Jan Beulich
14 March 2014 15:23
We can't just add in the hole shift value, as the hole may be at or
above the 44-bit boundary. Instead we need to determine the total bit
count until reaching 32 significant (not squashed out) bits in PFN
representations.

Signed-off-by: Jan Beulich <jbeulich@suse.com>

--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -180,6 +180,19 @@ void dump_pageframe_info(struct domain *
spin_unlock(&d->page_alloc_lock);
}

+static unsigned int __init noinline _domain_struct_bits(void)
+{
+ unsigned int bits = 32 + PAGE_SHIFT;
+ unsigned int sig = hweight32(~pfn_hole_mask);
+ unsigned int mask = pfn_hole_mask >> 32;
+
+ for ( ; bits < BITS_PER_LONG && sig < 32; ++bits, mask >>= 1 )
+ if ( !(mask & 1) )
+ ++sig;

This logic took a few looks to be sure what it was doing, let alone correct. Without the patch comment I'd be even a bit more lost. I think the changeset comment, or similar, should be included as a code comment for this new function.

I'm sure given there is only one hole there is a simpler form of this based on checking whether bottom_shift > 44. But this is more general I suppose, so probably better despite being harder to understand.

Acked-by: Keir Fraser <keir@xen.org>
+ return bits;
+}
+
struct domain *alloc_domain_struct(void)
{
struct domain *d;
@@ -187,7 +200,10 @@ struct domain *alloc_domain_struct(void)
* We pack the PDX of the domain structure into a 32-bit field within
* the page_info structure. Hence the MEMF_bits() restriction.
*/
- unsigned int bits = 32 + PAGE_SHIFT + pfn_pdx_hole_shift;
+ static unsigned int __read_mostly bits;
+
+ if ( unlikely(!bits) )
+ bits = _domain_struct_bits();

BUILD_BUG_ON(sizeof(*d) > PAGE_SIZE);
d = alloc_xenheap_pages(0, MEMF_bits(bits));