3 - Fully reactivate OST imports after reconnection (512, others)
4 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5 - More graceful error handling for truncating on dead OST (515)
6 - Don't error out unless we're actually accessing dead stripes (474)
7 - Fix garbage sizes when stripes are missing (410)
8 - LRU counters were broken, causing constant lock purge (433, 432)
9 - garbage on read from stripes with failed OSTs (441)
10 - mark OSCs as active before reconnecting during recovery (438)
11 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
12 - lfind did not preserve OST order in output (443)
13 - symlinks cause hung clients, incorrect data (439)
14 - stop dereferencing request after dropping refcount (457)
15 - don't LASSERT(spin_is_locked) on non-SMP (455)
16 - fixes for many rename() bugs
17 - fstat didn't correctly synchronize attributes (399)
18 - server must handle lock cancellation during blocking AST prep (487)
19 - bulk descriptors were free()d too soon (511)
20 - fix paths in lconf, which would load incorrect modules (451, 507)
21 - fix confusing lconf 'host not found' error message (386)
23 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
26 - fix many simultaneous client startup (392)
27 - fix dentry->d_it clobbering
28 - credentials weren't being shipped for readdir/getattr operations
29 - remove invalid assertions triggered during some concurrent MD
31 - proper Lustre versions added (336, 389)
32 - fix memory leak for create error case (398)
33 - fix LOV locking bug that would get cli/srv out of sync
34 - fix echo client over LOV (409)
35 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
36 - fix double-O_EXCL intent crash (424)
37 - avoid sending multiple lock CANCELs (352)
39 - MDS can do multi-client recovery (modulo bugs in new code)
41 - many updates, edits, cleanups
43 2002-11-18 Phil Schwan <phil@clusterfs.com>
46 - fix null d_it dereference (346)
47 - fix full OST/dbench hang (333)
48 - fix permission problem with file removal (286)
49 - fix removal of OSCs from LOV when they fail
50 - fix NULL deref during bulk timeout (214)
51 - fix problems related to multiple filesystems on one MDS (241)
52 - fixed serious subtle metadata locking bugs
53 - free locks on clients when inodes are removed due to memory
55 - fix inode pointer in lock data (285)
56 - partial support for multiple MDS on a single host (241)
57 - data locks weren't cancelled at clear_inode time (290, 311)
58 - intent locks could lead to unbounded lock growth (205)
59 - added a maximum lock count, an LRU list, and a flusher
60 - fix multiple rename (365)
61 - properly abstracted the echo client
62 - OSC locked 1 byte too many; fixed
63 - rewrote brw callback code:
64 - fixed recovery bugs related to LOVs (306)
65 - fixed too-many-pages-in-one-write crash (191)
66 - fixed (again) crash in sync_io_timeout (214)
67 - probably fixed callback-related race (385)
69 - Add capability to MDS protocol
70 - LDLM cancellations and callbacks on different portals
72 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
75 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
76 - defer open object destruction to close time (601981, 138)
77 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
78 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
79 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
80 - don't call obd_setattr in ll_file_release for destroyed objects
81 * protocol change to lustre_msg: move |version| and add |flags|
82 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
83 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
84 * added replay of create, unlink, link and rename operations during
85 MDS failover; recovery should be much more robust now
86 * remove failed OSCs from LOVs (only lov_create uses this so far)
87 * the lustre-HOWTO was brought (more) up to date (582544)
89 2002-10-23 Phil Schwan <phil@clusterfs.com>
92 - in-use dentries weren't being reused properly (617851)
93 - prevent multiple LDLM setup (599178)
94 - fix LOV size calculations for truncate (617853)
95 - fix client handling of MDS intent errors (POSIX)
96 - fix permission bug in lovstripe.c test (624321)
97 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
98 - truncate past end of file could corrupt data
99 - proper cleanup after timeouts, crashes, etc (592524, 550815)
100 - a race in recovery could return ETIMEDOUT to apps (623947)
101 - building outside the source directory was fixed
102 * the lustre-HOWTO was brought (more) up to date (582544)
103 * major progress was made on recovery functionality
105 2002-10-10 Phil Schwan <phil@clusterfs.com>
108 - recovery deadlock fix
109 - rm -rf causes LBUG fix (617817)
110 - file open by multiple tasks fix (618962)
111 - directory permissions bugs (602707 and 620007)
112 - journal_stop fixed with locking (611313)
113 - O_APPEND failures resolved (618273, perhaps 614459)
114 - lconf PATH fix (619770)
115 - IA64 build fix (621450)
116 - RPC buffer sizes scale with amount of memory
118 2002-10-01 Phil Schwan <phil@clusterfs.com>
121 - locks would be cancelled without throwing away data pages,
122 resulting in inconsistent data (605627)
123 - inode attributes were not always being refreshed (605627, 612449)
124 - lconf now continues to cleanup after lctl reports an error
125 - MDS now enforces user permissions (602707)
126 - lprocfs cleanup fixed, but not yet enabled (614157)
127 - fixed infinite server hang, should a client not respond to an AST
128 - avoid going into recovery if user calls readlink() with a buffer
129 that's too small (613941)
130 - AST RPCs no longer require replies (614867) -- this may be changed
131 - don't crash server if client sends an IOV that's too big (611336)
132 - fixed lock conversion deadlock (611892)
133 - fixed the following of symlinks (614622)
134 * recovery: the server can remove locks from a client that dies, other
135 clients can make progress
136 * more extN patch fixes
137 * compile-time configurable ptlrpc buffer allocations
139 - collaborative read cache document
140 - Lustre Lite Performance CDR document-in-progress
142 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
145 - fix typo in patch-2.4.18
147 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
150 - clear ptlrpc request each time in handle_incoming_request()
151 - unlink of files now destroys the object on the OST
153 2002-09-19 Peter Braam <braam@clusterfs.com>
155 * add hard link support
156 * change obdfile creation method
157 * kernel patch changed
159 2002-09-19 Peter Braam <braam@clusterfs.com>
162 - stack overflow bug in extN fixed
164 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
166 * documentation updates
167 - add man pages for config tools
168 - update tests/README to describe testing with new config tools
169 - finish metadata API descriptions
170 * bug fixes and cleanups
171 - statfs workaround for 16TB limit
172 - LOV stripe allocation improved, can stripe on subset of OSTs
173 - LOV file size/IO offset was wrong for files > 4GB in size
174 - object EA data was being dropped, caused files to be unreadable
175 - memory overflow with non-LOV OST caused memory corruption
176 - fixed regression tests to work with new config tools, obdfilter
177 - fixed bug when directory size became larger than 1 block
178 - fixed bug (for single client case) when PWD was deleted
179 - invalidate local directory pages when doing intent-based ops
180 - avoid LDLM oops when lock callback contained bad data
182 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
184 * documentation updates
185 * bug fixes and cleanups
186 - configuration tools
189 - 64-bit compile warnings
190 - 64-bit internal statfs data
192 * test_brw on persistent OST devices
196 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
198 * documentation updates
199 * bug fixes and cleanups
200 * configuration tools
202 2002-08-30 Peter J. Braam <braam@clusterfs.com>
205 * many small fixes to 0_5_4
206 * io/network handling
207 * thinkos in MDS operations
209 2002-08-24 Peter J. Braam <braam@clusterfs.com>
212 * crucial basic fixes to 0.5.3
213 * IOR, Iozone work over Elan
216 2002-08-07 Phil Schwan <phil@clusterfs.com>
217 * version 0_5_3, our first alpha
218 * we use the new Portals iovs
219 * documentation updates
220 * bug fixes and cleanups
221 * small changes in the DLM wire protocol
223 2002-07-25 Peter J. Braam <braam@clusterfs.com>
224 * version 0_5_1 with some initial stability,
225 * locking on MD and file I/O.
226 * documentation updates
227 * several bug fixes since 0.5.0
228 * small changes in wire protocol
230 2002-07-18 Phil Schwan <phil@clusterfs.com>
232 * delivered as Lustre Light Alpha
233 * fixed a crash after handling invalid MDS requests
234 * fixed directory pages for architectures with non-4k pages sizes
236 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
237 * release version v0_4_4
238 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
240 * Fixes a number of interruption problems with OST operations.
241 * Update documentation for portals header changes
242 * Move all wire protocol structs/defines to lustre_idl.h
243 * Fixes symlink length bug.
244 * Add tcpdump to repository.
246 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
247 * release version v0_4_3
248 * Fixes statfs for inodes on extN.
249 * Fixes bug in runtests which would delete /etc/hosts.
250 * Use 64-bit object IDs wherever possible (not into VFS though)
251 Remove ost_get_info, which is unused by lustre, and out of date.
253 2002-07-03 Peter Braam <braam@clusterfs.com>
254 * release version v0_4_2 Fixes a lookup error (type not passed)
255 * move forward to head of Portals
256 * move forward to latest Lustre kernel
258 2002-06-25 Peter Braam <braam@clusterfs.com>
259 * release version v0_4_1. Hopefully stable on single node use.