On Thu, Jun 15, 2023 at 07:58:31AM +0200, Christoph Hellwig wrote: > On Wed, Jun 14, 2023 at 02:48:52PM +0100, Matthew Wilcox (Oracle) wrote: > > This is the folio equivalent of memcpy_to_page(), but it handles large > > highmem folios. It may be a little too big to inline on systems that > > have CONFIG_HIGHMEM enabled but on systems we actually care about almost > > all the code will be eliminated. > > I suspect the right thing is to have the trivial version without kmapping > for !HIGHMEM inline, and a separate version with the kmap loop out of > line for HIGHMEM builds. > > Same for the next patch. Yeah, that's what I did to zero_user_segments(). As time goes by, I'm starting to care about performance of CONFIG_HIGHMEM systems less and less.
