1xparace(3) SAORD Documentation xparace(3)
2
6 XPA Race Conditions
7
9 Potential XPA race conditions and how to avoid them.
10
12 Currently, there is only one known circumstance in which XPA can get
13 (temporarily) deadlocked in a race condition: if two or more XPA
14 servers send messages to one another using an XPA client routine such
15 as XPASet(), they can deadlock while each waits for the other server to
16 respond. (This can happen if the servers call XPAPoll() with a time
17 limit, and send messages in between the polling call.) The reason this
18 happens is that both client routines send a string to the other server
19 to establish the handshake and then wait for the server response. Since
20 each client is waiting for a response, neither is able to enter its
21 event-handling loop and respond to the other's request. This deadlock
22 will continue until one of the timeout periods expire, at which point
23 an error condition will be triggered and the timed-out server will
24 return to its event loop.
25 Starting with version 2.1.6, this rare race condition can be avoided by
27 setting the XPA_IOCALLSXPA environment variable for servers that will
28 make client calls. Setting this variable causes all XPA socket IO calls
29 to process outstanding XPA requests whenever the primary socket is not
30 ready for IO. This means that a server making a client call will
31 (recursively) process incoming server requests while waiting for client
32 completion. It also means that a server callback routine can handle
33 incoming XPA messages if it makes its own XPA call. The semi-public
34 routine oldvalue=XPAIOCallsXPA(newvalue) can be used to turn this
35 behavior off and on temporarily. Passing a 0 will turn off IO process‐
36 ing, 1 will turn it back on. The old value is returned by the call.
37 By default, the XPA_IOCALLSXPA option is turned off, because we judge
39 that the added code complication and overhead involved will not be jus‐
40 tified by the amount of its use. Moreover, processing XPA requests
41 within socket IO can lead to non-intuitive results, since incoming
42 server requests will not necessarily be processed to completion in the
43 order in which they are received.
44 Aside from setting XPA_IOCALLSXPA, the simplest way to avoid this race
46 condition is to multi-process: when you want to send a client message,
47 simply start a separate process to call the client routine, so that the
48 server is not stopped. It probably is fastest and easiest to use fork()
49 and then have the child call the client routine and exit. But you also
50 can use either the system() or popen() routine to start one of the com‐
51 mand line programs and do the same thing. Alternatively, you can use
52 XPA's internal launch() routine instead of system(). Based on fork()
53 and exec(), this routine is more secure than system() because it does
54 not call /bin/sh.
55 Starting with version 2.1.5, you also can send an XPAInfo() message
57 with the mode string "ack=false". This will cause the client to send a
58 message to the server and then exit without waiting for any return mes‐
59 sage from the server. This UDP-like behavior will avoid the server
60 deadlock when sending short XPAInfo messages.
61
63 See xpa(n) for a list of XPA help pages
64version 2.1.12 January 26, 2010 xparace(3)