Re: [Xen-devel] [PATCHv9 4/4] gnttab: use per-VCPU maptrack free lists

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 20.05.15 at 17:54, <david.vrabel@xxxxxxxxxx> wrote:
>  static inline int
>  get_maptrack_handle(
>      struct grant_table *lgt)
>  {
> +    struct vcpu          *v = current;
>      int                   i;
>      grant_handle_t        handle;
>      struct grant_mapping *new_mt;
> -    unsigned int          new_mt_limit, nr_frames;
> +    unsigned int          nr_frames;
>  
> -    spin_lock(&lgt->maptrack_lock);
> +    handle = __get_maptrack_handle(lgt, v);
> +    if ( likely(handle != -1) )
> +        return handle;
>  
> -    while ( unlikely((handle = __get_maptrack_handle(lgt)) == -1) )
> -    {
> -        nr_frames = nr_maptrack_frames(lgt);
> -        if ( nr_frames >= max_maptrack_frames )
> -            break;
> +    nr_frames = nr_vcpu_maptrack_frames(v);
> +    if ( nr_frames >= max_maptrack_frames )

I think you want to check lgt->maptrack_pages here, to make sure
the global limit remains a per-domain one. Or else we'd need a dual
(Dom0/DomU) or even per-domain bound to avoid many-vCPU
guests being able to force Xen into using significant amounts of
memory for their maptrack.

> @@ -1447,6 +1484,17 @@ gnttab_setup_table(
>      gt = d->grant_table;
>      write_lock(&gt->lock);
>  
> +    /* Tracking of mapped foreign frames table */
> +    if ( (gt->maptrack = xzalloc_array(struct grant_mapping *,
> +                                       max_maptrack_frames * d->max_vcpus)) 
> == NULL )
> +        goto out3;

Not that vzalloc() meanwhile exists. But depending on the resolution
of the aspect above you may not even urgently need it here.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel