Re: [Xen-devel] [PATCH v5 3/4] mm: make MEMF_no_refcount pages safe to assign

[Jan Beulich <jbeulich@xxxxxxxx>]

On 29.01.2020 11:16, Paul Durrant wrote:
> --- a/xen/common/page_alloc.c
> +++ b/xen/common/page_alloc.c
> @@ -2287,11 +2287,14 @@ int assign_pages(
>  
>      for ( i = 0; i < (1 << order); i++ )
>      {
> +        unsigned long count_info = pg[i].count_info;
> +
>          ASSERT(page_get_owner(&pg[i]) == NULL);
> -        ASSERT(!pg[i].count_info);
> +        ASSERT(!(count_info & ~PGC_no_refcount));

I think this would benefit from being more strict: The flag may also
not be set if MEMF_no_refcount is clear in the flags passed in.

> @@ -2334,11 +2332,20 @@ struct page_info *alloc_domheap_pages(
>                                    memflags, d)) == NULL)) )
>           return NULL;
>  
> -    if ( d && !(memflags & MEMF_no_owner) &&
> -         assign_pages(d, pg, order, memflags) )
> +    if ( d && !(memflags & MEMF_no_owner) )
>      {
> -        free_heap_pages(pg, order, memflags & MEMF_no_scrub);
> -        return NULL;
> +        if ( memflags & MEMF_no_refcount )
> +        {
> +            unsigned long i;

With this you mean ...

> +            for ( i = 0; i < (1 << order); i++ )

... 1UL here. Or else "unsigned int" and 1U. The file isn't
very consistent with this, but at least locally to a single
functions things should match up.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel