Re: [RFC 1/3] mm/memory.c: introduce vm_insert_page(s)_mkspecial

[Xiaoguang Wang <xiaoguang.wang@xxxxxxxxxxxxxxxxx>]

hi,

> I relaly don't think we should add this amount of overhead to the
> core VM for a (relatively) irrelevant driver.
OK, but as what I have described in cover letter, both vm_insert_pages and
remap_pfn_range have performance or usage limits. Do you know any better
method to map block device io request's sgl pages to user space? I think 
block
device backend implemented in use space will benefit from this feature, 
thanks.

Regards,
Xiaoguang Wang