On Thu, Jul 23, 2009 at 11:31 PM, Pavel Roskin<pro...@gnu.org> wrote: > On Thu, 2009-07-23 at 11:38 +0200, Vladimir 'phcoder' Serbinenko wrote: >> Hello. According to xen some BIOSes update only lower 32-bit in mmap >> entries. To workaround this and not get high values in memory map >> zero-fill before calling BIOS > > I think the fix belongs to grub_get_mmap_entry(), not to all of its > callers. I'm ok with this too. The main reason for me to decide for one or another would be core.img size but I haven't checked yet which one is smaller. Actually asm part is kept to minimum and grub_machine_mmap_iterate is the C wrapper and I followed this idea. But as I said your solution is fine for me too. > Something like this: > > diff --git a/kern/i386/pc/startup.S b/kern/i386/pc/startup.S > index be258fb..5468ba8 100644 > --- a/kern/i386/pc/startup.S > +++ b/kern/i386/pc/startup.S > @@ -997,6 +997,14 @@ FUNCTION(grub_get_mmap_entry) > /* push ADDR */ > pushl %eax > > + /* clear the request area, buggy BIOSes may not clear it */ > + xor %edi, %edi > + movl %edi, 4(%eax) > + movl %edi, 8(%eax) > + movl %edi, 12(%eax) > + movl %edi, 16(%eax) > + movl %edi, 20(%eax) > + > /* place address (+4) in ES:DI */ > addl $4, %eax > movl %eax, %edi For me this coode looks weird because of using edi as zero register and eax as pointer. Could we perhaps move this code down and use eax as 0 and edi as pointer? > > I don't know whether it would interfere with the Apple CC workaround. > It would be great if we get rid of it, perhaps by using labels starting > with "L_". > > -- > Regards, > Pavel Roskin > > > _______________________________________________ > Grub-devel mailing list > Grub-devel@gnu.org > http://lists.gnu.org/mailman/listinfo/grub-devel >
-- Regards Vladimir 'phcoder' Serbinenko Personal git repository: http://repo.or.cz/w/grub2/phcoder.git _______________________________________________ Grub-devel mailing list Grub-devel@gnu.org http://lists.gnu.org/mailman/listinfo/grub-devel
