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)
8 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
11 - fix many simultaneous client startup (392)
12 - fix dentry->d_it clobbering
13 - credentials weren't being shipped for readdir/getattr operations
14 - remove invalid assertions triggered during some concurrent MD
16 - proper Lustre versions added (336, 389)
17 - fix memory leak for create error case (398)
18 - fix LOV locking bug that would get cli/srv out of sync
19 - fix echo client over LOV (409)
20 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
21 - fix double-O_EXCL intent crash (424)
22 - avoid sending multiple lock CANCELs (352)
24 - MDS can do multi-client recovery (modulo bugs in new code)
26 - many updates, edits, cleanups
28 2002-11-18 Phil Schwan <phil@clusterfs.com>
31 - fix null d_it dereference (346)
32 - fix full OST/dbench hang (333)
33 - fix permission problem with file removal (286)
34 - fix removal of OSCs from LOV when they fail
35 - fix NULL deref during bulk timeout (214)
36 - fix problems related to multiple filesystems on one MDS (241)
37 - fixed serious subtle metadata locking bugs
38 - free locks on clients when inodes are removed due to memory
40 - fix inode pointer in lock data (285)
41 - partial support for multiple MDS on a single host (241)
42 - data locks weren't cancelled at clear_inode time (290, 311)
43 - intent locks could lead to unbounded lock growth (205)
44 - added a maximum lock count, an LRU list, and a flusher
45 - fix multiple rename (365)
46 - properly abstracted the echo client
47 - OSC locked 1 byte too many; fixed
48 - rewrote brw callback code:
49 - fixed recovery bugs related to LOVs (306)
50 - fixed too-many-pages-in-one-write crash (191)
51 - fixed (again) crash in sync_io_timeout (214)
52 - probably fixed callback-related race (385)
54 - Add capability to MDS protocol
55 - LDLM cancellations and callbacks on different portals
57 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
60 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
61 - defer open object destruction to close time (601981, 138)
62 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
63 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
64 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
65 - don't call obd_setattr in ll_file_release for destroyed objects
66 * protocol change to lustre_msg: move |version| and add |flags|
67 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
68 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
69 * added replay of create, unlink, link and rename operations during
70 MDS failover; recovery should be much more robust now
71 * remove failed OSCs from LOVs (only lov_create uses this so far)
72 * the lustre-HOWTO was brought (more) up to date (582544)
74 2002-10-23 Phil Schwan <phil@clusterfs.com>
77 - in-use dentries weren't being reused properly (617851)
78 - prevent multiple LDLM setup (599178)
79 - fix LOV size calculations for truncate (617853)
80 - fix client handling of MDS intent errors (POSIX)
81 - fix permission bug in lovstripe.c test (624321)
82 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
83 - truncate past end of file could corrupt data
84 - proper cleanup after timeouts, crashes, etc (592524, 550815)
85 - a race in recovery could return ETIMEDOUT to apps (623947)
86 - building outside the source directory was fixed
87 * the lustre-HOWTO was brought (more) up to date (582544)
88 * major progress was made on recovery functionality
90 2002-10-10 Phil Schwan <phil@clusterfs.com>
93 - recovery deadlock fix
94 - rm -rf causes LBUG fix (617817)
95 - file open by multiple tasks fix (618962)
96 - directory permissions bugs (602707 and 620007)
97 - journal_stop fixed with locking (611313)
98 - O_APPEND failures resolved (618273, perhaps 614459)
99 - lconf PATH fix (619770)
100 - IA64 build fix (621450)
101 - RPC buffer sizes scale with amount of memory
103 2002-10-01 Phil Schwan <phil@clusterfs.com>
106 - locks would be cancelled without throwing away data pages,
107 resulting in inconsistent data (605627)
108 - inode attributes were not always being refreshed (605627, 612449)
109 - lconf now continues to cleanup after lctl reports an error
110 - MDS now enforces user permissions (602707)
111 - lprocfs cleanup fixed, but not yet enabled (614157)
112 - fixed infinite server hang, should a client not respond to an AST
113 - avoid going into recovery if user calls readlink() with a buffer
114 that's too small (613941)
115 - AST RPCs no longer require replies (614867) -- this may be changed
116 - don't crash server if client sends an IOV that's too big (611336)
117 - fixed lock conversion deadlock (611892)
118 - fixed the following of symlinks (614622)
119 * recovery: the server can remove locks from a client that dies, other
120 clients can make progress
121 * more extN patch fixes
122 * compile-time configurable ptlrpc buffer allocations
124 - collaborative read cache document
125 - Lustre Lite Performance CDR document-in-progress
127 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
130 - fix typo in patch-2.4.18
132 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
135 - clear ptlrpc request each time in handle_incoming_request()
136 - unlink of files now destroys the object on the OST
138 2002-09-19 Peter Braam <braam@clusterfs.com>
140 * add hard link support
141 * change obdfile creation method
142 * kernel patch changed
144 2002-09-19 Peter Braam <braam@clusterfs.com>
147 - stack overflow bug in extN fixed
149 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
151 * documentation updates
152 - add man pages for config tools
153 - update tests/README to describe testing with new config tools
154 - finish metadata API descriptions
155 * bug fixes and cleanups
156 - statfs workaround for 16TB limit
157 - LOV stripe allocation improved, can stripe on subset of OSTs
158 - LOV file size/IO offset was wrong for files > 4GB in size
159 - object EA data was being dropped, caused files to be unreadable
160 - memory overflow with non-LOV OST caused memory corruption
161 - fixed regression tests to work with new config tools, obdfilter
162 - fixed bug when directory size became larger than 1 block
163 - fixed bug (for single client case) when PWD was deleted
164 - invalidate local directory pages when doing intent-based ops
165 - avoid LDLM oops when lock callback contained bad data
167 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
169 * documentation updates
170 * bug fixes and cleanups
171 - configuration tools
174 - 64-bit compile warnings
175 - 64-bit internal statfs data
177 * test_brw on persistent OST devices
181 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
183 * documentation updates
184 * bug fixes and cleanups
185 * configuration tools
187 2002-08-30 Peter J. Braam <braam@clusterfs.com>
190 * many small fixes to 0_5_4
191 * io/network handling
192 * thinkos in MDS operations
194 2002-08-24 Peter J. Braam <braam@clusterfs.com>
197 * crucial basic fixes to 0.5.3
198 * IOR, Iozone work over Elan
201 2002-08-07 Phil Schwan <phil@clusterfs.com>
202 * version 0_5_3, our first alpha
203 * we use the new Portals iovs
204 * documentation updates
205 * bug fixes and cleanups
206 * small changes in the DLM wire protocol
208 2002-07-25 Peter J. Braam <braam@clusterfs.com>
209 * version 0_5_1 with some initial stability,
210 * locking on MD and file I/O.
211 * documentation updates
212 * several bug fixes since 0.5.0
213 * small changes in wire protocol
215 2002-07-18 Phil Schwan <phil@clusterfs.com>
217 * delivered as Lustre Light Alpha
218 * fixed a crash after handling invalid MDS requests
219 * fixed directory pages for architectures with non-4k pages sizes
221 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
222 * release version v0_4_4
223 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
225 * Fixes a number of interruption problems with OST operations.
226 * Update documentation for portals header changes
227 * Move all wire protocol structs/defines to lustre_idl.h
228 * Fixes symlink length bug.
229 * Add tcpdump to repository.
231 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
232 * release version v0_4_3
233 * Fixes statfs for inodes on extN.
234 * Fixes bug in runtests which would delete /etc/hosts.
235 * Use 64-bit object IDs wherever possible (not into VFS though)
236 Remove ost_get_info, which is unused by lustre, and out of date.
238 2002-07-03 Peter Braam <braam@clusterfs.com>
239 * release version v0_4_2 Fixes a lookup error (type not passed)
240 * move forward to head of Portals
241 * move forward to latest Lustre kernel
243 2002-06-25 Peter Braam <braam@clusterfs.com>
244 * release version v0_4_1. Hopefully stable on single node use.