS, allows '/' to
2639 have locks and avoids getattr rpc requests for it on every stat.
2642 Frequency : Always, for filesystems larger than 2TB
2644 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2645 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
2646 filesystems larger than 2TB to be mounted with mballoc enabled.
2649 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
2651 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
2652 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
2653 filesystems larger than 2TB is mounted then the kernel will
2654 silently corrupt the start of the filesystem. CONFIG_LBD is
2655 enabled for all CFS-supported kernels, but the possibility of
2656 this happening with a modified kernel config exists.
2658 Severity : enhancement
2660 Description: add client O_DIRECT support for 2.6 kernels
2661 Details : It is now possible to do O_DIRECT reads and writes to files
2662 in the Lustre client mountpoint on 2.6 kernel clients.
2664 Severity : enhancement
2666 Description: parallel glimpse, setattr, statfs, punch, destroy requests
2667 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
2668 parallel, not waiting for response from every OST before sending
2669 a rpc to the next OST.
2674 Description: setattr vs write race when updating file timestamps
2675 Details : Client processes that update a file timestamp into the past
2676 right after writing to the file (e.g. tar) it is possible that
2677 the updated file modification time can be reset to the current
2678 time due to a race between processing the setattr and write RPC.
2680 Severity : enhancement
2682 Description: Bring 'lfs find' closer in line with regular Linux find.
2683 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
2684 -print0 options and obtains all the needed info through the lustre
2687 Severity : enhancement
2689 Description: support up to 1024 configured devices on one node
2690 Details : change obd_dev array from statically allocated to dynamically
2691 allocated structs as they are first used to reduce memory usage
2696 Description: Flush dirty partially truncated pages during truncate
2697 Details : Immediatelly flush partially truncated pages in filter_setattr,
2698 this way we completely avoid having any pages in page cache on OST
2699 and can retire ugly workarounds during writes to flush such pages.
2704 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
2705 Details : protect inode from truncation within vfs_unlink() context
2706 just take a reference before calling vfs_unlink() and release it
2707 when parent's i_sem is free.
2712 Description: last_id value checked outside lock on OST caused LASSERT failure
2713 Details : If there were multiple MDS->OST object precreate requests in
2714 flight, it was possible that the OST's last object id was checked
2715 outside a lock and incorrectly tripped an assertion. Move checks
2716 inside locks, and discard old precreate requests.
2719 Frequency : always, if extents are used on OSTs
2721 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
2722 Details : OSTs using the extents format would not zero the high 16 bits of
2723 the index physical block number. This is not a problem for any
2724 OST filesystems smaller than 16TB, and no kernels support ext3
2725 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
2726 new/modified files) and can be fixed for existing filesystems
2727 with e2fsprogs-1.39-cfs1.
2732 Description: import connection selection may be incorrect if timer wraps
2733 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
2734 import connections to be ignored if the 32-bit jiffies counter
2735 wraps. Use a 64-bit jiffies counter.
2738 Frequency : very large clusters immediately after boot
2740 Description: LNET request buffers exhausted under heavy short-term load
2741 Details : If a large number of client requests are generated on a service
2742 that has previously never seen so many requests it is possible
2743 that the request buffer growth cannot keep up with the spike in
2744 demand. Instead of dropping incoming requests, they are held in
2745 the LND until the RPC service can accept more requests.
2748 Frequency : Sometimes during replay
2750 Description: Assertion failure in ll_local_open after replay.
2751 Details : If replay happened on an open request reply before we were able
2752 to set replay handler, reply will become not swabbed tripping the
2753 assertion in ll_local_open. Now we set the handler right after
2754 recognising of open request
2757 Frequency : very rare
2759 Description: kernel reports "badness in vsnprintf"
2760 Details : Reading from the "recovery_status" /proc file in small chunks
2761 may cause a negative length in lprocfs_obd_rd_recovery_status()
2762 call to vsnprintf() (which is otherwise harmless). Exit early
2763 if there is no more space in the output buffer.
2765 Severity : enhancement
2767 Description: clear OBD RPC statistics by writing to them
2768 Details : It is now possible to clear the OBD RPC statistics by writing
2769 to the "stats" file.
2774 Description: Client mtime is not the same on different clients after utimes
2775 Details : In some cases, the client was using the utimes() syscall on
2776 a file cached on another node. The clients now validate the
2777 ctime from the MDS + OSTs to determine which one is right.
2782 Description: Inability to activate failout mode
2783 Details : lconf script incorrectly assumed that in pythong string's numeric
2784 value is used in comparisons.
2787 Frequency : always with multiple stripes per file
2789 Description: Inefficient object allocation for mutli-stripe files
2790 Details : When selecting which OSTs to stripe files over, for files with
2791 a stripe count that divides evenly into the number of OSTs,
2792 the MDS is always picking the same starting OST for each file.
2793 Return the OST selection heuristic to the original design.
2798 Description: mount failures may take full timeout to return an error
2799 Details : Under some heavy load conditions it is possible that a
2800 failed mount can wait for the full obd_timeout interval,
2801 possibly several minutes, before reporting an error.
2802 Instead return an error as soon as the status is known.
2804 Frequency : quota enabled and large files being deleted
2806 Description: releasing more than 4GB of quota at once hangs OST
2807 Details : If a user deletes more than 4GB of files on a single OST it
2808 will cause the OST to spin in an infinite loop. Release
2809 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
2814 Description: statfs data retrieved from /proc may be stale or zero
2815 Details : When reading per-device statfs data from /proc, in the
2816 {kbytes,files}_{total,free,avail} files, it may appear
2817 as zero or be out of date.
2820 Frequency : systems with MD RAID1 external journal devices
2822 Description: lconf's call to blkid is confused by RAID1 journal devices
2823 Details : Use the "blkid -l" flag to locate the MD RAID device instead
2824 of returning all block devices that match the journal UUID.
2827 Frequency : always, for aggregate stripe size over 4GB
2829 Description: assertion fails when trying to use 4GB stripe size
2830 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
2831 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
2834 Frequency : always on ppc64
2836 Description: the first write on an ext3 filesystem with mballoc got stuck
2837 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
2838 perform endianness conversion.
2840 ------------------------------------------------------------------------------
2842 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
2844 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
2845 INTEROPERATE with older versions automatically. Please read the
2846 user documentation before upgrading any part of a live system.
2847 * WARNING: Lustre networking configuration changes are required with
2848 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
2851 * Support for kernels:
2852 2.6.9-22.0.2.EL (RHEL 4)
2853 2.6.5-7.244 (SLES 9)
2854 2.6.12.6 vanilla (kernel.org)
2857 Severity : enhancement
2858 Bugzilla : 7981/8208
2859 Description: Introduced Lustre Networking (LNET)
2860 Details : LNET is new networking infrastructure for Lustre, it includes
2861 a reorganized network configuration mode (see the user
2862 documentation for full details) as well as support for routing
2863 between different network fabrics. Lustre Networking Devices
2864 (LNDs) for the supported network fabrics have also been
2865 created for this new infrastructure.
2867 Severity : enhancement
2868 Description: Introduced Access control lists
2869 Details : clients can set ACLs on files and directories in order to have
2870 more fine-grained permissions than the standard Unix UGO+RWX.
2871 The MDS must be started with the "-o acl" mount option.
2873 Severity : enhancement
2874 Description: Introduced filesystem quotas
2875 Details : Administrators may now establish per-user quotas on the
2878 Severity : enhancement
2880 Description: Configuration change for the XT3
2881 The PTLLND is now used to run Lustre over Portals on the XT3
2882 The configure option(s) --with-cray-portals are no longer used.
2883 Rather --with-portals=<path-to-portals-includes> is used to
2884 enable building on the XT3. In addition to enable XT3 specific
2885 features the option --enable-cray-xt3 must be used.
2890 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
2891 Details : A race between allocating a new llcd and re-getting the llcd_lock
2892 allowed another thread to grab newly-allocated llcd.
2894 Severity : enhancement
2896 Description: 2.6 OST async journal commit and locking fix to improve performance
2897 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
2898 2.6 kernels are now async as they already were in 2.4 kernels so
2899 that they can commit concurrently with the network bulk transfer.
2900 For block-allocated files the filter allocation semaphore is held
2901 to avoid filesystem fragmentation during allocation. BKL lock
2902 removed for 2.6 xattr operations where it is no longer needed.
2907 Description: lconf incorrectly determined whether two IP networks could talk
2908 Details : In some more complicated routing and multiple-network
2909 configurations, lconf will avoid trying to make a network
2910 connection to a disjoint part of the IP space. It was doing the
2911 math incorrectly for one set of cases.
2916 Description: Fix for potential infinite loop processing records in an llog.
2917 Details : If an llog record is corrupted/zeroed, it is possible to loop
2918 forever in llog_process(). Validate the llog record length
2919 and skip the remainder of the block on error.
2922 Frequency : occasional (liblustre only)
2924 Description: liblustre could not open files whose last component is a symlink
2925 Details : sysio_path_walk() would incorrectly pass the open intent to
2926 intermediate path components.
2929 Frequency : rare (liblustre only with non-standard tuning)
2930 Bugzilla : 7201 (7350)
2931 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
2932 Details : llu_lookup_finish_locks() tries to set lock data on a lock
2933 after it has been released, only do this for referenced locks
2935 Severity : enhancement
2937 Description: specifying an (invalid) directory default stripe_size of -1
2938 would reset the directory default striping
2939 Details : stripe_size -1 was used internally to signal directory stripe
2940 removal, now use "all default" to signal dir stripe removal
2941 as a directory striping of "all default" is not useful
2944 Frequency : common for large clusters running liblustre clients
2946 Description: doing an ls when liblustre clients are running is slow
2947 Details : sending a glimpse AST to a liblustre client waits for every AST
2948 to time out, as liblustre clients will not respond. Since they
2949 cannot cache data we refresh the OST lock LVB from disk instead.
2951 Severity : enhancement
2953 Description: doing an ls at the same time as file IO can be slow
2954 Details : enqueue and other "small" requests can be blocked behind many
2955 large IO requests. Create a new OST IO portal for non-IO
2956 requests so they can be processed faster.
2959 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
2961 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
2962 created file with mode 000
2963 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
2964 allow an ftruncate() as a non-root user to a file with mode 000.
2965 Lustre now respects this flag to disable mode checks when
2966 truncating a file owned by the user
2969 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
2972 Description: Revoking locks from clients that went dead or catatonic might take
2974 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
2975 cancellation of such locks instant on servers without waiting for
2976 any reply from clients. Clients drops these locks when cancel
2977 notification from server is received without replying.
2980 Frequency : liblustre-only, when liblustre client dies or becomes busy
2982 Description: Doing ls on Linux clients can take a long time with active
2984 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
2985 granting such locks to it in the first place if possible. Locks
2986 are taken by proxy on the OST during the read or write and
2987 dropped immediately afterward. Add connect flags handling, do
2988 not grant locks to liblustre clients for glimpse ASTs.
2990 Severity : enhancement
2992 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
2993 Details : Existing read-ahead algorithm is tuned for the case of streamlined
2994 sequential reads and behaves badly with applications doing random
2995 reads. Improve it by reading ahead at least read region, and
2996 avoiding excessive large RPC for small reads.
2998 Severity : enhancement
3000 Description: Creating more than 1000 files for a single job may cause a load
3001 imbalance on the OSTs if there are also a large number of OSTs.
3002 Details : qos_prep_create() uses an OST index reseed value that is an
3003 even multiple of the number of available OSTs so that if the
3004 reseed happens in the middle of the object allocation it will
3005 still utilize the OSTs as uniformly as possible.
3010 Description: OST or MDS may oops in ping_evictor_main()
3011 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
3012 but doesn't restart at beginning of obd_exports_timed list
3015 Severity : enhancement
3017 Description: improve by-nid export eviction on the MDS and OST
3018 Details : allow multiple exports with the same NID to be evicted at one
3019 time without re-searching the exports list.
3022 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
3024 Description: MDS may oops in groups_free()
3025 Details : in rare race conditions a newly allocated group_info struct is
3026 freed again, and this can be NULL. The 2.4 compatibility code
3027 for groups_free() checked for a NULL pointer, but 2.6 did not.
3030 Frequency : common for liblustre clients doing little filesystem IO
3031 Bugzilla : 9352, 7313
3032 Description: server may evict liblustre clients accessing contended locks
3033 Details : if a client is granted a lock or receives a completion AST
3034 with a blocking AST already set it would not reply to the AST
3035 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
3038 Frequency : lfs setstripe, only systems with more than 160 OSTs
3040 Description: unable to set striping with a starting offset beyond OST 160
3041 Details : llapi_create_file() incorrectly limited the starting stripe
3042 index to the maximum single-file stripe count.
3045 Frequency : LDAP users only
3047 Description: lconf did not handle in-kernel recovery with LDAP properly
3048 Details : lconf/LustreDB get_refs() is searching the wrong namespace
3050 Severity : enhancement
3052 Description: bind OST threads to NUMA nodes to improve performance
3053 Details : all OST threads are uniformly bound to CPUs on a single NUMA
3054 node and do their allocations there to localize memory access
3056 Severity : enhancement
3058 Description: llmount can determine client NID directly from Myrinet (GM)
3059 Details : the client NID code from gmnalnid was moved directly into
3060 llmount, removing the need to use this or specifying the
3061 client NID explicitly when mounting GM clients with zeroconf
3064 Frequency : if client is started with down MDS
3066 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
3067 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
3068 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
3069 ptlrpc_import_delayed_req() because we didn't check if the
3070 request was interrupted, and we also didn't break out of the
3071 event loop if there was no timeout
3076 Description: data loss during non-page-aligned writes to a single file from
3077 both multiple nodes and multiple threads on one node at same time
3078 Details : updates to KMS and lsm weren't protected by common lock. Resulting
3079 inconsistency led to false short-reads, that were cached and later
3080 used by ->prepare_write() to fill in partially written page,
3081 leading to data loss.
3084 Frequency : always, if lconf --abort_recovery used
3086 Description: lconf --abort_recovery fails with 'Operation not supported'
3087 Details : lconf was attempting to abort recovery on the MDT device and not
3090 Severity : enhancement
3092 Description: remove cleanup logs
3093 Details : replace lconf-generated cleanup logs with lustre internal
3094 cleanup routines. Eliminates the need for client-cleanup and
3097 Severity : enhancement
3099 Description: add support for EAs (user and system) on lustre filesystems
3100 Details : it is now possible to store extended attributes in the Lustre
3101 client filesystem, and with the user_xattr mount option it
3102 is possible to allow users to store EAs on their files also
3104 Severity : enhancement
3106 Description: Add possibility (config option) to show minimal available OST free
3108 Details : When compiled with --enable-mindf configure option, statfs(2)
3109 (and so, df) will return least minimal free space available from
3110 all OSTs as amount of free space on FS, instead of summary of
3111 free spaces of all OSTs.
3113 Severity : enhancement
3115 Description: do not expand extent locks acquired on OST-side
3116 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
3117 by server: they are not cached anyway.
3120 Frequency : when mmap is used/binaries executed from Lustre
3122 Description: Unmmap pages before throwing them away from read cache.
3123 Details : llap_shrink cache now attempts to unmap pages before discarding
3124 them (if unmapping failed - do not discard). SLES9 kernel has
3125 extra checks that trigger if this unmapping is not done first.
3130 Description: lconf didn't resolve symlinks before checking to see whether a
3131 given mountpoint was already in use
3134 Frequency : when migrating failover services
3135 Bugzilla : 6395, 9514
3136 Description: When migrating a subset of services from a node (e.g. failback
3137 from a failover service node) the remaining services would
3138 time out and evict clients.
3139 Details : lconf --force (implied by --failover) sets the global obd_timeout
3140 to 5 seconds in order to quickly disconnect, but this caused
3141 other RPCs to time out too quickly. Do not change the global
3142 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
3144 Severity : enhancement
3145 Frequency : if MDS is started with down OST
3146 Bugzilla : 9439,5706
3147 Description: Allow startup/shutdown of an MDS without depending on the
3148 availability of the OSTs.
3149 Details : Asynchronously call mds_lov_synchronize during MDS startup.
3150 Add appropriate locking and lov-osc refcounts for safe
3151 cleaning. Add osc abort_inflight calls in case the
3152 synchronize never started.
3155 Frequency : occasional (Cray XT3 only)
3157 Description: root not authorized to access files in CRAY_PORTALS environment
3158 Details : The client process capabilities were not honoured on the MDS in
3159 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
3160 been accessed by an authorized user then root was able to access
3161 the file on the local client also. The root user capabilities
3162 are now allowed on the MDS, as this environment has secure UID.
3165 Frequency : occasional
3167 Description: ldiskfs "too long searching" message happens too often
3168 Details : A debugging message (otherwise harmless) prints too often on
3169 the OST console. This has been reduced to only happen when
3170 there are fragmentation problems on the filesystem.
3175 Description: Division by zero in statfs when all OSCs are inactive
3176 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
3177 lov_statfs divides by value returned by lov_get_stripecnt().
3181 Bugzilla : 9489, 3273
3182 Description: First write from each client to each OST was only 4kB in size,
3183 to initialize client writeback cache, which caused sub-optimal
3184 RPCs and poor layout on disk for the first writen file.
3185 Details : Clients now request an initial cache grant at (re)connect time
3186 and so that they can start streaming writes to the cache right
3187 away and always do full-sized RPCs if there is enough data.
3188 If the OST is rebooted the client also re-establishes its grant
3189 so that client cached writes will be honoured under the grant.
3194 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
3195 Details : Now I/O RPCs go to different portal number and (presumably) fast
3196 lock requests (and glimses) and other RPCs get their own service
3197 threads pool that should be able to service those RPCs
3200 Severity : enhancement
3202 Description: Ability to exchange lustre version between client and servers and
3203 issue warnings at client side if client is too old. Also for
3204 liblustre clients there is ability to refuse connection of too old
3206 Details : New 'version' field is added to connect data structure that is
3207 filled with version info. That info is later checked by server and
3211 Frequency : rare, liblustre only.
3212 Bugzilla : 9296, 9581
3213 Description: Two simultaneous writes from liblustre at offset within same page
3214 might proceed at the same time overwriting eachother with stale
3216 Details : I/O lock withing llu_file_prwv was released too early, before data
3217 actually was hitting the wire. Extended lock-holding time until
3218 server acknowledges receiving data.
3221 Frequency : extremely rare. Never observed in practice.
3223 Description: avoid generating lustre_handle cookie of 0.
3224 Details : class_handle_hash() generates handle cookies by incrementing
3225 global counter, and can hit 0 occasionaly (this is unlikely, but
3226 not impossible, because initial value of cookie counter is
3227 selected randonly). Value of 0 is used as a sentinel meaning
3228 "unassigned handle" --- avoid it. Also coalesce two critical
3229 sections in this function into one.
3231 Severity : enhancement
3233 Description: allow liblustre clients to delegate truncate locking to OST
3234 Details : To avoid overhead of locking, liblustre client instructs OST to
3235 take extent lock in ost_punch() on client's behalf. New connection
3236 flag is added to handle backward compatibility.
3238 Severity : enhancement
3239 Bugzilla : 4928, 7341, 9758
3240 Description: allow number of OST service threads to be specified
3241 Details : a module parameter allows the number of OST service threads
3242 to be specified via "options ost ost_num_threads={N}" in the
3243 OSS's /etc/modules.conf or /etc/modprobe.conf.
3247 Bugzilla : 6146, 9635, 9895
3248 Description: servers crash with bad pointer in target_handle_connect()
3249 Details : In rare cases when a client is reconnecting it was possible that
3250 the connection request was the last reference for that export.
3251 We would temporarily drop the export reference and get a new
3252 one, but this may have been the last reference and the export
3253 was just destroyed. Get new reference before dropping old one.
3255 Severity : enhancement
3256 Frequency : if client is started with failover MDS
3258 Description: Allow multiple MDS hostnames in the mount command
3259 Details : Try to read the configuration from all specified MDS
3260 hostnames during a client mount in case the "primary"
3263 Severity : enhancement
3265 Description: Stop sending data to evicted clients as soon as possible.
3266 Details : Check if the client we are about to send or are sending data to
3267 was evicted already. (Check is done every second of waiting,
3268 for which l_wait_event interface was extended to allow checking
3269 of exit condition at specified intervals).
3272 Frequency : rare, normally only when NFS exporting is done from client
3274 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
3276 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
3279 Frequency : always, if srand() is called before liblustre initialization
3281 Description: Liblustre uses system PRNG disturbing its usage by user application
3282 Details : Introduce internal to lustre fast and high-quality PRNG for
3283 lustre usage and make liblustre and some other places in generic
3284 lustre code to use it.
3286 Severity : enhancement
3287 Bugzilla : 9477, 9557, 9870
3288 Description: Verify that the MDS configuration logs are updated when xml is
3289 Details : Check if the .xml configuration logs are newer than the config
3290 logs stored on the MDS and report an error if this is the case.
3291 Request --write-conf, or allow starting with --old_conf.
3293 Severity : enhancement
3295 Description: Handle symlinks in the path when checking if Lustre is mounted.
3296 Details : Resolve intermediate symlinks when checking if a client has
3297 mounted a filesystem to avoid duplicate client mounts.
3302 Description: lconf can hit an error exception but still return success.
3303 Details : The lconf command catches the Command error exception at the top
3304 level script context and will exit with the associated exit
3305 status, but doesn't ensure that this exit status is non-zero.
3310 Description: failure of ptlrpc thread startup can cause oops
3311 Details : Starting a ptlrpc service thread can fail if there are a large
3312 number of threads or the server memory is very fragmented.
3313 Handle this without oopsing.
3316 Frequency : always, only if liblustre and non-default acceptor port was used
3318 Description: liblustre cannot connect to servers with non-default acceptor port
3319 Details : tcpnal_set_default_params() was not called and was therefore
3320 ignoring the environment varaible TCPNAL_PORT, as well as other
3321 TCPNAL_ environment variables
3326 Description: two objects could be created on the same OST for a single file
3327 Details : If an OST is down, in some cases it was possible to create two
3328 objects on a single OST for a single file. No problems other
3329 than potential performance impact and spurious error messages.
3333 Bugzilla : 5681, 9562
3334 Description: Client may oops in ll_unhash_aliases
3335 Details : Client dcache may become inconsistent in race condition.
3336 In some cases "getcwd" can fail if the current directory is
3342 Description: Inode refcounting problems in NFS export code
3343 Details : link_raw functions used to call d_instantiate without obtaining
3344 extra inode reference first.
3348 Bugzilla : 9942, 9903
3349 Description: Referencing freed requests leading to crash, memleaks with NFS.
3350 Details : We used to require that call to ll_revalidate_it was always
3351 followed by ll_lookup_it. Also with revalidate_special() it is
3352 possible to call ll_revalidate_it() twice for the same dentry
3353 even if first occurence returned success. This fix changes semantic
3354 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
3355 reference on a request referred from the intent.
3356 ll_intent_release() then releases such a request.
3359 Frequency : rare, normally benchmark loads only
3361 Description: unlinked inodes were kept in memory on the client
3362 Details : If a client is repeatedly creating and unlinking files it
3363 can accumulate a lot of stale inodes in the inode slab cache.
3364 If there is no other client load running this can cause the
3365 client node to run out of memory. Instead flush old inodes
3366 from client cache that have the same inode number as a new inode.
3369 Frequency : SLES9 2.6.5 kernel and long filenames only
3370 Bugzilla : 9969, 10379
3371 Description: utime reports stale NFS file handle
3372 Details : SLES9 uses out-of-dentry names in some cases, which confused
3373 the lustre dentry revalidation. Change it to always use the
3377 Frequency : rare, unless heavy write-truncate concurrency is continuous
3378 Bugzilla : 4180, 6984, 7171, 9963, 9331
3379 Description: OST becomes very slow and/or deadlocked during object unlink
3380 Details : filter_destroy() was holding onto the parent directory lock
3381 while truncating+unlinking objects. For very large objects this
3382 may block other threads for a long time and slow overall OST
3383 responsiveness. It may also be possible to get a lock ordering
3384 deadlock in this case, or run out of journal credits because of
3385 the combined truncate+unlink. Solution is to do object truncate
3386 first in one transaction without parent lock, and then do the
3387 final unlink in a new transaction with the parent lock. This
3388 reduces the lock hold time dramatically.
3391 Frequency : rare, 2.4 kernels only
3393 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
3394 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
3395 before unmounting their filesystem in order to avoid deadlock.
3396 The kernel_locked() function in 2.4 kernels only checks whether
3397 the kernel lock is held, not whether it is this process that is
3398 holding it as 2.6 kernels do.
3403 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
3404 Details : The client ptlrpc code may be trying to reconnect to a down
3405 server before a previous connection attempt has timed out.
3406 Increase the reconnect interval to be longer than the connection
3407 timeout interval to avoid sending duplicate connections to
3411 Frequency : echo_client brw_test command
3413 Description: fix echo_client to work with OST preallocated code
3414 Details : OST preallocation code (5137) didn't take echo_client IO path
3415 into account: echo_client calls filter methods outside of any
3416 OST thread and, hence, there is no per-thread preallocated
3417 pages and buffers to use. Solution: hijack pga pages for IO. As
3418 a byproduct, this avoids unnecessary data copying.
3422 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
3423 Description: Client can oops in mdc_commit_close() after open replay
3424 Details : It was possible for the MDS to return an open request with no
3425 transaction number in mds_finish_transno() if the client was
3426 evicted, but without actually returning an error. Clients
3427 would later try to replay that open and may trip an assertion
3428 Simplify the client close codepath, and always return an error
3429 from the MDS in case the open is not successful.
3432 Frequency : rare, 2.6 OSTs only
3434 Description: OST may deadlock under high load on fragmented files
3435 Details : If there was a heavy load and highly-fragmented OST filesystems
3436 it was possible to have all the OST threads deadlock waiting on
3437 allocation of biovecs, because the biovecs were not released
3438 until the entire RPC IO was completed. Instead, release biovecs
3439 as soon as they are complete to ensure forward IO progress.
3441 Severity : enhancement
3443 Description: Support for specifying external journal device at mount
3444 Details : If an OST or MDS device is formatted with an external journal
3445 device, this device major/minor is stored in the ext3 superblock
3446 and may not be valid for failover. Allow detecting and
3447 specifying the external journal at mount time.
3452 Description: Mounting an MDS with pending unlinked files may cause oops
3453 Details : target_finish_recovery() calls mds_postrecov() which returned
3454 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
3455 considers this an error and immediately begins cleaning up the
3456 lov, just after starting the mds_lov process
3458 Severity : enhancement
3460 Description: Implement 'lfs df' to report actual free space on per-OST basis
3461 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
3462 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
3463 usage of objects; '-h' to report in human readable format.
3465 ------------------------------------------------------------------------------
3467 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
3474 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
3475 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
3476 at startup during orphan unlinking. Instead, skip buddy update
3477 before setup, as it will be regenerated after recovery is complete.
3482 Description: If an OST is inactive, its locks might reference stale inodes.
3483 Details : lov_change_cbdata() must iterate over all namespaces, even if
3484 they are inactive to clear inode references from the lock.
3486 Severity : enhancement
3487 Frequency : occasional, if non-standard max_dirty_mb used
3489 Description: Client will block write RPCs if not enough grant
3490 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
3491 then the client will block writes while waiting for another RPC
3492 to complete instead of consuming its dirty limit. With change
3493 we get improved performance when max_dirty_mb is small.
3495 Severity : enhancement
3496 Bugzilla : 3389, 6253
3497 Description: Add support for supplementary groups on the MDS.
3498 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
3499 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
3500 lookups for user supplementary groups into a cache.
3504 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
3505 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
3506 for liblustre O_EXCL is enforced by MDS.
3510 Description: readdir never returns NULL in liblustre.
3511 Details : Corrected llu_iop_getdirentries logic, to return offset of next
3512 dentry in struct dirent.
3516 Frequency : liblustre only, depends on application IO pattern
3517 Description: liblustre clients evicted if not contacting servers
3518 Details : Don't put liblustre clients into the ping_evictor list, so
3519 they will not be evicted by the pinger ever.
3521 Severity : enhancement
3523 Description: Add ability to evict clients by NID from MDS.
3524 Details : By echoing "nid:$NID" string into
3525 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
3526 $NID would be instantly evicted from this MDS and from all active
3527 OSTs connected to it.
3531 Description: Do not query file size twice, somewhat slowing stat(2) calls.
3532 Details : lookup_it_finish() used to query file size from OSTs that was not
3537 Description: service threads change working directory to that of init
3538 Details : Starting lustre service threads may pin the working directory
3539 of the parent thread, making that filesystem busy. Threads
3540 now change to the working directory of init to avoid this.
3544 Frequency : during shutdown only
3545 Description: shutdown with a failed MDS or OST can cause unmount to hang
3546 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
3551 Frequency : 2.6 only
3552 Description: chmod/chown may include an extra supplementary group
3553 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
3554 supplementary group and if none is specified this is used.
3557 Bugzilla : 5479 (6816)
3559 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
3560 Details : If lookup is in progress on a file that is unlinked we might try
3561 to revalidate the inode and fail in revalidate after lookup is
3562 complete and ll_file_open() enqueues the open again but
3563 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
3566 Frequency : always, if lconf --abort_recovery used
3568 Description: lconf --abort_recovery fails with 'Operation not supported'
3569 Details : lconf was attempting to abort recovery on the MDT device and not
3572 ------------------------------------------------------------------------------
3574 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
3579 Frequency : rare (only unsupported configurations with a node running as an
3581 Bugzilla : 6514, 5137
3582 Description: Mounting a Lustre file system on a node running as an OST could
3584 Details : OSTs now preallocates memory needed to write out data at
3585 startup, instead of when needed, to avoid having to
3586 allocate memory in possibly low memory situations.
3587 Specifically, if the file system is mounted on on OST,
3588 memory pressure could force it to try to write out data,
3589 which it needed to allocate memory to do. Due to the low
3590 memory, it would be unable to do so and the node would
3591 become unresponsive.
3593 Severity : enhancement
3595 Description: Addition of lconf --service command line option
3596 Details : lconf now accepts a '--service <arg>' option, which is
3597 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
3599 Severity : enhancement
3601 Description: Failover mode is now the default for OSTs.
3602 Details : By default, OSTs will now run in failover mode. To return to
3603 the old behaviour, add '--failout' to the lmc line for OSTs.
3605 Severity : enhancement
3607 Description: Health checks are now provided for MDS and OSTs
3608 Details : Additional detailed health check information on MSD and OSTs
3609 is now provided through the procfs health_check value.
3612 Frequency : occasional, depends on IO load
3614 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
3615 after numerous create/delete cycles
3616 Details : The ext3 inode allocation policy would not allocate new inodes
3617 very well on the OSTs because there are no new directories
3618 being created. Instead we look for groups with free space if
3619 the parent directories are nearly full.
3624 Description: Network or server problems during mount may cause partially
3625 mounted clients instead of returning an error.
3626 Details : The config llog parsing code may overwrite the error return
3627 code during mount error handling, returning success instead
3633 Description: MDS can fail to allocate large reply buffers
3634 Details : After long uptimes the MDS can fail to allocate large reply
3635 buffers (e.g. zconf client mount config records) due to memory
3636 fragmentation or consumption by the buffer cache. Preallocate
3637 some large reply buffers so that these replies can be sent even
3638 under memory pressure.
3642 Frequency : rare (liblustre)
3643 Description: fsx running with liblustre complained that using truncate() to
3644 extend the file doesn't work. This patch corrects that issue.
3645 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
3646 ATTR_SIZE and lsm use in llu_setattr_raw.
3650 Frequency : rare, only 2.6 kernels
3651 Description: Unusual file access patterns on the MDS may result in inode
3652 data being lost in very rare circumstances.
3653 Details : Bad interaction between the ea-in-inode patch and the "no-read"
3654 code in the 2.6 kernel caused the inode and/or EA data not to
3655 be read from disk, causing single-file corruption.
3659 Frequency : rare, only 2.6 filesystems using extents
3660 Description: Heavy concurrent write and delete load may cause data corruption.
3661 Details : It was possible under high-load situations to have an extent
3662 metadata block in the block device cache from a just-unlinked
3663 file overwrite a newly-allocated data block. We now unmap any
3664 metadata buffers that alias just-allocated data blocks.
3668 Frequency : filesystems with default stripe_count larger than 77
3669 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
3670 Details : lconf specifies an inode size of 4096 bytes when the default
3671 stripe_count is larger than 77. This conflicts with the default
3672 inode density of 1 per 4096 bytes. Allocate smaller inodes in
3673 this case to avoid pinning too much memory for large EAs.
3675 ------------------------------------------------------------------------------
3677 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
3682 Frequency : rare (extremely heavy IO load with hundreds of clients)
3684 Description: Client is evicted, gets IO error writing to file
3685 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
3686 caused clients to be evicted for AST timeouts. The fixes in
3687 bug 5192 mean we no longer need to have such short AST timeouts
3688 so ldlm_timeout has been increased.
3691 Frequency : occasional during --force or --failover shutdown under load
3692 Bugzilla : 5949, 4834
3693 Description: Server oops/LBUG if stopped with --force or --failover under load
3694 Details : a collection of import/export refcount and cleanup ordering
3695 issues fixed for safer force cleanup
3698 Frequency : only filesystems larger than 120 OSTs
3699 Bugzilla : 5990, 6223
3700 Description: lfs getstripe would oops on a very large filesystem
3701 Details : lov_getconfig used kfree on vmalloc'd memory
3704 Frequency : only filesystems exporting via NFS to Solaris 10 clients
3705 Bugzilla : 6242, 6243
3706 Description: reading from files that had been truncated to a non-zero size
3707 but never opened returned no data
3708 Details : ll_file_read() reads zeros from no-object files to EOF
3713 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
3714 Details : There's little harm in aborting MDS/OSS recovery and letting it
3715 try again, so I removed the LASSERT and return an error instead.
3717 Severity : enhancement
3719 Description: New debugging infrastructure for tracking down data corruption
3720 Details : The I/O checksum code was replaced to: (a) control it at runtime,
3721 (b) cover more of the client-side code path, and (c) try to narrow
3722 down where problems occurred
3726 Bugzilla : 3819, 4364, 4397, 6313
3727 Description: Racing close and eviction MDS could cause assertion in mds_close
3728 Details : It was possible to get multiple mfd references during close and
3729 client eviction, leading to one thread referencing a freed mfd.
3731 Severity: : enhancement
3732 Bugzilla : 3262, 6359
3733 Description: Attempts to reconnect to servers are now more aggressive.
3734 Details : This builds on the enhanced upcall-less recovery that was added
3735 in 1.4.2. When trying to reconnect to servers, clients will
3736 now try each server in the failover group every 10 seconds. By
3737 default, clients would previously try one server every 25 seconds.
3742 Description: After recovery, certain operations trigger a failed
3743 assertion on a client.
3744 Details : Failing over an mds, using lconf -d --failover, while a
3745 client was doing a readdir() call would cause the client to
3746 LBUG after recovery completed and the readdir() was resent.
3748 Severity : enhancement
3750 Description: Default groups are now added by lconf
3751 Details : You can now run lconf --group <servicename> without having to
3752 manually add groups with lmc.
3755 Frequency : occasional
3757 Description: Nodes with an elan id of 0 trigger a failed assertion
3760 Frequency : always when accessing e.g. tty/console device nodes
3762 Description: tty and some other devices nodes cannot be used on lustre
3763 Details : file's private_data field is used by device data and lustre
3764 values in there got lost. New field was added to struct file to
3765 store fs-specific private data.
3768 Frequency : when exporting Lustre via NFS
3770 Description: NFSD failed occasionally when looking up a path component
3771 Details : NFSD is looking up ".." which was broken in ext3 directories
3772 that had grown large enough to become hashed.
3775 Frequency : Clusters with multiple interfaces not on the same subnet
3777 Description: Nodes will repeatedly try to reconnect to an interface which it
3778 cannot reach and report an error to the log.
3779 Details : Extra peer list entries will be created by lconf with some peers
3780 unreachable. lconf now validates the peer before adding it.
3783 Frequency : Only if a default stripe is set on the filesystem root.
3785 Description: Setting a default stripe on the filesystem root prevented the
3786 filesystem from being remounted.
3787 Details : The client was sending extra request flags in the root getattr
3788 request and did not allocate a reply buffer for the dir EA.
3791 Frequency : occasional, higher if lots of files are accessed by one client
3792 Bugzilla : 6159, 6097
3793 Description: Client trips assertion regarding lsm mismatch/magic
3794 Details : While revalidating inodes the VFS looks up inodes with ifind()
3795 and in rare cases can find an inode that is being freed.
3796 The ll_test_inode() code will free the lsm during ifind()
3797 when it finds an existing inode and then the VFS later attaches
3798 this free lsm to a new inode.
3802 Bugzilla : 6422, 7030
3803 Description: MDS deadlock between mkdir and client eviction
3804 Details : Creating a new file via mkdir or mknod (starting a transaction
3805 and getting the ns lock) can deadlock with client eviction
3806 (gets ns lock and trying to finish a synchronous transaction).
3809 Frequency : occasional
3810 Description: While starting a server, the fsfilt_ext3 module could not be
3812 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
3813 kernels. Previously, lconf would still use the ext3 name
3814 when trying to load modules. Now, it will correctly use
3815 ext3 on 2.4 and ldiskfs on 2.6.
3817 Severity : enhancement
3818 Description: The default stripe count has been changed to 1
3819 Details : The interpretation of the default stripe count (0, to lfs
3820 or lmc) has been changed to mean striping across a single
3821 OST, rather than all available. For general usage we have
3822 found a stripe count of 1 or 2 works best.
3824 Severity : enhancement
3825 Description: Add support for compiling against Cray portals.
3826 Details : Conditional compiling for some areas that are different
3830 Frequency : occasional
3831 Bugzilla : 6409, 6834
3832 Description: Creating files with an explicit stripe count may lead to
3833 a failed assertion on the MDS
3834 Details : If some OSTs are full or unavailable, creating files may
3835 trigger a failed assertion on the MDS. Now, Lustre will
3836 try to use other servers or return an error to the
3840 Frequency : occasional
3842 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
3843 caused lock timeout during readahead (since 1.4.2).
3844 Details : Processes doing readahead might match a lock that hasn't been
3845 granted yet if there are overlapping and conflicting lock
3846 requests. The readahead process waits on ungranted lock
3847 (original lock is CBPENDING), while OST waits for that process
3848 to cancel CBPENDING read lock and eventually evicts client.
3850 Severity : enhancement
3852 Description: Initial enabling of flock support for clients
3853 Details : Implements fcntl advisory locking and file status functions.
3854 This feature is provided as an optional mount flag (default
3855 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
3856 locking are implemented yet, and those that are are not guaranteed
3857 to be completely correct in production environments.
3858 mount -t lustre -o [flock|noflock] ...
3861 Frequency : occasional
3863 Description: OSTs running 2.4 kernels but with extents enabled might trip an
3864 assertion in the ext3 JBD (journaling) layer.
3865 Details : The b_committed_data struct is protected by the big kernel lock
3866 in 2.4 kernels, serializing journal_commit_transaction() and
3867 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
3868 access to this struct. In 2.6 kernels there is finer grained
3869 locking to improve SMP performance of the JBD layer.
3873 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
3878 Description: merge and cleanup kernel patches.
3881 -----------------------------------------------------------------------------
3883 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
3885 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
3886 versions of Lustre. Please update all servers and clients to
3887 version 1.4.2 or later at the same time. You must also run
3888 "lconf --write-conf {config}.xml" on the MDS while it is stopped
3889 to update the configuration logs.
3891 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
3892 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
3893 - swab llog records of type '0' so we get proper header size/idx (5861)
3894 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
3895 - fix rename of one directory over another leaking an inode (5953)
3896 - avoid SetPageDirty on 2.6 (5981)
3897 - don't re-add just-being-destroyed locks to the waiting list (5653)
3898 - when creating new directories, inherit the parent's custom
3899 striping settings if present parent (3048)
3900 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
3901 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
3902 - don't hold client locks on temporary worklist from l_lru (5666)
3903 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
3904 - automatically evict dead clients (5921)
3905 - Update file size properly in create+truncate+fstat case (6196)
3906 - Do not unhash mountpoint dentries, do not allow removal of
3908 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
3909 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
3910 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
3911 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
3912 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
3913 - Correctly handle reads of files with no objects (6243)
3914 - lctl recover will also mark a device active if deactivate used (5933)
3916 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
3917 - allow --write-conf on an MDS with different nettype than client (5619)
3918 - don't write config llogs to MDS for mounts not from that MDS (5617)
3919 - lconf should create multiple TCP connections from a client (5201)
3920 - init scripts are now turned off by default; run chkconfig --on
3921 lustre and chkconfig --on lustrefs to use them
3922 - upcalls are no longer needed for clients to recover to failover
3924 - add --abort-recovery option to lconf to abort recovery on device
3926 - add support for an arbitrary number of OSTs (3026)
3927 - Quota support protocol changes.
3928 - forward compatibility changes to wire structs (6007)
3929 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
3930 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
3931 - improved "lustre status" script
3932 - initialize blocksize for non-regular files (6062)
3933 - added --disable-server and --disable-client configure options (5782)
3934 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
3935 - Vanilla 2.4.29 support
3936 - increase maximum number of obd devices to 520 (6242)
3937 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
3938 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
3939 SLES 9 2.6.5 kernels
3940 - limit stripes per file to 160 (the maximum EA size) (6093)
3942 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
3945 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
3946 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
3947 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
3948 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
3949 - small ext3 extents cleanups and fixes (5733)
3950 - improved mballoc code, several small races and bugs fixed (5733, 5638)
3951 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
3952 - avoid needless client->OST connect, fix handle mismatch (5317)
3953 - fix DLM error path that led to out-of-sync client, long delays (5779)
3954 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
3955 - fix several locking issues related to i_size (5492,5624,5654,5672)
3956 - don't move pending lock onto export if it is already evicted (5683)
3957 - fix kernel oops when creating .foo in unlinked directory (5548)
3958 - fix deadlock in obdfilter statistics vs. object create (5811)
3959 - use time_{before,after} to avoid timer jiffies wrap (5882)
3960 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
3961 - Do not leak request if server was not able to process it (5154)
3962 - If mds_open unable to find parent dir, make that negative lookup(5154)
3963 - don't create new directories with extent-mapping (5909, 5936)
3965 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
3966 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
3967 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
3968 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
3969 - support for RHEL4 kernel on i686 (5773)
3970 - provide error messages when incompatible logs are encountered (5898)
3972 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
3973 * version 1.4.0.10 (1.4.1 release candidate 1)
3975 - don't keep a lock reference when lock is not granted (4238)
3976 - unsafe list practices (rarely) led to infinite eviction loop (4908)
3977 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
3978 - drop import inflight refcount on signal_completed_replay error (5255)
3979 - unlock page after async write error during send (3677)
3980 - handle missing objects in filter_preprw_read properly (5265)
3981 - no transno return for symlink open, don't save no-trasno open (3440)
3982 - don't try to complete elan receive that already failed (4012)
3983 - free RPC server reply state on error (5406)
3984 - clean up thread from ptlrpc_start_thread() on error (5160)
3985 - readahead could read extra page into cache that wasn't ejected (5388)
3986 - prevent races in class_attach/setup/cleanup/detach (5260)
3987 - don't dereference de->d_inode after l_dput of de (5458)
3988 - use "int" for stripe value returned from lock_to_stripe (5544)
3989 - mballoc allocation and error-checking fixes in 2.6 (5504)
3990 - block device patches to fix I/O request sizes in 2.6 (5482)
3991 - look up hostnames for IB nals (5602)
3992 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
3993 - don't start multiple acceptors for the same port (5277)
3994 - fix incorrect LASSERT in mds_getattr_name (5635)
3995 - export a proc file for general "ping" checking (5628)
3996 - fix "lfs check" to not block when the MDS is down (5628)
3998 - service request history (4965)
3999 - put {ll,lov,osc}_async_page structs in a single slab (4699)
4000 - create an "evict_client" /proc entry on OSTs, like the MDS has
4001 - fix mount usage message, return errors per mount(8) (5168)
4002 - change grep [] to grep "[]" in tests so they work in more UMLs
4003 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
4004 - remove ancient LOV_MAGIC_V0 EA support (5047)
4005 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
4006 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
4007 - allow readahead allocations to fail when low on memory (5383)
4008 - mmap locking landed again, after considerable improvement (2828)
4009 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
4011 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
4014 - send OST transaction number in read/write reply to free req (4966)
4015 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4016 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4017 - the watchdog thread now runs as interruptible (5246)
4018 - flock/lockf fixes (but it's still disabled, pending 5135)
4019 - don't use EXT3 constants in llite code (5094)
4020 - memory shortage at startup could cause assertion (5176)
4022 - reorganization of lov code
4023 - single portals codebase
4025 - add extents/mballoc support (5025)
4026 - direct I/O reads in the obdfilter (4048)
4027 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
4029 tbd Cluster File Systems, Inc. <info@clusterfs.com>
4032 - send OST transaction number in read/write reply to free req (4966)
4033 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
4034 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
4035 - the watchdog thread now runs as interruptible (5246)
4036 - handle missing objects in filter_preprw_read properly (5265)
4037 - unsafe list practices (rarely) led to infinite eviction loop (4908)
4038 - drop import inflight refcount on signal_completed_replay error (5255)
4039 - unlock page after async write error during send (3677)
4040 - return original error code on reconstructed replies (3761)
4041 - no transno return for symlink open, don't save no-trasno open (3440)
4043 - add pid to ldlm debugging output (4922)
4044 - bump the watchdog timeouts -- we can't handle 30sec yet
4045 - extra debugging for orphan dentry/inode bug (5259)
4047 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
4050 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
4051 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
4052 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
4053 - fix NULL dereference in /proc/sys/portals/routes (4827)
4054 - allow failed mdc_close() operations to be interrupted (4561)
4055 - stop precreate on OST before MDS would time out on it (4778)
4056 - don't send partial-page writes before EOF from client (4410)
4057 - discard client grant for sub-page writes on large-page clients (4520)
4058 - don't free dentries not owned by NFS code, check generation (4806)
4059 - fix lsm leak if mds_create_objects() fails (4801)
4060 - limit debug_daemon file size, always print CERROR messages (4789)
4061 - use transno after validating reply (3892)
4062 - process timed out requests if import state changes (3754)
4063 - update mtime on OST during writes, return in glimpse (4829)
4064 - add mkfsoptions to LDAP (4679)
4065 - use ->max_readahead method instead of zapping global ra (5039)
4066 - don't interrupt __l_wait_event() during strace
4068 - add software watchdogs to catch hung threads quickly (4941)
4069 - make lustrefs init script start after nfs is mounted
4070 - fix CWARN/ERROR duplication (4930)
4071 - return async write errors to application if possible (2248)
4072 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
4073 - print NAL number in %x format (4645)
4074 - update barely-supported suse-2.4.21-171 series (4842)
4075 - support for sles 9 %post scripts
4076 - support for building 2.6 kernel-source packages
4077 - support for sles km_* packages
4079 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
4082 - ignore -ENOENT errors in osc_destroy (3639)
4083 - notify osc create thread that OSC is being cleaned up (4600)
4084 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
4085 - reconstruct ost_handle() like mds_handle() (4657)
4086 - create a new thread to do import eviction to avoid deadlock (3969)
4087 - let lconf resolve symlinked-to devices (4629)
4088 - don't unlink "objects" from directory with default EA (4554)
4089 - hold socknal file ref over connect in case target is down (4394)
4090 - allow more than 32000 subdirectories in a single directory (3244)
4091 - fix blocks count for O_DIRECT writes (3751)
4092 - OST returns ENOSPC from object create when no space left (4539)
4093 - don't send truncate RPC if file size isn't changing (4410)
4094 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
4095 - bind to privileged port in socknal and tcpnal (3689)
4097 - rate limit CERROR/CWARN console message to avoid overload (4519)
4098 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
4099 - basic mmap support (3918)
4100 - kernel patch series update from b1_4 (4711)
4102 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
4105 - avoid crash during MDS cleanup with OST shut down (2775)
4106 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
4107 - don't use bad inodes on the MDS (3744)
4108 - dynamic object preallocation to improve recovery speed (4236)
4109 - don't hold spinlock over lock dumping or change debug flags (4401)
4110 - don't zero obd_dev when it is force cleaned (3651)
4111 - print grants to console if they go negative (4431)
4112 - "lctl deactivate" will stop automatic recovery attempts (3406)
4113 - look for existing locks in ldlm_handle_enqueue() (3764)
4114 - don't resolve lock handle twice in recovery avoiding race (4401)
4115 - revalidate should check working dir is a directory (4134)
4117 - don't always mark "slow" obdfilter messages as errors (4418)
4119 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
4122 - don't close LustreDB during write_conf until it is done (3860)
4123 - fix typo in lconf for_each_profile (3821)
4124 - allow dumping logs from multiple threads at one time (3820)
4125 - don't allow multiple threads in OSC recovery (3812)
4126 - fix debug_size parameters (3864)
4127 - fix mds_postrecov to initialize import for llog ctxt (3121)
4128 - replace config semaphore with spinlock (3306)
4129 - be sure to send a reply for a CANCEL rpc with bad export (3863)
4130 - don't allow enqueue to complete on a destroyed export (3822)
4131 - down write_lock before checking llog header bitmap (3825)
4132 - recover from lock replay timeout (3764)
4133 - up llog sem before sending rpc (3652)
4134 - reduce ns lock hold times when setting kms (3267)
4135 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
4136 - fix NULL deref and obd_dev leak on setup error (3312)
4137 - replace some LBUG about llog ops with error handling (3841)
4138 - don't match INVALID dentries from d_lookup and spin (3784)
4139 - hold dcache_lock while marking dentries INVALID and hashing (4255)
4140 - fix invalid assertion in ptlrpc_set_wait (3880)
4142 - add libwrap support for the TCP acceptor (3996)
4143 - add /proc/sys/portals/routes for non-root route listing (3994)
4144 - allow setting MDS UUID in .xml (2580)
4145 - print the stack of a process that LBUGs (4228)
4147 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
4150 - don't cleanup request in ll_file_open() on failed MDS open (3430)
4151 - make sure to unset replay flag from failed open requests (3440)
4152 - if default stripe count is 0, use OST count for inode size (3636)
4153 - update parent mtime/ctime on client for create/unlink (2611)
4154 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
4155 - free recovery state on server during a forced cleanup (3571)
4156 - unregister_reply for resent reqs (3063)
4157 - loop back devices mounting and status check on 2.6 (3563)
4158 - fix resource-creation race that can provoke i_size == 0 (3513)
4159 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
4160 - more debugging for page-accounting assertion (3746)
4161 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
4162 - avoid deadlock after precreation failure (3758)
4163 - fix race and lock order deadlock in orphan handling (3450, 3750)
4164 - add validity checks when grabbing inodes from l_ast_data (3599)
4166 - add /proc/.../recovery_status to obdfilter (3428)
4167 - lightweight CDEBUG infrastructure, debug daemon (3668)
4168 - change default OSC RPC parameters to be better on small clusters
4169 - turn off OST read cache for files smaller than 32MB
4170 - install man pages and include them in rpms (3100)
4171 - add new init script for (un)mounting lustre filesystems (2593)
4172 - run chkconfig in %post for init scripts (3701)
4173 - drop scimac NAL (unmaintained)
4175 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
4178 - clean kiobufs before and after use (3485)
4179 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
4180 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
4181 - change init script to fail more clearly if not run as root (1528)
4182 - allow clients to reconnect during replay (1742)
4183 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
4184 - don't do DNS lookups on NIDs too small for IP addresses (3442)
4185 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
4186 - fix cond_resched (3554)
4187 - only evict unfinished clients after recovery (3515)
4188 - allow bulk resend, prevent data loss (3570)
4189 - dynamic ptlrpc request buffer allocation (2102)
4190 - don't allow unlinking open directory if it isn't empty (2904)
4191 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
4192 - remove extraneous obd dereference causing LASSERT failure (3334)
4193 - don't use get_cycles() when creating temp. files on the mds (3156)
4194 - hold i_sem when setting i_size in ll_extent_lock() (3564)
4195 - handle EEXIST for set-stripe, set proper directory name (3336)
4197 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
4198 - fix ksocknal_fmb_callback() error messages (2918)
4200 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
4203 - don't copy lvb into (possibly NULL) reply on error (2983)
4204 - don't deref dentry after dput, don't free lvb on error (2922)
4205 - use the kms to determine writeback rpc length (2947)
4206 - increment oti_logcookies when osc is inactive (2948)
4207 - update client's i_blocks count via lvb messages (2543)
4208 - handle intent open/close of special files properly (1557)
4209 - mount MDS with errors=remount-ro, like obdfilter (2009)
4210 - initialize lock handle to avoid ASSERT on error cleanup (3057)
4211 - don't use cancelling-locks' kms values (2947)
4212 - use highest lock extent for kms, not last one (2925)
4213 - don't dereference ERR_PTR() dentry in error handling path (3107)
4214 - fix thread race in portals_debug_dumplog() (3122)
4215 - create lprocfs device entries at setup instead of at attach (1519)
4216 - common AST error handler, don't evict client on completion race (3145)
4217 - zero nameidata in detach_mnt in 2.6 (3118)
4218 - verify d_inode after revalidate_special is valid in 2.6 (3116)
4219 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
4220 - initialize RPC timeout timer earlier for 2.6 (3219)
4221 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
4222 - print nal/nid for unknown nid (3258)
4223 - additional checks for oscc recovery before doing precreate (3284)
4224 - fix ll_extent_lock() error return code for 64-bit systems (3043)
4225 - don't crash in mdc_close for bad permissions on open (3285)
4226 - zero i_rdev for non-device files (3147)
4227 - clear page->private before handing to FS, better assertion (3119)
4228 - tune the read pipeline (3236)
4229 - fix incorrect decref of invalidated dentry (2350)
4230 - provide read-ahead stats and refine rpc in flight stats (3328)
4231 - don't hold journal transaction open across create RPC (3313)
4232 - update atime on MDS at close time (3265)
4233 - close LDAP connection when recovering to avoid server load (3315)
4234 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
4235 - don't leak open file on MDS after open resend (3325)
4236 - serialize filter_precreate and filter_destroy_precreated (3329)
4237 - loop device shouldn't call sync_dev() for nul device (3092)
4238 - clear page cache after eviction (2766)
4239 - resynchronize MDS->OST in background (2824)
4240 - refuse to mount the same filesystem twice on same mountpoint (3394)
4241 - allow llmount to create routes for mounting behind routers (3320)
4242 - push lock cancellation to blocking thread for glimpse ASTs (3409)
4243 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
4244 - fix rare problem with rename on htree directories (3417)
4246 - allow default OST striping configuration per directory (1414)
4247 - fix compilation for qswnal for 2.6 kernels (3125)
4248 - increase maximum number of MDS request buffers for large systems
4249 - change liblustreapi to be useful for external progs like lfsck (3098)
4250 - increase local configuration timeout for slow disks (3353)
4251 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
4253 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
4256 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
4257 - don't overwrite extent policy data in reply if lock was blocked (2901)
4258 - drop filter export grants atomically with removal from device (2663)
4259 - del obd_self_export from work_list in class_disconnect_exports (2908)
4260 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
4261 - swab reply message in mdc_close, other PPC fixes (2464)
4262 - fix destroying of named logs (2325)
4263 - overwrite old logs when running lconf --write_conf (2264)
4264 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
4265 - fix race in target_handle_connect (2898)
4266 - mds_reint_create() should take same inode create lock (2926)
4267 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
4268 - don't close files for self_export to avoid uninitialized obd (2936)
4269 - allow MDS with the same name as client node (2939)
4270 - hold dentry reference for closed log files for unlink (2325)
4271 - reserve space for all logs during transactions (2059)
4272 - don't evict page beyond end of stripe extent (2925)
4273 - don't oops on a deleted current working directory (2399)
4274 - handle hard links to targets without a parent properly (2517)
4275 - don't dereference NULL lock when racing during eviction (2867)
4276 - don't grow lock extents when lots of conflicting locks (2919)
4278 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
4281 - account for cache space usage on clients to avoid data loss (974)
4282 - lfsck support in lustre kernel code (2349)
4283 - reduce journal credits needed for BRW writes (2370)
4284 - orphan handling to avoid losing space on client/server crashes
4285 - ptlrpcd can be blocked, stopping ALL progress (2477)
4286 - use lock value blocks to assist in proper KMS, faster stat (1021)
4287 - takes i_sem instead of DLM locks internally on obdfilter (2720)
4288 - recovery for initial connections (2355)
4289 - fixes for mds_cleanup_orphans (1934)
4290 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
4291 - block all file creations until orphan recovery completes (1901)
4292 - client remove rq_connection from request struct (2423)
4293 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
4294 - recovery timer race (2670)
4295 - mdc_close recovey bug (2532)
4296 - ptlrpc cleanup bug (2710)
4297 - mds timeout on local locks (2588)
4298 - namespace lock held during RPCs (2431)
4299 - handle interrupted sync write properly (2503)
4300 - don't try to handle a message that hasn't been replied to (2699)
4301 - client assert failure during cleanup after abort recovery (2701)
4302 - leak mdc device after failed mount (2712)
4303 - ptlrpc_check_set allows timedout requests to complete (2714)
4304 - wait for inflight reqs when ptlrpcd finishes (2710)
4305 - make sure unregistered services are removed from the srv_list
4306 - reset bulk XID's when resending them (caught by 1138 test)
4307 - unregister_bulk after timeout
4308 - fix lconf error (2694)
4309 - handle write after unfinished setstripe, stripe-only getstripe (2388)
4310 - readahead locks pages, leaves pending causing memory pressure (2673)
4311 - increase OST request buffers to 4096 on large machines (2729)
4312 - fix up permission of existing directories in simple_mkdir (2661)
4313 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
4314 - don't assign transno to errored transactions (2742)
4315 - don't delete objects on OST if given a bogus objid from MDS (2751)
4316 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
4317 - if rq_no_resend, then timeout request after recovery (2432)
4318 - fix MDS llog_logid record size, 64-bit array alignment (2733)
4319 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
4320 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
4321 - MDS recovery shouldn't delete objects that it has given out (2730)
4322 - if enqueue arrives after completion, don't clobber LVB (2819)
4323 - don't unlock pages twice when trigger_group_io returns error (2814)
4324 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
4325 - don't write pages to disk if there was an error (1450)
4326 - don't ping imports that have recovery disabled (2676)
4327 - take buffered bytes into account when balancing socknal conn (2817)
4328 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
4329 - reconnect unlink llog connection after MDS reconnects to OST (2816)
4330 - remove little-endian swabbing of llog records (1987)
4331 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
4332 - retry reposting request buffers if they fail (1191)
4333 - grow extent at grant time to avoid granting a revoked lock (2809)
4334 - lock revoke doesn't evict page if covered by a second lock (2765)
4335 - disable VM readahead to avoid reading outside lock extents (2805)
4337 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
4338 - updated kernel patches for hp-2.4.20 kernel (2681)
4340 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
4343 - fix truncated write corruption (2366)
4344 - fix for failed assertion in iopen_connect_dentry (1792,2517)
4346 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
4347 - ensure len(uuid) < 37 in lmc (1171)
4348 - fix ia64 OOPS in llog_test (2255)
4349 - zero end of page at obdfilter for partial page writes (2648)
4350 - don't leave stale dentries around after renames (bug 2428)
4351 - fix timeouts when evicting a client with a single lock held (2642)
4352 - set deadline for the initial HELLO message to drain (2634)
4353 - print out dotted-quad IP addresses in the socknal (2302)
4355 - additional debugging for MDS client eviction problem (2443)
4356 - fix mkfsoptions support for osts (2603, 2604)
4358 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
4361 - add series for the vanilla 2.6.0 kernel
4362 - add series for the vanilla 2.4.24 kernel
4363 - add series for a cray x86/64 UL kernel drop
4364 - fix xattr patches for the vanilla 2.4.19 series
4366 - generate true UUIDs in lmc (1171)
4367 - have portals stack dumping break in UML (2466)
4368 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
4369 - allocate with _NOFS in ldlm to avoid deadlock (1933)
4370 - wake callback waiting threads on client eviction (2460)
4371 - Add --ptldebug and --subsystem to lmc (1719)
4372 - update assertion to allow safe interrupt allocation
4373 - set rq_no_resend for cancel requests (2432)
4374 - recalculate ptlrpcd timeout after resend (2494)
4375 - call vfs_rmdir when removing pending directories (2368)
4376 - fix renaming a file to itself (2429)
4377 - lmc creates a default one-stripe lov (2454)
4378 - expand procfs space to handle large clusters (2326)
4379 - increase UML stack to avoid overflow
4380 - update lconf's list of debug and subsystem masks
4381 - fix lfs find --obd (2510)
4382 - /proc tunable for disabling filter read caching (2591)
4383 - stop rpm packages from altering slapd.conf (2301)
4384 - disable nagle in the socknal under 0conf (2578)
4385 - choose mds inode size based on stripe count (2572)
4386 - fix kernel-source rpm problems (2516)
4388 - add --disable-doc to avoid pdf generation (2421)
4389 - update documentation, tests, type-os, comments
4390 - avoid format warnings on ia64
4391 - remove the TOE NAL
4392 - tiny code cleanups by removing unused fields
4394 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
4397 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
4398 - remove the most blatant lies from BUILDING file (2371)
4399 - change default debug level to reasonable production setting
4400 - reduce client side cache size to reduce cache flush time
4401 - reduce max RPCs in flight to avoid unnecessary file fragmentation
4402 - make TCP zerocopy and pinger support enabled by default (2476)
4403 - sync writes completed after process exits caused crashes (2319)
4404 - maintain correct mount count on the MDS (2356)
4405 - backout 1557, because 2316 wasn't really fixed
4406 - better file I/O statistics gathering in /proc
4407 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
4408 - another kernel patch to fix zero-copy TCP function export
4409 - don't take duplicate lock when processing re-sent getattr (2420)
4410 - lctl uses obd_self_export instead of creating new conn (2353)
4411 - MDS/OST recovery case which requires object creation asserted (2425)
4412 - move lfs from /usr/sbin to /usr/bin in packages
4413 - fix race between mds_client_add and mds_client_free (2417)
4414 - use kmalloc instead of slabs in portals (2430)
4415 - don't create duplicate records when a failover MDS is present (2442)
4416 - remove unnecessary mount age check (2332)
4417 - don't remove directory inodes from locks prematurely (2451)
4418 - don't break if MDS service name is the same as hostname (2103)
4419 - fix races in client write RPC generation when cache full (2482)
4421 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
4424 - remove now-unused request->rq_obd (278)
4425 - if an allocation fails, print out how much memory we've used (1933)
4426 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
4427 - add the "configurable stack size" patch to most series files (1256)
4428 - ability to write large log records, for 100+ OST configs (2306)
4429 - fix NULL deref when filter_prep fails (2314)
4430 - fix operator precedence error in filter_sync
4431 - dynamic allocation of socknal TX descriptors (2315)
4432 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
4433 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
4434 - ensure that CWARN messages in Portals always get to the syslog
4435 - __init/__exit are not for prototype decls (ldlm_init/exit)
4436 - x86-64 compile warning fixes
4437 - fix gateway LMC keyword conflict (2318)
4438 - fix MDS lock inversions in getattr/reint paths (1844)
4439 - fix a rare lock re-ordering bug, which caused deadlock (2322)
4440 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
4441 - DLM race condition prevented some lock evictions (2328)
4442 - ENOMEM detection and retry on socknal sends (2230)
4443 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
4444 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
4445 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
4446 - better error messages when a client is rejected during recovery (1505)
4447 - avoid cancelling locks which were never granted, after failure (2330)
4448 - fix i_sem/journal inversion in mds_client_add (2333)
4449 - fix truncate/getattr lock cycle deadlock (2334)
4450 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
4451 - fix two rare exit paths which could leak an l_lock() ref (2321)
4452 - fencepost error in MDS/OST orphan recovery (2226)
4453 - make log record alignment 8 bytes (1988)
4454 - lstripe now fails when requested offset > ost_count (2237)
4455 - ensure that all kernel series have a complete list.h (1607)
4456 - fix crashes in special-file operations (2316)
4457 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
4459 - allow configurable automake binary, for testing new versions
4460 - small update to the lfs documentation
4462 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
4464 * fix negative export reference count in fsfilt_sync (2312)
4466 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
4467 * release candidate 0.9.1
4469 - orphans are moved into the PENDING directory for possible recovery
4470 - replayed opens now open by fid for orphan/rename safety (1042)
4471 - last close of an orphan inode generates a transno (683)
4472 - chdir() and mount() now pin the directory entry (1020)
4473 - avoid CERROR in normal ll_setattr_raw() error case (1500)
4474 - discard very old requests without processing them (1502)
4475 - remove some common, well-understood CERRORs (1505)
4476 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
4477 - clear "grant" flags in OST replies until OST grant code lands (1644)
4478 - fix read performance by not clobbering i_blksize on client (1598)
4479 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
4480 - make LRU size a /proc tunable, clears locks when reduced (707)
4481 - fix some lprocfs rot that prevented ptlbd from loading (1732)
4482 - server locks take references on exports now (1558)
4483 - build fixes for 2.4.20-rh trees (1663)
4484 - return an error from lov_create if all OSCs are inactive (1751)
4485 - fix import levels when a reconnect happens without a timeout (1597)
4486 - exit early from mds_open if we get a lookup error (1749)
4487 - partial page read at EOF wouldn't wait for disk before sending (1642)
4488 - avoid NULL deref in obdfilter when reading page past EOF (1592)
4489 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
4490 - fix LBUG in ll_it_open_error with rc = -2 (1861)
4491 - write/truncate lock inversion (1639)
4492 - Don't auto-load obdclass, portals modules during cleanup (1495)
4493 - fix timestamps from jumping to "now" (1763)
4494 - extra journal assertions (1648)
4495 - add an extra multiunlink test (1771)
4496 - fix read_record/write_record API (1776)
4497 - fix leak of offset_extent, possible incorrect i_size later (1772)
4498 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
4499 - Debugging for the kqswnal_get_idle_tx problems (1820)
4500 - Allow recovery to be attempted multiple times (1536)
4501 - Write out MDS last_rcvd file after it is first created (1600)
4502 - Fix tx_descriptor leak in failed transmit situations (1827)
4503 - ext3 journaling fixes for assertion failure after IO error (1871)
4504 - class_export_put() on freed export after completion AST error (1896)
4505 - Fix revalidate looping in VFS (1322)
4506 - Don't access a freed export during MDS_REINT timeout (1521)
4507 - Add open-unlink recovery support on the MDS (1673,1764)
4508 - Return an error if no MDS data was read from last_rcvd (1946)
4509 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
4510 - Don't setup a disk device that doesn't match exported UUID (317)
4511 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
4512 - avoid committing NULL handle in force close
4513 - local.sh is now a one-stripe LOV configuration
4514 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
4515 - fix ext3 htree duplicate directory entry corruption (1516)
4516 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
4517 - update to new LOV EA format (2097)
4518 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
4520 2003-06-15 Phil Schwan <phil@clusterfs.com>
4523 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
4524 - per-import/export recovery handling (958, 931, 959)
4525 - multiple last-rcvd slots, for serving multiple FSes (949)
4526 - connections are again shared between multiple imp/exports (963, 964)
4527 - "umount -f" would hang if any requests needed to be sent (393, 978)
4528 - avoid pinning large req buffer by copying for queued messages (989)
4529 - add "uuid" to "lctl device" command to help upcalls (991)
4530 - "open" RPCs with transnos would confuse recovery counters (1037)
4531 - do proper endian conversion of all wire messages (288, 340, 891)
4532 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
4533 - call waiting locks callback from LDLM recovery thread (1127, 1151)
4534 - fix ptlrpc_connection leak in target_handle_connect (1174)
4535 - fix import refcounting bug in OST and MDS cleanup (1134)
4536 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
4537 - fix very unlikely obd_types race condition (501)
4538 - remove osc_open hack for echo_client (1187)
4539 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
4540 - a failure in read_inode2 leads to deadlock (1139)
4541 - cancel ack-locks as soon as transaction is committed (1072)
4542 - fix major leaks and crashes in the bulk I/O path (937, 1057)
4543 - make sure to commitrw after any preprw to avoid deadlock (1162)
4544 - failing to execute a file in a lustre FS would lock inode (1203)
4545 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
4546 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
4547 - obd_uuid2tgt would walk off the end of the list (1255)
4548 - on IA64 the osc would give portals incorrect bulk size (1258)
4549 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
4550 - fix lock inversion caused by new llite matching code (1282)
4551 - limit the number of dirty pages on a client to 10MB (1286)
4552 - timed out locks were not being corrected cancelled (1289)
4553 - fix O_DIRECT above 4GB on IA-32 (1292)
4554 * major user-visible changes
4555 - fail out/fail over policy now controlled by the upcall (993)
4557 - add OBD_PING to check server availability and failure (954)
4558 - lustre messages are now sent in sending host order (288, 340, 891)
4559 - add eadatalen to MDS getattr reply (340)
4560 - OST read replies may contain second buffer, with per-page status (593)
4562 2003-03-11 Phil Schwan <phil@clusterfs.com>
4565 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
4566 - failed open()s could cause deadlock; fixed (867, 869)
4567 - stop cancelling OST locks when files are closed (481)
4568 - overlapping XID spaces caused network corruption (851, 853)
4569 - fix unsafe fsfilt counter arithmetic; change to atomic_t
4570 - setattr_raw added, to do single-RPC, server-side setattrs
4571 - lmc/lconf syntax change for OST UUIDs
4572 - fix crashy race condition between ptlrpc_free_req and osc_close
4573 - don't use request in mdc_enqueue if we hit a timeout (889)
4574 - don't set the inode i_size for regular files from the MDS (896)
4575 - handle out of order completion AST (842)
4576 - don't LBUG if a lock request times out after receiving AST (913)
4577 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
4578 - if a bad lock AST arrives, send an error instead of dropping entirely
4579 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
4580 - fix leak in bulk IO when only partially completed (899, 900, 926)
4581 - fix O_DIRECT for ia64 (55)
4582 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
4583 - C-z after timeout could hang a process forever; fixed (977)
4585 - client-side I/O cache (678, 924, 929, 941, 970)
4587 - READPAGE and SETATTRs which don't take server-side locks get
4590 2003-02-11 Phil Schwan <phil@clusterfs.com>
4593 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
4594 - Fix fsfilt_extN_readpage() to read a full page of directory
4595 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
4596 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
4597 - Fix imbalanced LOV object allocation and out-of-bound access (469)
4598 - Most intent operations were removed, in favour of a new RPC mode
4599 that does a single RPC to the server and bypasses most of the VFS
4600 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
4601 - Aggressively cancel local locks on DLM servers
4602 - mds_reint_unlink sends EA to the client if it's the last nlink.
4603 client uses that EA to unlink OST objects.
4604 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
4605 - recursive symlinks were fixed (439)
4606 - fixed NULL deref in DEBUG_REQ
4607 - filter_update_lastobjid no longer calls sync, which annoyed extN
4608 - fixed multi-client small-writes to a single file problem (445)
4609 - fixed mtime updates during file writes (607)
4610 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
4611 - fixed bug in obd_brw_read/write() (under guise of testing 367)
4612 - fixed Linux OST size reporting problem (444, 656)
4613 - OST now updates object mtime with writes or setattr (607, 619)
4614 - client verifies file size before zeroing page past EOF (445)
4615 - OST now writes last allocated objid to disk with allocation (108)
4616 - LOV on echo now works (409)
4618 - mds_reint_unlink sends a new buffer, with the EA included. this
4619 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
4620 set if a regular file was being unlinked, and it was the last link
4621 - use PtlGet from the target for bulk writes (315)
4622 - OST now updates object mtime with writes or setattr (607, 619)
4623 - LDLM now has a grant-time callback to revalidate locked items, if
4625 - Many MDS operations were reorganized to combat race conditions
4627 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
4628 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
4630 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4633 - Fully reactivate OST imports after reconnection (512, others)
4634 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
4635 - More graceful error handling for truncating on dead OST (515)
4636 - Don't error out unless we're actually accessing dead stripes (474)
4637 - Fix garbage sizes when stripes are missing (410)
4638 - LRU counters were broken, causing constant lock purge (433, 432)
4639 - garbage on read from stripes with failed OSTs (441)
4640 - mark OSCs as active before reconnecting during recovery (438)
4641 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
4642 - lfind did not preserve OST order in output (443)
4643 - symlinks cause hung clients, incorrect data (439)
4644 - stop dereferencing request after dropping refcount (457)
4645 - don't LASSERT(spin_is_locked) on non-SMP (455)
4646 - fixes for many rename() bugs
4647 - fstat didn't correctly synchronize attributes (399)
4648 - server must handle lock cancellation during blocking AST prep (487)
4649 - bulk descriptors were free()d too soon (511)
4650 - fix paths in lconf, which would load incorrect modules (451, 507)
4651 - fix confusing lconf 'host not found' error message (386)
4652 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
4653 - fix race condition in mdc_blocking_ast() for inode access (526)
4654 - fix lov_unpackmd() unpacking wrong number of stripes (537)
4655 - fix lov_set_osc_active() marking wrong OSC inactive (440)
4656 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
4657 - fix multiple writes of stripe MD to MDS (358, maybe 519)
4658 - fix lstripe in several ways (kernel side) (527)
4659 - fix request leak in ldlm_cli_enqueue (262)
4660 - incorrect OSC was marked inactive after OST failure
4661 - call mds_fs_cleanup before unmounting filesystem (524)
4662 - fix races between taking ns_lock and ldlm_lock_change_resource
4663 - fix races updating LOV export open file list
4664 - fix lov_enqueue error path, avoid decref-ing bad lock handle
4665 - fix recovery NULL deref in ldlm_cli_cancel_unused
4666 - fix some DLM races by using new hash table for lock handles (419)
4667 - permit the client to specify desired inodes, at replay
4668 - duplicate requests when we queue them for replay reintegration
4669 - fix last_rcvd offset calculation
4670 - sync after each recovered transaction, so we always make progress
4671 - never, not always, ERESTART requests without transnos
4672 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
4674 - skip replay if the MDS says that the client is already connected
4675 - don't check for a recovery-enabled export to match lctl's UUID
4676 - don't INC_USE_COUNT for phantom exports
4677 - don't crash when cleaning up phantom exports (567)
4678 - don't double-finish or set replay data for errored mdc_open requests
4679 - abort requests when they time out, so we don't get old replies
4680 - send/receive replies for AST messages again
4681 - if the client says that it doesn't have the lock, cancel it on the
4683 - if we timeout during I/O, don't try to cancel an in-use lock; instead
4684 mark it as destroyed, it will all work out when decref is called
4685 - fix module use counts (22, 581)
4687 - ASTs now expect a reply (server cancels lock on error reply)
4689 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
4692 - fix many simultaneous client startup (392)
4693 - fix dentry->d_it clobbering
4694 - credentials weren't being shipped for readdir/getattr operations
4695 - remove invalid assertions triggered during some concurrent MD
4697 - proper Lustre versions added (336, 389)
4698 - fix memory leak for create error case (398)
4699 - fix LOV locking bug that would get cli/srv out of sync
4700 - fix echo client over LOV (409)
4701 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
4702 - fix double-O_EXCL intent crash (424)
4703 - avoid sending multiple lock CANCELs (352)
4705 - MDS can do multi-client recovery (modulo bugs in new code)
4707 - many updates, edits, cleanups
4709 2002-11-18 Phil Schwan <phil@clusterfs.com>
4712 - fix null d_it dereference (346)
4713 - fix full OST/dbench hang (333)
4714 - fix permission problem with file removal (286)
4715 - fix removal of OSCs from LOV when they fail
4716 - fix NULL deref during bulk timeout (214)
4717 - fix problems related to multiple filesystems on one MDS (241)
4718 - fixed serious subtle metadata locking bugs
4719 - free locks on clients when inodes are removed due to memory
4721 - fix inode pointer in lock data (285)
4722 - partial support for multiple MDS on a single host (241)
4723 - data locks weren't cancelled at clear_inode time (290, 311)
4724 - intent locks could lead to unbounded lock growth (205)
4725 - added a maximum lock count, an LRU list, and a flusher
4726 - fix multiple rename (365)
4727 - properly abstracted the echo client
4728 - OSC locked 1 byte too many; fixed
4729 - rewrote brw callback code:
4730 - fixed recovery bugs related to LOVs (306)
4731 - fixed too-many-pages-in-one-write crash (191)
4732 - fixed (again) crash in sync_io_timeout (214)
4733 - probably fixed callback-related race (385)
4735 - Add capability to MDS protocol
4736 - LDLM cancellations and callbacks on different portals
4738 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
4741 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
4742 - defer open object destruction to close time (601981, 138)
4743 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
4744 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
4745 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
4746 - don't call obd_setattr in ll_file_release for destroyed objects
4747 * protocol change to lustre_msg: move |version| and add |flags|
4748 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
4749 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
4750 * added replay of create, unlink, link and rename operations during
4751 MDS failover; recovery should be much more robust now
4752 * remove failed OSCs from LOVs (only lov_create uses this so far)
4753 * the lustre-HOWTO was brought (more) up to date (582544)
4755 2002-10-23 Phil Schwan <phil@clusterfs.com>
4758 - in-use dentries weren't being reused properly (617851)
4759 - prevent multiple LDLM setup (599178)
4760 - fix LOV size calculations for truncate (617853)
4761 - fix client handling of MDS intent errors (POSIX)
4762 - fix permission bug in lovstripe.c test (624321)
4763 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
4764 - truncate past end of file could corrupt data
4765 - proper cleanup after timeouts, crashes, etc (592524, 550815)
4766 - a race in recovery could return ETIMEDOUT to apps (623947)
4767 - building outside the source directory was fixed
4768 * the lustre-HOWTO was brought (more) up to date (582544)
4769 * major progress was made on recovery functionality
4771 2002-10-10 Phil Schwan <phil@clusterfs.com>
4774 - recovery deadlock fix
4775 - rm -rf causes LBUG fix (617817)
4776 - file open by multiple tasks fix (618962)
4777 - directory permissions bugs (602707 and 620007)
4778 - journal_stop fixed with locking (611313)
4779 - O_APPEND failures resolved (618273, perhaps 614459)
4780 - lconf PATH fix (619770)
4781 - IA64 build fix (621450)
4782 - RPC buffer sizes scale with amount of memory
4784 2002-10-01 Phil Schwan <phil@clusterfs.com>
4787 - locks would be cancelled without throwing away data pages,
4788 resulting in inconsistent data (605627)
4789 - inode attributes were not always being refreshed (605627, 612449)
4790 - lconf now continues to cleanup after lctl reports an error
4791 - MDS now enforces user permissions (602707)
4792 - lprocfs cleanup fixed, but not yet enabled (614157)
4793 - fixed infinite server hang, should a client not respond to an AST
4794 - avoid going into recovery if user calls readlink() with a buffer
4795 that's too small (613941)
4796 - AST RPCs no longer require replies (614867) -- this may be changed
4797 - don't crash server if client sends an IOV that's too big (611336)
4798 - fixed lock conversion deadlock (611892)
4799 - fixed the following of symlinks (614622)
4800 * recovery: the server can remove locks from a client that dies, other
4801 clients can make progress
4802 * more extN patch fixes
4803 * compile-time configurable ptlrpc buffer allocations
4805 - collaborative read cache document
4806 - Lustre Lite Performance CDR document-in-progress
4808 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4811 - fix typo in patch-2.4.18
4813 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
4816 - clear ptlrpc request each time in handle_incoming_request()
4817 - unlink of files now destroys the object on the OST
4819 2002-09-19 Peter Braam <braam@clusterfs.com>
4821 * add hard link support
4822 * change obdfile creation method
4823 * kernel patch changed
4825 2002-09-19 Peter Braam <braam@clusterfs.com>
4828 - stack overflow bug in extN fixed
4830 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
4832 * documentation updates
4833 - add man pages for config tools
4834 - update tests/README to describe testing with new config tools
4835 - finish metadata API descriptions
4836 * bug fixes and cleanups
4837 - statfs workaround for 16TB limit
4838 - LOV stripe allocation improved, can stripe on subset of OSTs
4839 - LOV file size/IO offset was wrong for files > 4GB in size
4840 - object EA data was being dropped, caused files to be unreadable
4841 - memory overflow with non-LOV OST caused memory corruption
4842 - fixed regression tests to work with new config tools, obdfilter
4843 - fixed bug when directory size became larger than 1 block
4844 - fixed bug (for single client case) when PWD was deleted
4845 - invalidate local directory pages when doing intent-based ops
4846 - avoid LDLM oops when lock callback contained bad data
4848 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
4850 * documentation updates
4851 * bug fixes and cleanups
4852 - configuration tools
4855 - 64-bit compile warnings
4856 - 64-bit internal statfs data
4858 * test_brw on persistent OST devices
4860 * lprocfs (disabled)
4862 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
4864 * documentation updates
4865 * bug fixes and cleanups
4866 * configuration tools
4868 2002-08-30 Peter J. Braam <braam@clusterfs.com>
4871 * many small fixes to 0_5_4
4872 * io/network handling
4873 * thinkos in MDS operations
4875 2002-08-24 Peter J. Braam <braam@clusterfs.com>
4878 * crucial basic fixes to 0.5.3
4879 * IOR, Iozone work over Elan
4882 2002-08-07 Phil Schwan <phil@clusterfs.com>
4883 * version 0_5_3, our first alpha
4884 * we use the new Portals iovs
4885 * documentation updates
4886 * bug fixes and cleanups
4887 * small changes in the DLM wire protocol
4889 2002-07-25 Peter J. Braam <braam@clusterfs.com>
4890 * version 0_5_1 with some initial stability,
4891 * locking on MD and file I/O.
4892 * documentation updates
4893 * several bug fixes since 0.5.0
4894 * small changes in wire protocol
4896 2002-07-18 Phil Schwan <phil@clusterfs.com>
4898 * delivered as Lustre Light Alpha
4899 * fixed a crash after handling invalid MDS requests
4900 * fixed directory pages for architectures with non-4k pages sizes
4902 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
4903 * release version v0_4_4
4904 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
4906 * Fixes a number of interruption problems with OST operations.
4907 * Update documentation for portals header changes
4908 * Move all wire protocol structs/defines to lustre_idl.h
4909 * Fixes symlink length bug.
4910 * Add tcpdump to repository.
4912 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
4913 * release version v0_4_3
4914 * Fixes statfs for inodes on extN.
4915 * Fixes bug in runtests which would delete /etc/hosts.
4916 * Use 64-bit object IDs wherever possible (not into VFS though)
4917 Remove ost_get_info, which is unused by lustre, and out of date.
4919 2002-07-03 Peter Braam <braam@clusterfs.com>
4920 * release version v0_4_2 Fixes a lookup error (type not passed)
4921 * move forward to head of Portals
4922 * move forward to latest Lustre kernel
4924 2002-06-25 Peter Braam <braam@clusterfs.com>
4925 * release version v0_4_1. Hopefully stable on single node use.