]> git.ipfire.org Git - thirdparty/dhcp.git/blame - RELNOTES
Fix an uninitialized variable reference
[thirdparty/dhcp.git] / RELNOTES
CommitLineData
da411127 1 Internet Software Consortium DHCP Distribution
3922772a
TL
2 Version 3, Beta 2, Patchlevel 9
3 October 15, 2000
72c7bd79
TL
4
5 Release Notes
6
74f45f96
TL
7This is a development snapshot of Version 3 of the Internet Software
8Consortium DHCP Distribution.
72c7bd79 9
da411127 10 NEW FEATURES
16449d9c 11
da411127
TL
12Version 3, Beta 2 of the ISC DHCP Distribution includes the following
13features that are new since version 2.0:
29d5553a 14
da411127
TL
15 - DHCP Failover Protocol support
16 - OMAPI, an API for accessing and modifying the DHCP server and
17 client state.
18 - Conditional behaviour
19 - Storing arbitrary information on leases
20 - Address pools with access control
21 - Client classing
22 - Address allocation restriction by class
23 - Relay agent information option support
24 - Dynamic DNS updates
25 - Many bug fixes, performance enhancements, and minor new DHCP
26 protocol features.
72c7bd79 27
f67f6c4a
TL
28This beta release is quite new, and is running in production at only a
29few sites. We strongly recommend that you exercise caution in
30installing it. The 3.0 Beta 2 lease file is not backwards compatible
31with the 3.0 Beta 1 lease file, so if you have to go back, you will
32have to convert your lease files back to the 3.0 Beta 1 format - if
33you try to run a 3.0 Beta 1 server on a 3.0 Beta 2 lease file, it will
34cheerfully delete all your leases. Having said that, of course the
35way that new releases of open source software become stable is by
36people using them, finding the bugs, and reporting the bugs. So
37while we urge you to exercise caution, we would also appreciate any
38testing or production use you can safely do.
4ff4053b 39
ca4606b5
TL
40For information on how to install, configure and run this software,
41as well as how to find documentation and report bugs, please consult
42the README file.
43
da411127
TL
44The Dynamic DNS Update support is a descendent of an implementation
45done by Lans Carstensen and Brian Dols at Rose-Hulman Institute of
46Technology, Jim Watt at Perkin-Elmer, Irina Goble at Integrated
47Measurement Systems, and Brian Murrell at BC Tel Advanced
48Communications. I'd like to express my thanks to all of these good
49people here, both for working on the code and for prodding me into
50improving it.
2aa36519 51
3922772a
TL
52 Changes since 3.0 Beta 2 Patchlevel 8
53
54- Oops, there was another subtle math error in the header-length
55 bounds-checking.
56
57 Changes since 3.0 Beta 2 Patchlevel 7
848c2547
TL
58
59- Oops, forgot to byte-swap udp header length before bounds-checking it.
60
3922772a 61 Changes since 3.0 Beta 2 Patchlevel 6
0f6045f8 62
f8572308
TL
63- Fix a possible DoS attack where a client could cause the checksummer
64 to dump core. This was a read, not a write, so it shouldn't be
65 possible to exploit it any further than that.
66
67- Implement client- and server-side support for using the Client FQDN
68 option.
69
70- Support for other option spaces in the client has been added. This
71 means that it is now possible to define a vendor option space on the
72 client, request options in that space from the server (which must
73 define the same option space), and then use those options in the
74 client. This also allows NWIP and Client FQDN options to be used
75 meaningfully.
76
77- Add object initializer support. This means that objects can now be
78 initialized to something other than all-zeros when allocated, which
79 makes, e.g., the interface object support code a little more robust.
80
81- Fix an off-by-one bug in the host stuffer. This was causing host
82 deletes not the work, and may also have been causing OMAPI
83 connections to get dropped. Thanks to James Brister for tracking
84 this one down!
85
86- Fixed a core dump in the interface discovery code that is triggered
87 when there is no subnet declaration for an interface, but the server
88 decides to continue running. Thanks to Shane Kerr for tracking
89 down and fixing this problem.
90
91 Changes since 3.0 Beta 2 Patchlevel 5
92
0f6045f8
TL
93- Fix a bug in the recent enhancement to the interface discovery code
94 to support arbitrary-length interface lists.
95
96- Support NUL-terminated DHCP options when initializing client-script
97 environment.
98
99- Fix suffix operator.
100
101- Fix NetWare/IP option parsing.
102
103- Better error/status checking in dhcpctl initialization and omapi
104 connection code.
105
106- Fix a potential memory smash in dhcpctl code.
107
108- Fix SunOS4 and (maybe) Ultrix builds.
109
110- Fix a bug where a certain sort of incoming packet could cause a core
111 dump on Solaris (and probably elsewhere).
112
113- Add some more safety checks in error logging code.
114
115- Add support for ISC_R_INCOMPLETE in OMAPI protocol connection code.
116
117- Fix relay agent so that if an interface is specified on the command
118 line, the relay agent does not dump core.
119
120- Fix class matching so that match if can be combined with match or
121 spawn with.
122
123- Do not allow spurious leases in the lease database to introduce
124 potentially bogus leases into the in-memory database.
125
126- Fix a byte-order problem in the client hardware address type code
127 for OMAPI.
128
129- Be slightly less picky about what sort of hardware addresses OMAPI
130 can install in host declarations.
131
801de092
TL
132 Changes since 3.0 Beta 2 Patchlevel 4
133
134- Incorporated Peter Marschall's proposed change to array/record
135 parsing, which allows things like the slp-agent option to be encoded
136 correctly. Thanks very much to Peter for taking the initiative to
137 do this, and for doing such a careful job of it (e.g., updating the
138 comments)!
139
140- Added an encoding for the slp-agent option. :')
141
6ed7a93d
TL
142- Fixed SunOS 4 build. Thanks to Robert Elz for responding to my
143 request for help on this with patches!
144
145- Incorporated a change that should fix a problem reported by Philippe
146 Jumelle where when the network connection between two servers is
147 lost, they never reconnect.
148
149- Fix client script files other than that for NetBSD to actually use
150 make_resolv_conf as documented in the manual page.
151
152- Fix a bug in the packet handling code that could result in a core
153 dump.
154
155- Fix a bug in the bootp code where responses on the local net would
156 be sent to the wrong MAC address. Thanks to Jerry Schave for
157 catching this one.
158
490eb5e7
TL
159 Changes since 3.0 Beta 2 Patchlevel 3
160
161- In the DHCP client, execute client statements prior to using the values
162 of options, so that the client configuration can overried, e.g., the
163 lease renewal time.
164
165- Fix a reference counting error that would result in very reproducible
166 failures in updates, as well as occasional core dumps, if a zone was
167 declared without a key.
168
169- Fix some Linux 2.0 compilation problems.
170
171- Fix a bug in scope evaluation during execution of "on" statements that
172 caused values not to be recorded on leases.
173
174- If the dhcp-max-message-size option is specified in scope, and the
175 client didn't send this option, use the one specified in scope to
176 determine the maximum size of the response.
177
592d8153
TL
178 Changes since 3.0 Beta 2 Patchlevel 2
179
359b023e
TL
180- Fix a case where spawning subclasses were being allocated
181 incorrectly, resulting in a core dump.
182
592d8153
TL
183- Fix a case where the DHCP server might inappropriately NAK a
184 RENEWING client.
185
186- Fix a place dhcprequest() where static leases could leak.
187
188- Include memory.h in omapip_p.h so that we don't get warnings about
189 using memcmp().
190
2aa36519
TL
191 Changes since 3.0 Beta 2 Patchlevel 1
192
193- Notice when SIOCFIGCONF returns more data than fit in the buffer -
194 allocate a larger buffer, and retry. Thanks to Greg Fausak for
195 pointing this out.
196
197- In the server, if no interfaces were configured, report an error and
198 exit.
199
200- Don't ever record a state of 'startup'.
201
202- Don't try to evaluate the local failover binding address if none was
203 specified. Thanks to Joseph Breu for finding this.