Re: early fixmap causes kmap breakage

[Nick Piggin <npiggin@suse.de>]

On Tue, Dec 30, 2008 at 08:54:28AM +0100, Nick Piggin wrote:
> On Tue, Dec 30, 2008 at 07:13:44AM +0100, Ingo Molnar wrote:
> > 
> > * Nick Piggin <npiggin@suse.de> wrote:
> > 
> > > On Mon, Dec 29, 2008 at 03:17:31PM -0800, Andrew Morton wrote:
> > > > On Thu, 18 Dec 2008 22:15:43 +0100
> > > > Nick Piggin <npiggin@suse.de> wrote:
> > > > 
> > > > > Hi,
> > > > > 
> > > > > I've debugged a problem where i386+pae systems with more than a few CPUs
> > > > > blow up at boot in the kmap_atomic code.
> > > > 
> > > > ping?
> > > 
> > > No further progress here, I'm waiting on input for how to fix this 
> > > "nicely". Meantime, clearing the early fixmap pte I guess works, but you 
> > > lose a page... is it possible to put it into .initdata or is there some 
> > > issue with that? (I guess on a PAE kernel, 4K isn't a big deal).
> > 
> > yeah, 4K shouldnt be a big deal. Mind sending a patch for this?
> 
> I don't know if we should put a warning when we encounter this condition,
> to help catch other unintended uses...
> 
> But this at least clears out the early fixmap pte and provides a contiguous
> allocation...
> 
> ---
>  arch/x86/mm/init_32.c |   15 ++++++++++++++-
>  1 file changed, 14 insertions(+), 1 deletion(-)
> 
> Index: linux-2.6/arch/x86/mm/init_32.c
> ===================================================================
> --- linux-2.6.orig/arch/x86/mm/init_32.c
> +++ linux-2.6/arch/x86/mm/init_32.c
> @@ -155,6 +155,7 @@ page_table_range_init(unsigned long star
>  	unsigned long vaddr;
>  	pgd_t *pgd;
>  	pmd_t *pmd;
> +	pte_t *lastpte = NULL;
>  
>  	vaddr = start;
>  	pgd_idx = pgd_index(vaddr);
> @@ -166,7 +167,19 @@ page_table_range_init(unsigned long star
>  		pmd = pmd + pmd_index(vaddr);
>  		for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end);
>  							pmd++, pmd_idx++) {
> -			one_page_table_init(pmd);
> +			pte_t *pte;
> +
> +			pte = one_page_table_init(pmd);
> +			if (lastpte && lastpte + PTRS_PER_PTE != pte) {
> +				pte_t *newpte;
> +
> +				pmd_clear(pmd);

Hmm, no actually we do need a TLB flush here.

> +				newpte = one_page_table_init(pmd);
> +				BUG_ON(lastpte + PTRS_PER_PTE != newpte);
> +				memset(newpte, pte, PAGE_SIZE);

And here we probably need a loop doing set_pte just in case the
CPU speculatively loads an incomplete TLB...

> +				pte = lastpte;
> +			}
> +			lastpte = pte;
>  
>  			vaddr += PMD_SIZE;
>  		}