On Fri, 2005-04-08 at 15:32 +0200, Franck Bui-Huu wrote: > As I described in my previous email, bootmem.c does improper > pfn convertions into phys addr. This simple patch fixes that. ... > - bdata->node_bootmem_map = phys_to_virt(mapstart << PAGE_SHIFT); > - bdata->node_boot_start = (start << PAGE_SHIFT); > + bdata->node_bootmem_map = phys_to_virt(pfn_to_phys(mapstart)); > + bdata->node_boot_start = pfn_to_phys(start);
The only arch with phys_to_pfn() defined is UML, so the patch simply won't compile anything but UML on current kernels (unless I'm missing something). Could you try to give us a more complete description of your problem? I know your memory doesn't start at 0x0, but what problems does that cause? Does the mem_map[] allocation blow up, etc... If it's just mem_map[], That calculation could be fixed pretty easily. Something like +#ifdef CONFIG_CRAZY_MIPS_FOO_MEM_MAP_START... +extern unsigned long mem_map_start_pfn +#else +#define mem_map_start_pfn 0UL +#endif -#define pfn_to_page(pfn) (mem_map + (pfn)) +#define pfn_to_page(pfn) (mem_map + (pfn) - mem_map_start_pfn) (those names are horrid, please improve them, if you plan to do this) All of the zone (and allocator) calculations should be just fine, because it already has a zone_start_pfn. -- Dave - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
