Thread (17 messages) 17 messages, 4 authors, 2012-08-16

Re: [PATCH] mm: hugetlb: flush dcache before returning zeroed huge page to userspace

From: Michal Hocko <hidden>
Date: 2012-08-16 18:06:19
Also in: linux-mm, lkml

On Thu 16-08-12 18:34:59, Will Deacon wrote:
On Thu, Aug 16, 2012 at 06:25:27PM +0100, Michal Hocko wrote:
quoted
On Thu 16-08-12 17:09:54, Will Deacon wrote:
quoted
On Wed, Aug 08, 2012 at 05:26:07PM +0100, Michal Hocko wrote:
quoted
I guess the cleanest way is to hook into dequeue_huge_page_node and add
something like arch_clear_hugepage_flags.
I hooked into enqueue_huge_page instead, but how about something like this?:
Do you have any specific reason for that? enqueue_huge_page is called on
pages which potentially never get used so isn't that wasting a bit?
Not that it would be wrong I was just thinking why shouldn't we do it
when the page is actualy going to be used for sure.
I just did it that way to match the flag clearing for normal pages. I can
move it into dequeue if you think it's worthwhile but in the worst case it
just adds a clear_bit call, so I doubt it's measurable.
I do not have a strong opinion on that but the flags come in cleared when
they are freshly allocated (gather_surplus_pages) so then it would be
more appropriate in free_huge_page when enqueue_huge_page is called.

But this is just a nit.
Will
-- 
Michal Hocko
SUSE Labs
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help