On Fri, Jul 11, 2014 at 10:39:45PM +0800, Baoquan He wrote: > On 07/11/14 at 10:35am, Vivek Goyal wrote: > > On Fri, Jul 11, 2014 at 03:16:04PM +0800, Baoquan He wrote: > > > ->$ ./makedumpfile --mem-usage /proc/kcore > > > Excluding unnecessary pages : [100.0 %] | > > > > > > Page number of memory in different use > > > -------------------------------------------------- > > > TYPE PAGES EXCLUDABLE DESCRIPTION > > > pfn_zero 0x0000000000000000 yes Pages filled with zero > > > pfn_cache 0x000000000002b19f yes Cache pages > > > pfn_cache_private 0x0000000000042526 yes Cache pages + private > > > pfn_user 0x0000000000026bc3 yes User process data pages > > > pfn_free 0x0000000000133f79 yes Free pages > > > pfn_hwpoison 0x0000000000000000 yes Hwpoison pages > > > > Is there a need to display hwpoison pages? I think they are excluded by > > default. Otherwise reading them might kill makedumpfile process. > > You are right, hwpoison pages are excluded by default. But it truly > calculated in makedumpfile. I will change it as non-excludable like > below: > > TYPE PAGES EXCLUDABLE DESCRIPTION > pfn_hwpoison 0x0000000000000000 No Hwpoison pages I think we don't have to display hwpoison pages. It is of no use. Let us just display the pages which can be excluded using various filtering levels and total number of pages (without memory hole). That way people can do some math. amount of disk space required = total_pages - filtered pages; Thanks Vivek
