3 - fix many simultaneous client startup (392)
4 - fix dentry->d_it clobbering
5 - credentials weren't being shipped for readdir/getattr operations
6 - remove invalid assertions triggered during some concurrent MD
9 2002-11-18 Phil Schwan <phil@clusterfs.com>
12 - fix null d_it dereference (346)
13 - fix full OST/dbench hang (333)
14 - fix permission problem with file removal (286)
15 - fix removal of OSCs from LOV when they fail
16 - fix NULL deref during bulk timeout (214)
17 - fix problems related to multiple filesystems on one MDS (241)
18 - fixed serious subtle metadata locking bugs
19 - free locks on clients when inodes are removed due to memory
21 - fix inode pointer in lock data (285)
22 - partial support for multiple MDS on a single host (241)
23 - data locks weren't cancelled at clear_inode time (290, 311)
24 - intent locks could lead to unbounded lock growth (205)
25 - added a maximum lock count, an LRU list, and a flusher
26 - fix multiple rename (365)
27 - properly abstracted the echo client
28 - OSC locked 1 byte too many; fixed
29 - rewrote brw callback code:
30 - fixed recovery bugs related to LOVs (306)
31 - fixed too-many-pages-in-one-write crash (191)
32 - fixed (again) crash in sync_io_timeout (214)
33 - probably fixed callback-related race (385)
35 - Add capability to MDS protocol
36 - LDLM cancellations and callbacks on different portals
38 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
41 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
42 - defer open object destruction to close time (601981, 138)
43 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
44 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
45 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
46 - don't call obd_setattr in ll_file_release for destroyed objects
47 * protocol change to lustre_msg: move |version| and add |flags|
48 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
49 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
50 * added replay of create, unlink, link and rename operations during
51 MDS failover; recovery should be much more robust now
52 * remove failed OSCs from LOVs (only lov_create uses this so far)
53 * the lustre-HOWTO was brought (more) up to date (582544)
55 2002-10-23 Phil Schwan <phil@clusterfs.com>
58 - in-use dentries weren't being reused properly (617851)
59 - prevent multiple LDLM setup (599178)
60 - fix LOV size calculations for truncate (617853)
61 - fix client handling of MDS intent errors (POSIX)
62 - fix permission bug in lovstripe.c test (624321)
63 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
64 - truncate past end of file could corrupt data
65 - proper cleanup after timeouts, crashes, etc (592524, 550815)
66 - a race in recovery could return ETIMEDOUT to apps (623947)
67 - building outside the source directory was fixed
68 * the lustre-HOWTO was brought (more) up to date (582544)
69 * major progress was made on recovery functionality
71 2002-10-10 Phil Schwan <phil@clusterfs.com>
74 - recovery deadlock fix
75 - rm -rf causes LBUG fix (617817)
76 - file open by multiple tasks fix (618962)
77 - directory permissions bugs (602707 and 620007)
78 - journal_stop fixed with locking (611313)
79 - O_APPEND failures resolved (618273, perhaps 614459)
80 - lconf PATH fix (619770)
81 - IA64 build fix (621450)
82 - RPC buffer sizes scale with amount of memory
84 2002-10-01 Phil Schwan <phil@clusterfs.com>
87 - locks would be cancelled without throwing away data pages,
88 resulting in inconsistent data (605627)
89 - inode attributes were not always being refreshed (605627, 612449)
90 - lconf now continues to cleanup after lctl reports an error
91 - MDS now enforces user permissions (602707)
92 - lprocfs cleanup fixed, but not yet enabled (614157)
93 - fixed infinite server hang, should a client not respond to an AST
94 - avoid going into recovery if user calls readlink() with a buffer
95 that's too small (613941)
96 - AST RPCs no longer require replies (614867) -- this may be changed
97 - don't crash server if client sends an IOV that's too big (611336)
98 - fixed lock conversion deadlock (611892)
99 - fixed the following of symlinks (614622)
100 * recovery: the server can remove locks from a client that dies, other
101 clients can make progress
102 * more extN patch fixes
103 * compile-time configurable ptlrpc buffer allocations
105 - collaborative read cache document
106 - Lustre Lite Performance CDR document-in-progress
108 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
111 - fix typo in patch-2.4.18
113 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
116 - clear ptlrpc request each time in handle_incoming_request()
117 - unlink of files now destroys the object on the OST
119 2002-09-19 Peter Braam <braam@clusterfs.com>
121 * add hard link support
122 * change obdfile creation method
123 * kernel patch changed
125 2002-09-19 Peter Braam <braam@clusterfs.com>
128 - stack overflow bug in extN fixed
130 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
132 * documentation updates
133 - add man pages for config tools
134 - update tests/README to describe testing with new config tools
135 - finish metadata API descriptions
136 * bug fixes and cleanups
137 - statfs workaround for 16TB limit
138 - LOV stripe allocation improved, can stripe on subset of OSTs
139 - LOV file size/IO offset was wrong for files > 4GB in size
140 - object EA data was being dropped, caused files to be unreadable
141 - memory overflow with non-LOV OST caused memory corruption
142 - fixed regression tests to work with new config tools, obdfilter
143 - fixed bug when directory size became larger than 1 block
144 - fixed bug (for single client case) when PWD was deleted
145 - invalidate local directory pages when doing intent-based ops
146 - avoid LDLM oops when lock callback contained bad data
148 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
150 * documentation updates
151 * bug fixes and cleanups
152 - configuration tools
155 - 64-bit compile warnings
156 - 64-bit internal statfs data
158 * test_brw on persistent OST devices
162 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
164 * documentation updates
165 * bug fixes and cleanups
166 * configuration tools
168 2002-08-30 Peter J. Braam <braam@clusterfs.com>
171 * many small fixes to 0_5_4
172 * io/network handling
173 * thinkos in MDS operations
175 2002-08-24 Peter J. Braam <braam@clusterfs.com>
178 * crucial basic fixes to 0.5.3
179 * IOR, Iozone work over Elan
182 2002-08-07 Phil Schwan <phil@clusterfs.com>
183 * version 0_5_3, our first alpha
184 * we use the new Portals iovs
185 * documentation updates
186 * bug fixes and cleanups
187 * small changes in the DLM wire protocol
189 2002-07-25 Peter J. Braam <braam@clusterfs.com>
190 * version 0_5_1 with some initial stability,
191 * locking on MD and file I/O.
192 * documentation updates
193 * several bug fixes since 0.5.0
194 * small changes in wire protocol
196 2002-07-18 Phil Schwan <phil@clusterfs.com>
198 * delivered as Lustre Light Alpha
199 * fixed a crash after handling invalid MDS requests
200 * fixed directory pages for architectures with non-4k pages sizes
202 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
203 * release version v0_4_4
204 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
206 * Fixes a number of interruption problems with OST operations.
207 * Update documentation for portals header changes
208 * Move all wire protocol structs/defines to lustre_idl.h
209 * Fixes symlink length bug.
210 * Add tcpdump to repository.
212 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
213 * release version v0_4_3
214 * Fixes statfs for inodes on extN.
215 * Fixes bug in runtests which would delete /etc/hosts.
216 * Use 64-bit object IDs wherever possible (not into VFS though)
217 Remove ost_get_info, which is unused by lustre, and out of date.
219 2002-07-03 Peter Braam <braam@clusterfs.com>
220 * release version v0_4_2 Fixes a lookup error (type not passed)
221 * move forward to head of Portals
222 * move forward to latest Lustre kernel
224 2002-06-25 Peter Braam <braam@clusterfs.com>
225 * release version v0_4_1. Hopefully stable on single node use.