1 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
4 - Fully reactivate OST imports after reconnection (512, others)
5 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6 - More graceful error handling for truncating on dead OST (515)
7 - Don't error out unless we're actually accessing dead stripes (474)
8 - Fix garbage sizes when stripes are missing (410)
9 - LRU counters were broken, causing constant lock purge (433, 432)
10 - garbage on read from stripes with failed OSTs (441)
11 - mark OSCs as active before reconnecting during recovery (438)
12 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
13 - lfind did not preserve OST order in output (443)
14 - symlinks cause hung clients, incorrect data (439)
15 - stop dereferencing request after dropping refcount (457)
16 - don't LASSERT(spin_is_locked) on non-SMP (455)
17 - fixes for many rename() bugs
18 - fstat didn't correctly synchronize attributes (399)
19 - server must handle lock cancellation during blocking AST prep (487)
20 - bulk descriptors were free()d too soon (511)
21 - fix paths in lconf, which would load incorrect modules (451, 507)
22 - fix confusing lconf 'host not found' error message (386)
23 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
24 - fix race condition in mdc_blocking_ast() for inode access (526)
25 - fix lov_unpackmd() unpacking wrong number of stripes (537)
26 - fix lov_set_osc_active() marking wrong OSC inactive (440)
27 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
28 - fix multiple writes of stripe MD to MDS (358, maybe 519)
29 - fix lstripe in several ways (kernel side) (527)
30 - fix request leak in ldlm_cli_enqueue (262)
31 - incorrect OSC was marked inactive after OST failure
32 - call mds_fs_cleanup before unmounting filesystem (524)
33 - fix races between taking ns_lock and ldlm_lock_change_resource
34 - fix races updating LOV export open file list
35 - fix lov_enqueue error path, avoid decref-ing bad lock handle
36 - fix recovery NULL deref in ldlm_cli_cancel_unused
37 - fix some DLM races by using new hash table for lock handles (419)
38 - permit the client to specify desired inodes, at replay
39 - duplicate requests when we queue them for replay reintegration
40 - fix last_rcvd offset calculation
41 - sync after each recovered transaction, so we always make progress
42 - never, not always, ERESTART requests without transnos
43 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
45 - skip replay if the MDS says that the client is already connected
46 - don't check for a recovery-enabled export to match lctl's UUID
47 - don't INC_USE_COUNT for phantom exports
48 - don't crash when cleaning up phantom exports (567)
49 - don't double-finish or set replay data for errored mdc_open requests
50 - abort requests when they time out, so we don't get old replies
51 - send/receive replies for AST messages again
52 - if the client says that it doesn't have the lock, cancel it on the
54 - if we timeout during I/O, don't try to cancel an in-use lock; instead
55 mark it as destroyed, it will all work out when decref is called
56 - fix module use counts (22, 581)
58 - ASTs now expect a reply (server cancels lock on error reply)
60 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
63 - fix many simultaneous client startup (392)
64 - fix dentry->d_it clobbering
65 - credentials weren't being shipped for readdir/getattr operations
66 - remove invalid assertions triggered during some concurrent MD
68 - proper Lustre versions added (336, 389)
69 - fix memory leak for create error case (398)
70 - fix LOV locking bug that would get cli/srv out of sync
71 - fix echo client over LOV (409)
72 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
73 - fix double-O_EXCL intent crash (424)
74 - avoid sending multiple lock CANCELs (352)
76 - MDS can do multi-client recovery (modulo bugs in new code)
78 - many updates, edits, cleanups
80 2002-11-18 Phil Schwan <phil@clusterfs.com>
83 - fix null d_it dereference (346)
84 - fix full OST/dbench hang (333)
85 - fix permission problem with file removal (286)
86 - fix removal of OSCs from LOV when they fail
87 - fix NULL deref during bulk timeout (214)
88 - fix problems related to multiple filesystems on one MDS (241)
89 - fixed serious subtle metadata locking bugs
90 - free locks on clients when inodes are removed due to memory
92 - fix inode pointer in lock data (285)
93 - partial support for multiple MDS on a single host (241)
94 - data locks weren't cancelled at clear_inode time (290, 311)
95 - intent locks could lead to unbounded lock growth (205)
96 - added a maximum lock count, an LRU list, and a flusher
97 - fix multiple rename (365)
98 - properly abstracted the echo client
99 - OSC locked 1 byte too many; fixed
100 - rewrote brw callback code:
101 - fixed recovery bugs related to LOVs (306)
102 - fixed too-many-pages-in-one-write crash (191)
103 - fixed (again) crash in sync_io_timeout (214)
104 - probably fixed callback-related race (385)
106 - Add capability to MDS protocol
107 - LDLM cancellations and callbacks on different portals
109 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
112 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
113 - defer open object destruction to close time (601981, 138)
114 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
115 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
116 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
117 - don't call obd_setattr in ll_file_release for destroyed objects
118 * protocol change to lustre_msg: move |version| and add |flags|
119 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
120 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
121 * added replay of create, unlink, link and rename operations during
122 MDS failover; recovery should be much more robust now
123 * remove failed OSCs from LOVs (only lov_create uses this so far)
124 * the lustre-HOWTO was brought (more) up to date (582544)
126 2002-10-23 Phil Schwan <phil@clusterfs.com>
129 - in-use dentries weren't being reused properly (617851)
130 - prevent multiple LDLM setup (599178)
131 - fix LOV size calculations for truncate (617853)
132 - fix client handling of MDS intent errors (POSIX)
133 - fix permission bug in lovstripe.c test (624321)
134 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
135 - truncate past end of file could corrupt data
136 - proper cleanup after timeouts, crashes, etc (592524, 550815)
137 - a race in recovery could return ETIMEDOUT to apps (623947)
138 - building outside the source directory was fixed
139 * the lustre-HOWTO was brought (more) up to date (582544)
140 * major progress was made on recovery functionality
142 2002-10-10 Phil Schwan <phil@clusterfs.com>
145 - recovery deadlock fix
146 - rm -rf causes LBUG fix (617817)
147 - file open by multiple tasks fix (618962)
148 - directory permissions bugs (602707 and 620007)
149 - journal_stop fixed with locking (611313)
150 - O_APPEND failures resolved (618273, perhaps 614459)
151 - lconf PATH fix (619770)
152 - IA64 build fix (621450)
153 - RPC buffer sizes scale with amount of memory
155 2002-10-01 Phil Schwan <phil@clusterfs.com>
158 - locks would be cancelled without throwing away data pages,
159 resulting in inconsistent data (605627)
160 - inode attributes were not always being refreshed (605627, 612449)
161 - lconf now continues to cleanup after lctl reports an error
162 - MDS now enforces user permissions (602707)
163 - lprocfs cleanup fixed, but not yet enabled (614157)
164 - fixed infinite server hang, should a client not respond to an AST
165 - avoid going into recovery if user calls readlink() with a buffer
166 that's too small (613941)
167 - AST RPCs no longer require replies (614867) -- this may be changed
168 - don't crash server if client sends an IOV that's too big (611336)
169 - fixed lock conversion deadlock (611892)
170 - fixed the following of symlinks (614622)
171 * recovery: the server can remove locks from a client that dies, other
172 clients can make progress
173 * more extN patch fixes
174 * compile-time configurable ptlrpc buffer allocations
176 - collaborative read cache document
177 - Lustre Lite Performance CDR document-in-progress
179 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
182 - fix typo in patch-2.4.18
184 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
187 - clear ptlrpc request each time in handle_incoming_request()
188 - unlink of files now destroys the object on the OST
190 2002-09-19 Peter Braam <braam@clusterfs.com>
192 * add hard link support
193 * change obdfile creation method
194 * kernel patch changed
196 2002-09-19 Peter Braam <braam@clusterfs.com>
199 - stack overflow bug in extN fixed
201 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
203 * documentation updates
204 - add man pages for config tools
205 - update tests/README to describe testing with new config tools
206 - finish metadata API descriptions
207 * bug fixes and cleanups
208 - statfs workaround for 16TB limit
209 - LOV stripe allocation improved, can stripe on subset of OSTs
210 - LOV file size/IO offset was wrong for files > 4GB in size
211 - object EA data was being dropped, caused files to be unreadable
212 - memory overflow with non-LOV OST caused memory corruption
213 - fixed regression tests to work with new config tools, obdfilter
214 - fixed bug when directory size became larger than 1 block
215 - fixed bug (for single client case) when PWD was deleted
216 - invalidate local directory pages when doing intent-based ops
217 - avoid LDLM oops when lock callback contained bad data
219 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
221 * documentation updates
222 * bug fixes and cleanups
223 - configuration tools
226 - 64-bit compile warnings
227 - 64-bit internal statfs data
229 * test_brw on persistent OST devices
233 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
235 * documentation updates
236 * bug fixes and cleanups
237 * configuration tools
239 2002-08-30 Peter J. Braam <braam@clusterfs.com>
242 * many small fixes to 0_5_4
243 * io/network handling
244 * thinkos in MDS operations
246 2002-08-24 Peter J. Braam <braam@clusterfs.com>
249 * crucial basic fixes to 0.5.3
250 * IOR, Iozone work over Elan
253 2002-08-07 Phil Schwan <phil@clusterfs.com>
254 * version 0_5_3, our first alpha
255 * we use the new Portals iovs
256 * documentation updates
257 * bug fixes and cleanups
258 * small changes in the DLM wire protocol
260 2002-07-25 Peter J. Braam <braam@clusterfs.com>
261 * version 0_5_1 with some initial stability,
262 * locking on MD and file I/O.
263 * documentation updates
264 * several bug fixes since 0.5.0
265 * small changes in wire protocol
267 2002-07-18 Phil Schwan <phil@clusterfs.com>
269 * delivered as Lustre Light Alpha
270 * fixed a crash after handling invalid MDS requests
271 * fixed directory pages for architectures with non-4k pages sizes
273 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
274 * release version v0_4_4
275 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
277 * Fixes a number of interruption problems with OST operations.
278 * Update documentation for portals header changes
279 * Move all wire protocol structs/defines to lustre_idl.h
280 * Fixes symlink length bug.
281 * Add tcpdump to repository.
283 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
284 * release version v0_4_3
285 * Fixes statfs for inodes on extN.
286 * Fixes bug in runtests which would delete /etc/hosts.
287 * Use 64-bit object IDs wherever possible (not into VFS though)
288 Remove ost_get_info, which is unused by lustre, and out of date.
290 2002-07-03 Peter Braam <braam@clusterfs.com>
291 * release version v0_4_2 Fixes a lookup error (type not passed)
292 * move forward to head of Portals
293 * move forward to latest Lustre kernel
295 2002-06-25 Peter Braam <braam@clusterfs.com>
296 * release version v0_4_1. Hopefully stable on single node use.