syslocal(2) on 3b1 sick?

[Alex S. Crain]

In article <130 at zebra.UUCP> vern at zebra.UUCP (Vernon C. Hoxie) writes:
>
>	I'm trying to write a program to reset the clock on the 3b1.
>First, I want to be able to read the clock to see if things work. 
>This has to be done in the quickest time possible, so I am trying the
>syslocal(2) call special to the 3b1.  It has the form:
>		int syslocal(cmd, [, arg]
>From the manual (ugh) and the syslocal.h file, I get the command to be
>"SYSL_RDRTC" and the argument to be of struct rtc *x_rtc.  Now:
>		int x = syslocal(SYSL_RDRTC, x_rtc);
>gives x = -1.  So I added perror("Error id"); as the next instruction
>and got a "Bad address" response.

	syslocal wants the address of a structure, not a pointer, so the 
correct calling format would be:

	rtc x_rtc;
	int x = syslocal(SYSL_RDRTC, &x_rtc);

This is not terribly obvious, but makes sence if you think about it long 
enough, keeping in mind that C uses call-by-value argument passing.

>	How does a system function provide a bad address.  Oh! This is
>an AT&T machine (:>).

	This is a common mistake, and is more a characteristic of C and Unix
documentation than the vendor. rtx *x_rtc implies "the address of a structure"
without being specific about where that address comes from. The natural 
response is to use a pointer in that situation, and expect the system call to
return an address via the pointer. But a second glance shows that that would 
be impossible, because of call-by-value, and the correct method is to pass
the address of an existing structure.

-- 
					:alex.
					Systems Programmer
nerwin!alex at umbc3.umd.edu		UMBC
alex at umbc3.umd.edu