3 - LRU counters were broken, causing constant lock purge (433, 432)
4 - garbage on read from stripes with failed OSTs (441)
5 - mark OSCs as active before reconnecting during recovery (438)
6 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
7 - lfind did not preserve OST order in output (443)
8 - symlinks cause hung clients, incorrect data (439)
9 - stop dereferencing request after dropping refcount (457)
10 - don't LASSERT(spin_is_locked) on non-SMP (455)
11 - fixes for many rename() bugs
12 - fstat didn't correctly synchronize attributes (399)
13 - server must handle lock cancellation during blocking AST prep (487)
14 - bulk descriptors were free()d too soon (511)
15 - fix paths in lconf, which would load incorrect modules (451, 507)
16 - fix confusing lconf 'host not found' error message (386)
18 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
21 - fix many simultaneous client startup (392)
22 - fix dentry->d_it clobbering
23 - credentials weren't being shipped for readdir/getattr operations
24 - remove invalid assertions triggered during some concurrent MD
26 - proper Lustre versions added (336, 389)
27 - fix memory leak for create error case (398)
28 - fix LOV locking bug that would get cli/srv out of sync
29 - fix echo client over LOV (409)
30 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
31 - fix double-O_EXCL intent crash (424)
32 - avoid sending multiple lock CANCELs (352)
34 - MDS can do multi-client recovery (modulo bugs in new code)
36 - many updates, edits, cleanups
38 2002-11-18 Phil Schwan <phil@clusterfs.com>
41 - fix null d_it dereference (346)
42 - fix full OST/dbench hang (333)
43 - fix permission problem with file removal (286)
44 - fix removal of OSCs from LOV when they fail
45 - fix NULL deref during bulk timeout (214)
46 - fix problems related to multiple filesystems on one MDS (241)
47 - fixed serious subtle metadata locking bugs
48 - free locks on clients when inodes are removed due to memory
50 - fix inode pointer in lock data (285)
51 - partial support for multiple MDS on a single host (241)
52 - data locks weren't cancelled at clear_inode time (290, 311)
53 - intent locks could lead to unbounded lock growth (205)
54 - added a maximum lock count, an LRU list, and a flusher
55 - fix multiple rename (365)
56 - properly abstracted the echo client
57 - OSC locked 1 byte too many; fixed
58 - rewrote brw callback code:
59 - fixed recovery bugs related to LOVs (306)
60 - fixed too-many-pages-in-one-write crash (191)
61 - fixed (again) crash in sync_io_timeout (214)
62 - probably fixed callback-related race (385)
64 - Add capability to MDS protocol
65 - LDLM cancellations and callbacks on different portals
67 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
70 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
71 - defer open object destruction to close time (601981, 138)
72 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
73 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
74 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
75 - don't call obd_setattr in ll_file_release for destroyed objects
76 * protocol change to lustre_msg: move |version| and add |flags|
77 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
78 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
79 * added replay of create, unlink, link and rename operations during
80 MDS failover; recovery should be much more robust now
81 * remove failed OSCs from LOVs (only lov_create uses this so far)
82 * the lustre-HOWTO was brought (more) up to date (582544)
84 2002-10-23 Phil Schwan <phil@clusterfs.com>
87 - in-use dentries weren't being reused properly (617851)
88 - prevent multiple LDLM setup (599178)
89 - fix LOV size calculations for truncate (617853)
90 - fix client handling of MDS intent errors (POSIX)
91 - fix permission bug in lovstripe.c test (624321)
92 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
93 - truncate past end of file could corrupt data
94 - proper cleanup after timeouts, crashes, etc (592524, 550815)
95 - a race in recovery could return ETIMEDOUT to apps (623947)
96 - building outside the source directory was fixed
97 * the lustre-HOWTO was brought (more) up to date (582544)
98 * major progress was made on recovery functionality
100 2002-10-10 Phil Schwan <phil@clusterfs.com>
103 - recovery deadlock fix
104 - rm -rf causes LBUG fix (617817)
105 - file open by multiple tasks fix (618962)
106 - directory permissions bugs (602707 and 620007)
107 - journal_stop fixed with locking (611313)
108 - O_APPEND failures resolved (618273, perhaps 614459)
109 - lconf PATH fix (619770)
110 - IA64 build fix (621450)
111 - RPC buffer sizes scale with amount of memory
113 2002-10-01 Phil Schwan <phil@clusterfs.com>
116 - locks would be cancelled without throwing away data pages,
117 resulting in inconsistent data (605627)
118 - inode attributes were not always being refreshed (605627, 612449)
119 - lconf now continues to cleanup after lctl reports an error
120 - MDS now enforces user permissions (602707)
121 - lprocfs cleanup fixed, but not yet enabled (614157)
122 - fixed infinite server hang, should a client not respond to an AST
123 - avoid going into recovery if user calls readlink() with a buffer
124 that's too small (613941)
125 - AST RPCs no longer require replies (614867) -- this may be changed
126 - don't crash server if client sends an IOV that's too big (611336)
127 - fixed lock conversion deadlock (611892)
128 - fixed the following of symlinks (614622)
129 * recovery: the server can remove locks from a client that dies, other
130 clients can make progress
131 * more extN patch fixes
132 * compile-time configurable ptlrpc buffer allocations
134 - collaborative read cache document
135 - Lustre Lite Performance CDR document-in-progress
137 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
140 - fix typo in patch-2.4.18
142 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
145 - clear ptlrpc request each time in handle_incoming_request()
146 - unlink of files now destroys the object on the OST
148 2002-09-19 Peter Braam <braam@clusterfs.com>
150 * add hard link support
151 * change obdfile creation method
152 * kernel patch changed
154 2002-09-19 Peter Braam <braam@clusterfs.com>
157 - stack overflow bug in extN fixed
159 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
161 * documentation updates
162 - add man pages for config tools
163 - update tests/README to describe testing with new config tools
164 - finish metadata API descriptions
165 * bug fixes and cleanups
166 - statfs workaround for 16TB limit
167 - LOV stripe allocation improved, can stripe on subset of OSTs
168 - LOV file size/IO offset was wrong for files > 4GB in size
169 - object EA data was being dropped, caused files to be unreadable
170 - memory overflow with non-LOV OST caused memory corruption
171 - fixed regression tests to work with new config tools, obdfilter
172 - fixed bug when directory size became larger than 1 block
173 - fixed bug (for single client case) when PWD was deleted
174 - invalidate local directory pages when doing intent-based ops
175 - avoid LDLM oops when lock callback contained bad data
177 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
179 * documentation updates
180 * bug fixes and cleanups
181 - configuration tools
184 - 64-bit compile warnings
185 - 64-bit internal statfs data
187 * test_brw on persistent OST devices
191 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
193 * documentation updates
194 * bug fixes and cleanups
195 * configuration tools
197 2002-08-30 Peter J. Braam <braam@clusterfs.com>
200 * many small fixes to 0_5_4
201 * io/network handling
202 * thinkos in MDS operations
204 2002-08-24 Peter J. Braam <braam@clusterfs.com>
207 * crucial basic fixes to 0.5.3
208 * IOR, Iozone work over Elan
211 2002-08-07 Phil Schwan <phil@clusterfs.com>
212 * version 0_5_3, our first alpha
213 * we use the new Portals iovs
214 * documentation updates
215 * bug fixes and cleanups
216 * small changes in the DLM wire protocol
218 2002-07-25 Peter J. Braam <braam@clusterfs.com>
219 * version 0_5_1 with some initial stability,
220 * locking on MD and file I/O.
221 * documentation updates
222 * several bug fixes since 0.5.0
223 * small changes in wire protocol
225 2002-07-18 Phil Schwan <phil@clusterfs.com>
227 * delivered as Lustre Light Alpha
228 * fixed a crash after handling invalid MDS requests
229 * fixed directory pages for architectures with non-4k pages sizes
231 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
232 * release version v0_4_4
233 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
235 * Fixes a number of interruption problems with OST operations.
236 * Update documentation for portals header changes
237 * Move all wire protocol structs/defines to lustre_idl.h
238 * Fixes symlink length bug.
239 * Add tcpdump to repository.
241 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
242 * release version v0_4_3
243 * Fixes statfs for inodes on extN.
244 * Fixes bug in runtests which would delete /etc/hosts.
245 * Use 64-bit object IDs wherever possible (not into VFS though)
246 Remove ost_get_info, which is unused by lustre, and out of date.
248 2002-07-03 Peter Braam <braam@clusterfs.com>
249 * release version v0_4_2 Fixes a lookup error (type not passed)
250 * move forward to head of Portals
251 * move forward to latest Lustre kernel
253 2002-06-25 Peter Braam <braam@clusterfs.com>
254 * release version v0_4_1. Hopefully stable on single node use.