Andrew Morton wrote: > On Thu, 15 Feb 2007 18:25:00 -0800 Jeremy Fitzhardinge <jeremy at goop.org> wrote: > > >>Add a new mm function apply_to_page_range() which applies a given >>function to every pte in a given virtual address range in a given mm >>structure. This is a generic alternative to cut-and-pasting the Linux >>idiomatic pagetable walking code in every place that a sequence of >>PTEs must be accessed. >> >>Although this interface is intended to be useful in a wide range of >>situations, it is currently used specifically by several Xen >>subsystems, for example: to ensure that pagetables have been allocated >>for a virtual address range, and to construct batched special >>pagetable update requests to map I/O memory (in ioremap()). > > > There was some discussion about this sort of thing last week. The > consensus was that it's better to run the callback against a whole pmd's > worth of ptes, mainly to amortise the callback's cost (a lot). > > It was implemented in > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.20/2.6.20-mm1/broken-out/smaps-extract-pmd-walker-from-smaps-code.patch Speaking of that patch, I missed the discussion, but I'd hope it doesn't go upstream in its current form. We now have one way of walking range of ptes. The code may be duplicated a few times, but it is simple, we know how it works, and it is easy to get right because everyone does the same thing. We used to have about a dozen slightly different ways of doing this until Hugh spent the effort to standardise it all. Isn't it nice? If we want an ever-so-slightly lower performing interface for those paths that don't care to count every cycle -- which I think is a fine idea BTW -- it should be implemented in mm/memory.c and it should use our standard form of pagetable walking. -- SUSE Labs, Novell Inc. Send instant messages to your online friends http://au.messenger.yahoo.com
