On Thu, Oct 4, 2012 at 9:45 AM, Konrad Rzeszutek Wilk <konrad.w...@oracle.com> wrote: >> So could use ram under 1M to map that page table at first. > > Could or does this patch do it? And why 1MB?
can you or stefano could test attached patch on xen ? that will map the page table buffer that will be used. under 1M, still 4k there, so there will be no page table around 512M. >> >> so that will make it xen happy ? > > The issues that Xen faces are purely due to the fact that they > must be RO when they are in use. I believe (and without actually > checking it just to make sure) that it does not matter where > the page-tables are located. But with the current generic code > the location is quite linear: it starts with pgt_buf_start and > goes up to pgt_buf_top. So how would this patch move the location > of the page-table to be under 1MB? just page table buf's page table. Thanks Yinghai
fix_xen_with_init_mapping.patch
Description: Binary data