Re: [Xen-devel] [PATCH RFC] p2m: use GNTTABOP_unmap_and_duplicate if available

[David Vrabel <david.vrabel@citrix.com>]

On 31/07/13 17:07, Roger Pau Monne wrote:
> The new GNTTABOP_unmap_and_duplicate operation doesn't zero the
> mapping passed in new_addr, allowing us to perform batch unmaps in p2m
> code without requiring the use of a multicall.

Thanks. This looks like it should be a good performance improvement for
userspace-based backends.  Do you have any performance measurements for
(e.g.) qemu's disk backend?

> --- a/arch/x86/xen/p2m.c
> +++ b/arch/x86/xen/p2m.c
[...]
> +int m2p_remove_override_batch(struct page **pages,
> +		struct gnttab_map_grant_ref *kmap_ops, int count)
> +{

Is it possible to refactor this function and
m2p_remove_override_single() to share some common helper functions?

> +	struct gnttab_unmap_and_duplicate *unmap_ops = NULL;
> +	unsigned long *mfn = kcalloc(count, sizeof(mfn[0]), GFP_KERNEL);
> +	unsigned long flags;
> +	unsigned long pfn;
> +	unsigned long uninitialized_var(address);
> +	unsigned level;
> +	struct page *page;
> +	pte_t *ptep = NULL;
> +	int ret = 0, i;
> +
> +
> +	for (i = 0; i < count; i++) {
> +		page = pages[i];
> +		pfn = page_to_pfn(page);

The bits in this loop...

[...]
> +	for (i = 0; i < count; i++) {
> +		/* p2m(m2p(mfn)) == FOREIGN_FRAME(mfn): the mfn is already present
> +		 * somewhere in this domain, even before being added to the

... and this one look they should be common between the two.

David