On Mon, 2004-12-06 at 14:54, Mark Wong wrote: > add_memory(0000000040000000, 0000000008000000, 00000000 > __add_pages(c03bd300, 00040000, 32768, 00000000) > for: i: 0 > __add_section(c03bd300, 00040000, 00000000) > __add_section() phys_start_pfn: 00040000 > alloc_memsections(00040000, 00040000, 00008000) > set mem_section[8].phys_section: 8 > set phys_section[8]: 8 > __kmalloc_section_memmap() set page: c16e2000 set mapping to c0166510 > page_count(page): 1 > register_memory() with section: 8 > memory_block_action() > psection: 8 > first_page: f7100000 > last_page: f71fffe0 > phys_index: 8 > phys_section[ 8]: 8 > online_pages: onlining 0x8000 pages starting from pfn: 0x40000 > online_pages: onlining if 0x8000 pages done. Nothing unusual there, just standard debug output. The next step is to instrument the page fault path. We're probably just missing some kind of initialization in the AIO path. It's going to take some more in-depth digging. -- Dave
