Fix up mm/mincore.c error value cases

Hugh Dickins correctly points out that mincore() is actually _supposed_
to fail on an unmapped hole in the user address space, rather than
return valid ("empty") information about the hole. This just simplifies
the problem further (I had been misled by our previous confusing and
complicated way of doing mincore()).

Also, in the unlikely situation that we can't allocate a temporary
kernel buffer, we should actually return EAGAIN, not ENOMEM, to keep the
"unmapped hole" and "allocation failure" error cases separate.

Finally, add a comment about our stupid historical lack of support for
anonymous mappings. I'll fix that if somebody reminds me after 2.6.20
is out.

Acked-by: Hugh Dickins <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

mm/mincore.c

···
 	struct vm_area_struct *vma = find_vma(current->mm, addr);
 
 	/*
-	 * find_vma() didn't find anything: the address
-	 * is above everything we have mapped.
+	 * find_vma() didn't find anything above us, or we're
+	 * in an unmapped hole in the address space: ENOMEM.
 	 */
-	if (!vma) {
-		memset(vec, 0, pages);
-		return pages;
-	}
-
-	/*
-	 * find_vma() found something, but we might be
-	 * below it: check for that.
-	 */
-	if (addr < vma->vm_start) {
-		unsigned long gap = (vma->vm_start - addr) >> PAGE_SHIFT;
-		if (gap > pages)
-			gap = pages;
-		memset(vec, 0, gap);
-		return gap;
-	}
+	if (!vma || addr < vma->vm_start)
+		return -ENOMEM;
 
 	/*
 	 * Ok, got it. But check whether it's a segment we support
 	 * mincore() on. Right now, we don't do any anonymous mappings.
+	 *
+	 * FIXME: This is just stupid. And returning ENOMEM is 
+	 * stupid too. We should just look at the page tables. But
+	 * this is what we've traditionally done, so we'll just
+	 * continue doing it.
 	 */
 	if (!vma->vm_file)
 		return -ENOMEM;
···
 
 	tmp = (void *) __get_free_page(GFP_USER);
 	if (!tmp)
-		return -ENOMEM;
+		return -EAGAIN;
 
 	retval = 0;
 	while (pages) {