On Wed, 22 Jan 2014, Alex Thorlton wrote: > > I would expect that the alternative solution to a per-mm switch is to > > reserve the naturally aligned pages for a THP promotion. Have a threshold > > of pages pages that must be faulted before the full THP's worth of pages > > is allocated, zero'd and a huge pmd established. That would defer the > > THP setup costs until it was detected that it was necessary. > > I have some half-finished patches that I was working on a month or so > ago, to do exactly this (I think you were involved with some of the > discussion, maybe? I'd have to dig up the e-mails to be sure). After > cycling through numerous other methods of handling this problem, I still > like that idea, but I think it's going to require a decent amount of > effort to get finished. >
If you're going to go this route, I think a sane value would be max_ptes_none that controls when khugepaged would re-collapse a split hugepage into another hugepage after something does madvise(MADV_DONTNEED) to a region, for example. Otherwise it could collapse that memory into a hugepage even though it wasn't faulted as huge. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/