1 2002-11-18 Phil Schwan <phil@clusterfs.com>
4 - fix null d_it dereference (346)
5 - fix full OST/dbench hang (333)
6 - fix permission problem with file removal (286)
7 - fix removal of OSCs from LOV when they fail
8 - fix NULL deref during bulk timeout (214)
9 - fix problems related to multiple filesystems on one MDS (241)
10 - fixed serious subtle metadata locking bugs
11 - free locks on clients when inodes are removed due to memory
13 - fix inode pointer in lock data (285)
14 - partial support for multiple MDS on a single host (241)
15 - data locks weren't cancelled at clear_inode time (290, 311)
16 - intent locks could lead to unbounded lock growth (205)
17 - added a maximum lock count, an LRU list, and a flusher
18 - fix multiple rename (365)
19 - properly abstracted the echo client
20 - OSC locked 1 byte too many; fixed
21 - rewrote brw callback code:
22 - fixed recovery bugs related to LOVs (306)
23 - fixed too-many-pages-in-one-write crash (191)
24 - fixed (again) crash in sync_io_timeout (214)
25 - probably fixed callback-related race (385)
27 - Add capability to MDS protocol
28 - LDLM cancellations and callbacks on different portals
30 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
33 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
34 - defer open object destruction to close time (601981, 138)
35 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
36 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
37 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
38 - don't call obd_setattr in ll_file_release for destroyed objects
39 * protocol change to lustre_msg: move |version| and add |flags|
40 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
41 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
42 * added replay of create, unlink, link and rename operations during
43 MDS failover; recovery should be much more robust now
44 * remove failed OSCs from LOVs (only lov_create uses this so far)
45 * the lustre-HOWTO was brought (more) up to date (582544)
47 2002-10-23 Phil Schwan <phil@clusterfs.com>
50 - in-use dentries weren't being reused properly (617851)
51 - prevent multiple LDLM setup (599178)
52 - fix LOV size calculations for truncate (617853)
53 - fix client handling of MDS intent errors (POSIX)
54 - fix permission bug in lovstripe.c test (624321)
55 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
56 - truncate past end of file could corrupt data
57 - proper cleanup after timeouts, crashes, etc (592524, 550815)
58 - a race in recovery could return ETIMEDOUT to apps (623947)
59 - building outside the source directory was fixed
60 * the lustre-HOWTO was brought (more) up to date (582544)
61 * major progress was made on recovery functionality
63 2002-10-10 Phil Schwan <phil@clusterfs.com>
66 - recovery deadlock fix
67 - rm -rf causes LBUG fix (617817)
68 - file open by multiple tasks fix (618962)
69 - directory permissions bugs (602707 and 620007)
70 - journal_stop fixed with locking (611313)
71 - O_APPEND failures resolved (618273, perhaps 614459)
72 - lconf PATH fix (619770)
73 - IA64 build fix (621450)
74 - RPC buffer sizes scale with amount of memory
76 2002-10-01 Phil Schwan <phil@clusterfs.com>
79 - locks would be cancelled without throwing away data pages,
80 resulting in inconsistent data (605627)
81 - inode attributes were not always being refreshed (605627, 612449)
82 - lconf now continues to cleanup after lctl reports an error
83 - MDS now enforces user permissions (602707)
84 - lprocfs cleanup fixed, but not yet enabled (614157)
85 - fixed infinite server hang, should a client not respond to an AST
86 - avoid going into recovery if user calls readlink() with a buffer
87 that's too small (613941)
88 - AST RPCs no longer require replies (614867) -- this may be changed
89 - don't crash server if client sends an IOV that's too big (611336)
90 - fixed lock conversion deadlock (611892)
91 - fixed the following of symlinks (614622)
92 * recovery: the server can remove locks from a client that dies, other
93 clients can make progress
94 * more extN patch fixes
95 * compile-time configurable ptlrpc buffer allocations
97 - collaborative read cache document
98 - Lustre Lite Performance CDR document-in-progress
100 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
103 - fix typo in patch-2.4.18
105 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
108 - clear ptlrpc request each time in handle_incoming_request()
109 - unlink of files now destroys the object on the OST
111 2002-09-19 Peter Braam <braam@clusterfs.com>
113 * add hard link support
114 * change obdfile creation method
115 * kernel patch changed
117 2002-09-19 Peter Braam <braam@clusterfs.com>
120 - stack overflow bug in extN fixed
122 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
124 * documentation updates
125 - add man pages for config tools
126 - update tests/README to describe testing with new config tools
127 - finish metadata API descriptions
128 * bug fixes and cleanups
129 - statfs workaround for 16TB limit
130 - LOV stripe allocation improved, can stripe on subset of OSTs
131 - LOV file size/IO offset was wrong for files > 4GB in size
132 - object EA data was being dropped, caused files to be unreadable
133 - memory overflow with non-LOV OST caused memory corruption
134 - fixed regression tests to work with new config tools, obdfilter
135 - fixed bug when directory size became larger than 1 block
136 - fixed bug (for single client case) when PWD was deleted
137 - invalidate local directory pages when doing intent-based ops
138 - avoid LDLM oops when lock callback contained bad data
140 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
142 * documentation updates
143 * bug fixes and cleanups
144 - configuration tools
147 - 64-bit compile warnings
148 - 64-bit internal statfs data
150 * test_brw on persistent OST devices
154 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
156 * documentation updates
157 * bug fixes and cleanups
158 * configuration tools
160 2002-08-30 Peter J. Braam <braam@clusterfs.com>
163 * many small fixes to 0_5_4
164 * io/network handling
165 * thinkos in MDS operations
167 2002-08-24 Peter J. Braam <braam@clusterfs.com>
170 * crucial basic fixes to 0.5.3
171 * IOR, Iozone work over Elan
174 2002-08-07 Phil Schwan <phil@clusterfs.com>
175 * version 0_5_3, our first alpha
176 * we use the new Portals iovs
177 * documentation updates
178 * bug fixes and cleanups
179 * small changes in the DLM wire protocol
181 2002-07-25 Peter J. Braam <braam@clusterfs.com>
182 * version 0_5_1 with some initial stability,
183 * locking on MD and file I/O.
184 * documentation updates
185 * several bug fixes since 0.5.0
186 * small changes in wire protocol
188 2002-07-18 Phil Schwan <phil@clusterfs.com>
190 * delivered as Lustre Light Alpha
191 * fixed a crash after handling invalid MDS requests
192 * fixed directory pages for architectures with non-4k pages sizes
194 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
195 * release version v0_4_4
196 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
198 * Fixes a number of interruption problems with OST operations.
199 * Update documentation for portals header changes
200 * Move all wire protocol structs/defines to lustre_idl.h
201 * Fixes symlink length bug.
202 * Add tcpdump to repository.
204 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
205 * release version v0_4_3
206 * Fixes statfs for inodes on extN.
207 * Fixes bug in runtests which would delete /etc/hosts.
208 * Use 64-bit object IDs wherever possible (not into VFS though)
209 Remove ost_get_info, which is unused by lustre, and out of date.
211 2002-07-03 Peter Braam <braam@clusterfs.com>
212 * release version v0_4_2 Fixes a lookup error (type not passed)
213 * move forward to head of Portals
214 * move forward to latest Lustre kernel
216 2002-06-25 Peter Braam <braam@clusterfs.com>
217 * release version v0_4_1. Hopefully stable on single node use.