Re: [PATCH] x86: use pgd accessors when cloning a pgd range.

[Ian Campbell <Ian.Campbell@eu.citrix.com>]

On Wed, 2010-10-27 at 18:44 +0100, Borislav Petkov wrote:
> On Wed, Oct 27, 2010 at 10:31:37AM -0700, Jeremy Fitzhardinge wrote:
> > > Xen shouldn't have any users of this, since it's used for low-level
> > > operations like SMP bootstrap, suspend to RAM, reboot and low-level
> > > BIOS functionality.
> > >
> > Right, but it is being called smack in the middle of setup_arch().  It
> > looks like they could be hidden away in
> > native_pagetable_setup_start/done though.
> 
> I think I can put the second sync-back clone_pgd_range in
> native_pagetable_setup_done but the first one needs to take place right
> at the beginning of setup_arch() because a lot of code in-between relies
> on swapper_pg_dir containing proper entries.
> 
> Just to make sure I understand you correctly: best it would be for xen
> to not have the copy_pgd_range() calls in setup_arch, correct?

Yes, or to ensure that swapper_pg_dir has not yet been marked as a
pagetable at this point. Under Xen we are running on swapper_pg_dir when
setup_arch is called.

I think we could achieve this within the Xen pvops backend but e.g. the
following would likely work also.

(Note, untested. I also omitted the reindent for clarity at this point)

diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index 95a3274..2d89d5a 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -705,12 +705,14 @@ void __init setup_arch(char **cmdline_p)
 	 * copy kernel address range established so far and switch
 	 * to the proper swapper page table
 	 */
+	if (read_cr3() != __pa(swapper_pg_dir)) {
 	clone_pgd_range(swapper_pg_dir     + KERNEL_PGD_BOUNDARY,
 			initial_page_table + KERNEL_PGD_BOUNDARY,
 			KERNEL_PGD_PTRS);
 
 	load_cr3(swapper_pg_dir);
 	__flush_tlb_all();
+	}
 #else
 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
 #endif