|VxWorks Boot Host|
As of February 26, 2001, the boot host for VxWorks nodes was changed from b0dau30 to b0dap24. In November 2004, b0dap24 was decommissioned and we now use b0dap83 and b0dap84 as boot hosts. Also as of this time, disk access to b0dau30 using the "network" file driver, which uses rsh, no longer works. This means that if you open files with constructs like b0dau30:myfile, you will have to change the code. Network file access will only work to the boot host. nfs file access, to files in the /cdf/onln/code/cdfprod filesystem, will continue to work as before.
Note that it is strongly prefered that all configuration files necessary to run the experiment that are read by VxWorks nodes be located on the /cdf/onln/code/cdfprod disk. They should NOT be located on either the people disks or the /data1 disk.
It is known that there is a significant body of code that explicitly references node b0dau30 for rsh file I/O. This code must be changed before February 26. There are several options:
From a "C" program:
extern char sysBootHost; char filename; FILE outfile; ... sprintf(filename, "%s:/cdf/people1/cdf_cal/data/ped_0.vec", sysBootHost); outfile = fopen(filename, "rw"); ...From the VxWorks shell:
myfile = malloc(120) sprintf(&myfile, "%s:/cdf/people1/cdf_cal/data/ped_0.vec", &sysBootHost) .... free(myfile)Other methods are of course possible.
This WILL NOT WORK with b0dap83 or b0dap84 as the host. The processor will not boot. THIS IS INTENTIONAL. It is true that changing the above to the appropriate variation of
will work. This is fine for remote processors on test stands. However, absolutely no files should be loaded from home directories into the processors used for data taking. Only code in official product directories should be used. If you need a test version, you can create it in the official product area. Too many times people forget they have code in their home directories that is used in the processors and delete it. Then the processors don't boot and it can be difficult to figure out why. Furthermore, this may have deleted the best version of the code! Now is a good opportunity to fix these cases.