Re: [PATCH] use a nocache copy for bvecs in copy_from_iter_nocache()

[Al Viro]

On Fri, Oct 07, 2016 at 10:55:11AM -0500, Brian Boylston wrote:
> copy_from_iter_nocache() is only "nocache" for iovecs.  Enhance it to also
> use a nocache copy for bvecs.  This improves performance by 2-3X when
> splice()ing to a file in a DAX-mounted, pmem-backed file system.

> +static void memcpy_from_page_nocache(char *to, struct page *page, size_t offset, size_t len)
> +{
> +	char *from = kmap_atomic(page);
> +	__copy_from_user_inatomic_nocache(to, from, len);
> +	kunmap_atomic(from);
> +}

At the very least, it will blow up on any architecture with split
userland and kernel MMU contexts.  You *can't* feed a kernel pointer
to things like that and expect it to work.  At the very least, you
need to add memcpy_nocache() and have it default to memcpy(), with
non-dummy version on x86.  And use _that_, rather than messing with
__copy_from_user_inatomic_nocache()