On Sat, May 23, 2015 at 7:12 AM, hch(a)lst.de <hch(a)lst.de> wrote:
On Wed, May 13, 2015 at 06:35:55PM +0000, Williams, Dan J wrote:
> Jens, I'm wondering if you want to take this series(.) as patches or
> prepare a git branch to pull?
Honestly I don't think it should go anyway. It makes a big mess of
a structure without providing a real user for it. Given how we are
using the bio_vec for in-kernel page based I/O these days it seems
like a very dangerous idea.
There's nothing dangerous about the __pfn_t conversion of the block
layer in the !CONFIG_DEV_PFN case a __pfn_t based bio_vec is
bit-for-bit identical to a struct page based bio_vec. However, you're
right, I can't make the same claim about a scatterlist before and
after the change.
Hmm, we're missing a pfn-only block I/O user and we're missing the
second half of the implementation that provides __pfn_t_to_page() for
persistent memory. I'm looking to have a solution __pfn_t_to_page()
shortly, maybe that will allow the scatterlist changes to be
skipped... we'll see.