Re: process creation time increases linearly with shmem

[Rik van Riel]

On Fri, 26 Aug 2005, Nick Piggin wrote:

> > Skipping MAP_SHARED in fork() sounds like a good idea to me...
> 
> Indeed. Linus, can you remember why we haven't done this before?

Where "this" looks something like the patch below, shamelessly
merging Nick's and Andy's patches and adding the initialization
of retval.

I suspect this may be a measurable win on database servers with
a web frontend, where the connections to the database server are
set up basically for each individual query, and don't stick around
for a long time.

No, I haven't actually tested this patch - but feel free to go
wild while I sign off for the night.

Signed-off-by: Rik van Riel <riel@xxxxxxxxxx>

--- linux-2.6.12/kernel/fork.c.mapshared	2005-08-25 18:40:44.000000000 -0400
+++ linux-2.6.12/kernel/fork.c	2005-08-25 18:47:16.000000000 -0400
@@ -184,7 +184,7 @@
 {
 	struct vm_area_struct * mpnt, *tmp, **pprev;
 	struct rb_node **rb_link, *rb_parent;
-	int retval;
+	int retval = 0;
 	unsigned long charge;
 	struct mempolicy *pol;
 
@@ -265,7 +265,10 @@
 		rb_parent = &tmp->vm_rb;
 
 		mm->map_count++;
-		retval = copy_page_range(mm, current->mm, tmp);
+		/* Skip pte copying if page faults can take care of things. */
+		if (!file || !(tmp->vm_flags & VM_SHARED) ||
+						is_vm_hugetlb_page(vma))
+			retval = copy_page_range(mm, current->mm, tmp);
 		spin_unlock(&mm->page_table_lock);
 
 		if (tmp->vm_ops && tmp->vm_ops->open)
-
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/