On Mon, 28 Apr 2014 15:56:05 +0800 Baoquan He <bhe at redhat.com> wrote: > Hi Astushi, > > I remember earlier Jingbai posted a patchset on hugepage filtering, > later you pasted your hugepage filtering prototype patch, and people > thought yours is better. So now may I know the status and what's your > plan? > > Because in makedumpfile 1.5.6, this is not delivered. Your plan can help > us much. Hi Baoquan, I followed up on Atsushi's prototype, and there's a patch series sitting on my disk which just came too late for 1.5.6, so I held it back. It was also suggested that the mdf_pfn cleanup patch be applied first. That one is now in, and I'll send an updated series shortly. The first part of the change is the "Generic handling of multi-page exclusions", that is a method which can be used for excluding both free pages and compound pages (and without adding even more complexity to the calculation of the cycle size). The second part is handling of hugepages. This simply extends the algorithm so that all pages in a compound page are treated as one object. To make it work, additional info must be exported from the kernel. This is where I got stuck with my previous attempt. For now, this patch is already in the mm tree: http://ozlabs.org/~akpm/mmotm/broken-out/kexec-save-pg_head_mask-in-vmcoreinfo.patch Petr T
