* Dave Hansen (dave@sr71.net) wrote:

On 10/07/2013 01:21 PM, Robert C Jennings wrote:

quoted

+					} else {
+						if (vma)
+							zap_page_range(vma,
+								user_start,
+								(user_end -
+								 user_start),
+								NULL);
+						vma = find_vma_intersection(
+								current->mm,
+								useraddr,
+								(useraddr +
+								 PAGE_SIZE));
+						if (!IS_ERR_OR_NULL(vma)) {
+							user_start = useraddr;
+							user_end = (useraddr +
+								    PAGE_SIZE);
+						} else
+							vma = NULL;
+					}

This is pretty unspeakably hideous.  Was there truly no better way to do
this?

I was hoping to find a better way to coalesce pipe buffers and zap
entire VMAs (and it needs better documentation but your argument is with
structure and I agree). I would love suggestions for improving this but
that is not to say that I've abandoned it; I'm still looking for ways
to make this cleaner.

Doing find_vma() on a single page in the VMA rather than on each and
then zapping once provides a 50% runtime reduction for the writer when
tested with a 256MB vmsplice operation.  Based on the result I felt that
coalescing was justfied but the implementation is ugly.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>