Re: [RFC] de-asmify the x86-64 system call slowpath

[Linus Torvalds]

On Thu, Feb 6, 2014 at 2:24 PM, Kirill A. Shutemov <kirill@xxxxxxxxxxxxx> wrote:
>
> If we want to reduce number of page fault with less overhead we probably
> should concentrate on minor page fault -- populate pte around fault
> address which already in page cache. It should cover scripting use-case
> pretty well.

That's what my patch largely does. Except I screwed up and didn't use
FAULT_FLAG_ALLOW_RETRY in fault_around().

Anyway, my patch kind of works, but I'm starting to hate it. I think I
want to try to extend the "->fault()" interface to allow
filemap_fault() to just fill in multiple pages.

We alread have that "vmf->page" thing, we could make it a small array
easily. That would allow proper gang lookup, and much more efficient
"fill in multiple entries in one go" in mm/memory.c.

> Heh! `man 3 mmap' actually suggests MAP_POPULATE | MAP_NONBLOCK.
> What's the story beyond MAP_NONBLOCK?

It does nothing, afaik.

                  Linus
--
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/