Debian Codegraph @

rfc143.html

raw
1 <pre>Network Working Group W. Naylor
2 Request for Comments #143 J. Wong
3 NIC #6728 C. Kline
4 Categories: D.1, D.3 J. Postel
5 Obsoletes: None UCLA - NMC
6 Updates: <a href="./rfc123">123</a>, <a href="./rfc145">145</a> 3 May 1971
7
8 <span class="h1">Regarding Proferred Official ICP</span>
9
10 We should like to comment on a race condition discovered in the ICP as
11 proposed in NWG/RFC #123. The problem arises when the server attempts
12 to initiate a second connection to the user's receive socket and the
13 first connection is not yet closed. Using a similar notation to that of
14 NWG/RFC #123 the following table illustrates the sequence of events in
15 the proferred and proposed ICP. The last two columns indicate which
16 actions must be completed before the current action may be initiated.
17 User and Server are third level programs, and UNCP and SNCP are the
18 users NCP and Servers NCP respectively. Allocates have not been
19 included since they add nothing to the argument.
20
21 Required Predecessors
22 ---------------------
23 Reference # Action Initiator "Proferred" Proposed
24 ----------- ------ --------- ----------- --------
25 1 Listen(L,32) Server -- --
26
27 2 Init(U,L,32) User -- --
28
29 3 RTS(U,L,'l') UNCP 2 2
30
31 4 STR(L,U,32) SNCP 1 and 3 1 and 3
32
33 5 Send(L,S) Server 4 4
34
35 6 SEND('l',S) SNCP 5 5
36
37 7 RECEIVE('l',S) UNCP 6 6
38
39 8 Receive(U,S) User 7 7
40
41 9 Close(L) Server 5 5
42
43 10 CLS(L,U) SNCP 9 and 7 9 and 7
44
45 11 Close(U) User 8 not used
46
47 12 CLS(U,L) UNCP 11 10
48
49
50
51
52 <span class="grey"> [Page 1]</span></pre>
53 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span>
54 <span class="grey">NWG Regarding Proferred Official ICP <a href="./rfc143">RFC 143</a></span>
55
56
57 Required Predecessors
58 ---------------------
59 Reference # Action Initiator "Proferred" Proposed
60 ----------- ------ --------- ----------- --------
61
62 13 Init(S,U+1,B ) Server 9 9
63 u
64
65 14 RTS(S,U+1,'l' ) SNCP 13 13
66 2
67
68 15 Init(S+1,U,B ) Server 13 14 and 18
69 s
70
71 16 STR(S+1,U,B ) SNCP 15 15
72 s
73
74 17 Init(U+1,S,B ) User 11 12
75 u
76
77 18 STR(U+1,S,B ) UNCP 17 17
78 u
79
80 19 Init(U,S+1,B ) User 17 17
81 s
82
83 20 RTS(U,S+1,'l' ) UNCP 19 19
84 3
85
86 Note that in the Proferred Order column, 16 can occur before 12 in which
87 case UNCP would find socket U in use and probably return a CLS (U,S+1).
88 The Server would probably then assume the User was finished with the
89 conversation.
90
91 The above problem is resolved by eliminating the Close from one side and
92 causing that side to wait for the CLS from the other side before doing
93 an Init. We propose that eliminating the user's Close (U) is the best
94 solution. (The user NCP must of course return a CLS in response to the
95 CLS sent by the server NCP).
96
97 The Server's Close (L) leads more quickly to the reuse of socket L thus
98 the serving of another user.
99
100
101
102
103
104
105
106
107
108 <span class="grey"> [Page 2]</span></pre>
109 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span>
110 <span class="grey">NWG Regarding Proferred Official ICP <a href="./rfc143">RFC 143</a></span>
111
112
113 To clarify the above discussion which may seem confusing at first
114 glance, let us demonstrate the problem in the language of RFC #123.
115
116 Server User
117 ------ ----
118 (S1) Listen(L,32) (U1) Init(U,L,32)
119
120 (S2) [Wait for match] (U2)
121
122 (S3) Send(L,S) (U3) Receive(U,S)
123
124 (S4) Close(L) (U4) Close(U)
125
126 (S5) Init(S,U+1,B ) (U5) Init(U+1,S,B )
127 u u
128
129 (S6) Init(S+1,U,B ) (U6) Init(U,S+1,B )
130 s s
131
132 Notice that since server and user are independent (probably in different
133 hosts), server could execute (S6) before user executes (U4) and could
134 receive an error back from user's NCP that socket U is busy. Similarly,
135 user could execute (U6) before server executes (S4) and could receive an
136 error back from his own NCP that socket U is not yet closed (assuming an
137 implementation where sockets are kept busy until a CLS match).
138
139 Various modifications could be made to ICP to solve this problem. We
140 propose the following ICP:
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164 <span class="grey"> [Page 3]</span></pre>
165 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span>
166 <span class="grey">NWG Regarding Proferred Official ICP <a href="./rfc143">RFC 143</a></span>
167
168
169 Server User
170 ------ ----
171 Listen(L,32) Init(U,L,32)
172
173 [Wait for match]
174
175 Send(L,S) Receive(U,S)
176
177 Close(L) [Wait for CLS]
178
179 Init(S,U+1,B ) Init(U+1,S,B )
180 u u
181
182 [Wait for match] Init(U,S+1,B )
183 s
184
185 Init(S+1,U,B )
186 s
187
188 This ICP assumes the following:
189
190 <span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>. The user can inquire or is notified of the fact that one of his</span>
191 <span class="h2"> connections has been closed.</span>
192
193 <span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>. The server can inquire or is notified that a connection for which</span>
194 <span class="h2"> he has done an Init (or Listen) is now open.</span>
195
196 Both of the above seem basic to any NCP - user interface.
197
198 This race condition problem would not exist had the dynamic reconnection
199 features of RFC #36 been included in the NCP protocol and had dynamic
200 reconnection been used in this ICP.
201
202
203 [ This RFC was put into machine readable form for entry ]
204 [ into the online RFC archives by Walter Pienciak 1/98 ]
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220 [Page 4]
221 </pre>