Creating Remote Processes

With PTOP, the creation of the slave process and the set up of the communications channel is done by the POPEN call in the master and the first GET and ACCEPT (or REJECT) calls in the slave. The POPEN call specifies the remote node's location, the name of the program on the remote node to be created, and various process-creation parameters. The dsnum parameter returned by POPEN identifies the slave process and its communications channel. On the slave side, the GET and ACCEPT calls complete the set up.

With NetIPC and RPM, the actions of creating the slave process and setting up the communications channel are split up. RPM handles the process creation, while NetIPC handles the communications. The calls necessary for these tasks are more complicated than the POPEN and GET/ACCEPT calls. The master process creates a call socket and registers it in the local socket registry. It passes the name of the socket and the local node name to the slave by using the RPM string feature of RPMCREATE. The master calls RPMCREATE to create the slave, including any process-creation parameters. The slave retrieves the master socket and node names from the RPM strings. It creates its own socket, looks up the master's socket, and establishes a virtual-circuit connection between the two sockets. The steps for each side are given below.

Creating Remote Processes: In the Master Program

To convert the PTOP intrinsic listed below, perform the following steps.

Syntax

dsnum := POPEN (location,progname [,itag][,entryname] [,param][,flags][,stacksize][,dlsize][,maxdata][,bufsize])

Get the local node name. You can use the NSINFO intrinsic, provided that you execute a DSLINE or a REMOTE command prior to the call. This should normally be the case for a PTOP application. You can use local node name and length item numbers (19 and 18) in NSINFO.




	NOTE: `NSINFO` is supported on NS 3000/V starting with UB-delta-1 MIT and on NS 3000/XL with Release 1.1.)

Create a TCP call socket for the master program and name the socket, using the IPCCREATE and IPCNAME intrinsics. You can use a randomly generated name from IPCNAME.
Build the opt array for the RPMCREATE call, using the INITOPT and ADDOPT intrinsics. In the opt array, include RPM strings (opt code 20000) for the socket name and the local node name. (These will be used by the slave program to set up the virtual circuit connection.) If any process-creation options are included in the POPEN call (entryname, param, flags, stacksize, dlsize, or maxdata), include them in the opt array with the corresponding RPM opt codes:
Parameter
Opt Code
entry name
22001
param
22002
flags
22003
stacksize
22004
dlsize
22005
maxdata
22006
Call RPMCREATE to create the slave process on the remote node. Use the progname and location parameters as they appear in the POPEN call, although you will have to supply the (byte) lengths of progname and location. You should set the dependent bit of the flag parameter, so the slave will terminate if the master does. Save the program descriptor, returned by RPMCREATE, for a future RPMKILL.
NOTE: RPMCREATE requires the program to be linked with Process Handling (PH) capability. PTOP does not require PH capability.
Call IPCRECVCN to wait for the connection request from the slave. Save the returned virtual-circuit descriptor for subsequent IPCSEND and IPCRECV calls.
Now that the virtual-circuit connection has been set up, call IPCSHUTDOWN to delete the master's call socket and IPCNAMERASE to delete the socket name.
If a tag is specified, call IPCSEND to send the tag on the virtual circuit to the slave.
If the slave could respond to the POPEN with either an ACCEPT or a REJECT, call IPCRECV to receive a one-byte accept or reject indication from the slave.
If a tag is specified, call IPCRECV to receive the tag from the slave (but see the comment on IPCRECV under "Exchanging Data").