RE: IA64 non-contiguous memory space bugs

["Chen, Kenneth W" <kenneth.w.chen@intel.com>]

David Gibson wrote on Tuesday, February 21, 2006 6:26 PM
> Chen, Kenneth W wrote on Tuesday, February 21, 2006 5:32 PM
> > free_pgtables() has partial crap that the check of is_hugepage_only_range()
> > should be done on the entire vma range, not just the first hugetlb page.
> > Though, it's not possible to have a hugetlb vma while having normal page
> > instantiated inside that vma.  So the bug is mostly phantom.  For
> > correctness, it should be fixed.
> 
> Actually, from ppc64's point of view, the problem with the test is
> that the whole vma could be *less* than HPAGE_SIZE - we don't test
> that the address is aligned before checking is_hugepage_only_range().
> We thus can call hugetlb_free_pgd_range() on normal page VMAs - which
> we only get away with because the ppc64 hugetlb_free_pgd_range() is
> (so far) an alias for the normal free_pgd_range().
> 
> Your patch below is insufficient, because there's a second test of
> is_hugepage_only_range() further down.  However, instead of tweaking
> the tested ranges, I think what we really want to do is check for
> is_vm_hugetlb_page() instead.
> 
> I was worried before, but now that you point out it's the 'end'
> address which really matters, not the ceiling, that might be
> sufficient.  Um.. except that hugepages, unlike normal pages these
> days don't necessarily clean up all possible pagetable pages on
> unmap...  crud.  Still the patch below ought to be an improvement.


I agree.  It's an improvement with your patch.

For ia64:
Acked-by: Ken Chen <kenneth.w.chen@intel.com> 



> Index: working-2.6/mm/memory.c
> ===================================================================
> --- working-2.6.orig/mm/memory.c	2006-02-22 10:42:14.000000000 +1100
> +++ working-2.6/mm/memory.c	2006-02-22 13:22:07.000000000 +1100
> @@ -277,7 +277,7 @@ void free_pgtables(struct mmu_gather **t
>  		anon_vma_unlink(vma);
>  		unlink_file_vma(vma);
>  
> -		if (is_hugepage_only_range(vma->vm_mm, addr, HPAGE_SIZE)) {
> +		if (is_vm_hugetlb_page(vma)) {
>  			hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
>  				floor, next? next->vm_start: ceiling);
>  		} else {
> @@ -285,8 +285,7 @@ void free_pgtables(struct mmu_gather **t
>  			 * Optimization: gather nearby vmas into one call down
>  			 */
>  			while (next && next->vm_start <= vma->vm_end + PMD_SIZE
> -			  && !is_hugepage_only_range(vma->vm_mm, next->vm_start,
> -							HPAGE_SIZE)) {
> +			       && !is_vm_hugetlb_page(vma->vm_mm)) {
>  				vma = next;
>  				next = vma->vm_next;
>  				anon_vma_unlink(vma);

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/