update for 4.11 Security Officer-supported branches

[Barry Bouwsma]

[sorry for the delay in this reply; I can't remember if I already
 sent a response...]

On Wed, 12 Jan 2005 11:46:13 -0800, Julian Elischer wrote:

> >>While a 4.12 will PROBABLY not happen, I do plan on continued MFCs of 
> >>important
> >>changes to RELENG_4 as I do not envision my custommers moving to 5.x 

> >Thanks, Julian, and I'd like to contemplate getting a few of my MFC

> >One of these things I've found very handy is the `mount' option
> >introduced in 5.x that allows one to specify with -F an alternate
> >to /etc/fstab to be used.

> we can probably MFC anything that doesn't break any existing ABI.

I'll plead ignorance and stupidity.  Would an ABI change be something
like what I needed to do in another project, changing
-  int (* seek) (data_source_t *source, long offset, int whence);
+  int (* seek) (data_source_t *source, ogg_int64_t offset, int whence);
in order to fix largefile support, when this is something expected
to be called by userland programs?  I'll guess yes.


> >In order to adopt this change, I had to add to libc in 4.x as well
> >as butcher the `mount' code.  Does such a change stand a chance of
> >being added to 4.x, or are infrastructure changes required this way,
> >like to libc, off-limits outside my own hive of personal hackery?

> libc is not usually a target in a legacy branch. Addign a new function to it
> may be ok in some cases but probably best to not do it.

I'll leave the call to you, or anyone else who knows better than I.

As far as I know, it's not easily possible to implement the `-F' mount
option without this libc change, but I haven't played with it as a MFC
did the job for me.  The diff below is based on what was -current when
this option was introduced, and is fairly clean, as follows -- the diff
needed for `mount' itself would follow later, if this diff to RELENG_4
libc does not raise serious objections:


--- /stand/FreeBSD4-src/source-hacks/lib/libc/gen/fstab.c-ORIG	Fri Jan 28 00:06:15 2000
+++ /stand/FreeBSD4-src/source-hacks/lib/libc/gen/fstab.c	Tue Jul 27 10:30:39 2004
@@ -57,11 +57,40 @@
 static FILE *_fs_fp;
 static struct fstab _fs_fstab;
 static int LineNo = 0;
+static char *path_fstab;
+static char fstab_path[PATH_MAX];
+static int fsp_set = 0;
 
 static void error __P((int));
 static void fixfsfile __P((void));
 static int fstabscan __P((void));
 
+void
+setfstab(const char *file)
+{
+
+	if (file == NULL) {
+		path_fstab = _PATH_FSTAB;
+	} else {
+		strncpy(fstab_path, file, PATH_MAX);
+		fstab_path[PATH_MAX - 1] = '\0';
+		path_fstab = fstab_path;
+	}
+	fsp_set = 1;
+
+	return;
+}
+
+const char *
+getfstab (void)
+{
+
+	if (fsp_set)
+		return (path_fstab);
+	else
+		return (_PATH_FSTAB);
+}
+
 static void
 fixfsfile()
 {
@@ -229,7 +258,13 @@
 		LineNo = 0;
 		return(1);
 	}
-	if ((_fs_fp = fopen(_PATH_FSTAB, "r")) != NULL) {
+	if (fsp_set == 0) {
+		if (issetugid())
+			setfstab(NULL);
+		else
+			setfstab(getenv("PATH_FSTAB"));
+	}
+	if ((_fs_fp = fopen(path_fstab, "r")) != NULL) {
 		LineNo = 0;
 		return(1);
 	}
@@ -244,6 +279,8 @@
 		(void)fclose(_fs_fp);
 		_fs_fp = NULL;
 	}
+
+	fsp_set = 0;
 }
 
 static void
@@ -254,7 +291,7 @@
 	char num[30];
 
 	(void)_write(STDERR_FILENO, "fstab: ", 7);
-	(void)_write(STDERR_FILENO, _PATH_FSTAB, sizeof(_PATH_FSTAB) - 1);
+	(void)_write(STDERR_FILENO, path_fstab, strlen(path_fstab));
 	(void)_write(STDERR_FILENO, ":", 1);
 	sprintf(num, "%d: ", LineNo);
 	(void)_write(STDERR_FILENO, num, strlen(num));



I'm pretty sure (but not positive) that these diffs do not have
tentacles afflicting other code, and should be safe to apply to a
virginal RELENG_4 system without major breakage.


thanks
barry bouwsma