Re: [patch] voluntary-preempt-2.6.8.1-P1

[Ingo Molnar]

* Ingo Molnar <mingo@xxxxxxx> wrote:

> > Yes, Ingo identified an issue with copy_page_range, I don't think it's
> > fixed yet.  See the voluntary-preempt-2.6.8.1-P0 thread.
> 
> right, it's not fixed yet. It's not a trivial critical section - we
> are holding two locks and are mapping two atomic kmaps.

fortunately it's easier than i thought - neither the source pmd nor the
target pmd can go away so we can simply drop the locks, reschedule, and
remap. Does the patch below (ontop of -P3) fix the copy_page_range()
latencies you are seeing?

	Ingo

--- linux/mm/memory.c.orig	
+++ linux/mm/memory.c	
@@ -337,6 +337,15 @@ cont_copy_pte_range_noset:
 				}
 				src_pte++;
 				dst_pte++;
+				if (voluntary_need_resched()) {
+					pte_unmap_nested(src_pte);
+					pte_unmap(dst_pte);
+					spin_unlock(&src->page_table_lock);
+					cond_resched_lock(&dst->page_table_lock);
+					spin_lock(&src->page_table_lock);
+					dst_pte = pte_offset_map(dst_pmd, address);
+					src_pte = pte_offset_map_nested(src_pmd, address);
+				}
 			} while ((unsigned long)src_pte & PTE_TABLE_MASK);
 			pte_unmap_nested(src_pte-1);
 			pte_unmap(dst_pte-1);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/