RE: hugetlb demand paging patch part [2/3]

[Chen, Kenneth W]

>>>> David Gibson wrote on Thursday, April 15, 2004 12:17 AM
> > diff -Nurp linux-2.6.5/mm/memory.c linux-2.6.5.htlb/mm/memory.c
> > +++ linux-2.6.5.htlb/mm/memory.c	2004-04-13 12:02:31.000000000 -0700
> > @@ -769,11 +769,6 @@ int get_user_pages(struct task_struct *t
> >  		if ((pages && vm_io) || !(flags & vma->vm_flags))
> >  			return i ? : -EFAULT;
> >
> > -		if (is_vm_hugetlb_page(vma)) {
> > -			i = follow_hugetlb_page(mm, vma, pages, vmas,
> > -						&start, &len, i);
> > -			continue;
> > -		}
> >  		spin_lock(&mm->page_table_lock);
> >  		do {
> >  			struct page *map = NULL;
>
> Ok, I notice that you've removed the follow_hugtlb_page() function
> (and from the arch specific stuff, as well).  As far as I can tell,
> this isn't actually related to demand-paging, in fact as far as I can
> tell this function is unnecessary

That was the reason I removed the function because it is no longer used
with demand paging.


> should already work for huge pages.  In particular the path in
> get_user_pages() which can call handle_mm_fault() (which won't work on
> hugepages without your patch) should never get triggered, since
> hugepages are all prefaulted.

> Does that sound right?  In other words, do you think the patch below,
> which just kills off follow_hugetlb_page() is safe, or have I missed
> something?
>
> Index: working-2.6/mm/memory.c
> ===================================================================
> --- working-2.6.orig/mm/memory.c	2004-04-13 11:42:42.000000000 +1000
> +++ working-2.6/mm/memory.c	2004-04-15 17:03:01.421905400 +1000
> @@ -766,16 +766,13 @@
> [snip]
>  		spin_lock(&mm->page_table_lock);
>  		do {
>  			struct page *map;
>  			int lookup_write = write;
>  			while (!(map = follow_page(mm, start, lookup_write))) {
> +				/* hugepages should always be prefaulted */
> +				BUG_ON(is_vm_hugetlb_page(vma));
>  				/*
>  				 * Shortcut for anonymous pages. We don't want
>  				 * to force the creation of pages tables for

This portion is incorrect, because it will trigger BUG_ON all the time
for faulting hugetlb page.

Yes, killing follow_hugetlb_page() is safe because follow_page() takes
care of hugetlb page.  See 2nd patch posted earlier in
hugetlb_demanding_generic.patch


-
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/