Re: [Xen-devel] [RFC PATCH V2 1/4] xen/hap: Increase hap page pool size for more vcpus support

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 31/08/17 06:01, Lan Tianyu wrote:
> This patch is to increase hap page pool size to support more vcpus in single 
> VM.
>
> Signed-off-by: Lan Tianyu <tianyu.lan@xxxxxxxxx>
> ---
>  xen/arch/x86/mm/hap/hap.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/xen/arch/x86/mm/hap/hap.c b/xen/arch/x86/mm/hap/hap.c
> index cdc77a9..96a7ed0 100644
> --- a/xen/arch/x86/mm/hap/hap.c
> +++ b/xen/arch/x86/mm/hap/hap.c
> @@ -464,6 +464,7 @@ void hap_domain_init(struct domain *d)
>  int hap_enable(struct domain *d, u32 mode)
>  {
>      unsigned int old_pages;
> +    unsigned int pages;
>      unsigned int i;
>      int rv = 0;
>  
> @@ -473,7 +474,14 @@ int hap_enable(struct domain *d, u32 mode)
>      if ( old_pages == 0 )
>      {
>          paging_lock(d);
> -        rv = hap_set_allocation(d, 256, NULL);
> +
> +        /* Increase hap page pool with max vcpu number. */
> +        if ( d->max_vcpus > 128 )
> +            pages = 256;
> +        else
> +            pages = 512;
> +
> +        rv = hap_set_allocation(d, pages, NULL);

What effect is this intended to have?  hap_enable() is always called
when d->max_vcpus is 0.

d->max_vcpus isn't chosen until a subsequent hypercall.  (This is one of
many unexpected surprised from multi-vcpu support having been hacked on
the side of existing Xen support, rather than being built in to the
createdomain hypercall).

~Andrew

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