🚀 go-pugleaf

The above question was a stupid one.  Please pretend like I never posted it !!

Thanks,


Paul-

In article <1992Aug12.154841.26732@midway.uchicago.edu>
	eer1@midway.uchicago.edu (Paul Woods) writes:
>According to the man page on "hostid", the hostid can be changed.  so why when
>I try to change it, it tells me that it does not want to let me.  Is it truly
>possible to change it?  I thought that it was.

In article <1992Aug12.160802.27393@midway.uchicago.edu>
	eer1@midway.uchicago.edu (Paul Woods) writes:
>The above question was a stupid one.

No, it's a good question, but it's come up before (last November).

Here's a slightly edited version:

NeXT didn't remove sethostid() from the kernel--they merely
munged its dispatch vector in the version of sdmach they
distribute to customers.

Every UNIX system call is assigned a number;
sethostid() is #143 <syscall.h>

These numbers are used to index into the "sysent" array.
Each element is an 8-byte structure:  <sys/systm.h>

struct sysent
{
	short	sy_narg;		/* total number of arguments */
	short	sy_parallel;		/* can execute in parallel */
	int	(*sy_call)();		/* handler */
} sysent[];

This is what NeXT ships in 2.x:

sysent[143] = { 0, 1, errsys }
  errsys always fails with EINVAL

This is what it's supposed to be:

sysent[143] = { 1, 1, sethostid }
  this is the code that's supposed to be executed

This can be corrected by binary-patching 3 bytes in the kernel:

[This is for 2.1 ONLY!!!]
% cmp -l sdmach xxmach
579602   0   1
579607 240 217
579608 214 206
^^^^^^ ^^^^^^^
decimal  octal

Once you make this change, sethostid() will work the way it's
supposed to, and the hostid(1) command will work as documented.

|Note that the addresses differ in different kernel versions;
|e.g., for 2.2 it looks like this:
|
|% cmp -l sdmach xxmach
|593462   0   1
|593467 236 215
|593468 100  72

[Any of the 3.0 prerelease sites care to comment?]

					-=EPS=-