On Mon, Jun 02, 2014 at 11:19:18AM -0700, Dave Hansen wrote:
> On 06/02/2014 10:14 AM, Naoya Horiguchi wrote:
> > Yes, that's necessary to consider (but I haven't done, sorry),
> > so I'm thinking of moving this definition to the new file
> > include/uapi/linux/pagecache.h and let it be imported from the
> > userspace programs. Is it fine?
>
> Yep, although I'd probably also explicitly separate the definitions of
> the user-exposed ones from the kernel-internal ones. We want to make
> this hard to screw up.
>
> I can see why we might want to expose dirty and writeback out to
> userspace, especially since we already expose the aggregate, system-wide
> view in /proc/meminfo. But, what about PAGECACHE_TAG_TOWRITE? I really
> can't think of a good reason why userspace would ever care about it or
> consider it different from PAGECACHE_TAG_DIRTY.
I guess that TOWRITE tag might be useful to predict IO behavior
("which pages are to be writeback next" type of information).
But it's not clear to me how. I hope that DB developers have some
idea about good usecases of this tag for userspace.
Thanks,
Naoya Horiguchi
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>