Re: [RFC] iproute2: Fix meta match u32 with 0xffffffff

On Tue, 12 Apr 2011 09:56:49 +0200
Thomas Graf [off-list ref] wrote:

On Mon, 2011-04-11 at 11:52 -0700, Stephen Hemminger wrote: 

quoted

The value 0xffffffff is a valid mask and bstrtoul() would return
ULONG_MAX which was the error value. Resolve the problem by separating
return value and error indication.
 
-unsigned long bstrtoul(const struct bstr *b)
+int bstrtoul(const struct bstr *b, unsigned long *lp)
 {
 	char *inv = NULL;
-	unsigned long l;
 	char buf[b->len+1];
 
+	if (b->len == 0)
+		return -EINVAL;
+
 	memcpy(buf, b->data, b->len);
 	buf[b->len] = '\0';
 
-	l = strtoul(buf, &inv, 0);
-	if (l == ULONG_MAX || inv == buf)
-		return ULONG_MAX;
+	*lp = strtoul(buf, &inv, 0);
+	if (inv == buf)
+		return -EINVAL;
+
+	if (*lp == ULONG_MAX || errno == ERANGE)
+		return -ERANGE;
 
-	return l;
+	return 0;
 }

This is definitely much better but we still can't parse ULONG_MAX
as string representative. Checking glibc docs, the only way to do it is
to ignore the return value for error checking and look errno.

I think the error case is ret == ULONG_MAX && errno == ERANGE
If there is no error, then strtoul doesn't set errno.



--