[PATCH 3/6] xen: modify memory ops to be NUMA-aware

[Ryan Harper <ryanh@us.ibm.com>]

This patch modifies three memory operations to be NUMA-aware:

increase_reservation
populate_physmap
memory_exchange

These three operations request memory from the domain heap and have been
modified to distribute the request across the physical cpus of the
target domain evenly.  This make memory local to the physical cpus
within the domain available for the guest.

-- 
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
(512) 838-9253   T/L: 678-9253
diffstat output:
 memory.c |   42 ++++++++++++++++++++++++++++++++++--------
 1 files changed, 34 insertions(+), 8 deletions(-)

Signed-off-by: Ryan Harper <ryanh@us.ibm.com>
---
diff -r ac0a7a5a6425 xen/common/memory.c
--- a/xen/common/memory.c	Mon Jul  3 17:12:37 2006
+++ b/xen/common/memory.c	Sat Jul  8 12:27:19 2006
@@ -40,6 +40,12 @@
     struct page_info *page;
     unsigned long i;
     xen_pfn_t mfn;
+    int max_vcpu_id = 0;
+    struct vcpu *v;
+
+    for_each_vcpu (d, v) 
+        if ( v->vcpu_id > max_vcpu_id )
+            max_vcpu_id = v->vcpu_id;
 
     if ( !guest_handle_is_null(extent_list) &&
          !guest_handle_okay(extent_list, nr_extents) )
@@ -56,9 +62,11 @@
             *preempted = 1;
             return i;
         }
-
-        if ( unlikely((page = alloc_domheap_pages(
-            d, extent_order, memflags)) == NULL) )
+        /* spread each allocation across the total number of 
+         * vcpus allocated to this domain */
+        if ( unlikely((page = __alloc_domheap_pages( d, 
+            (d->vcpu[i % (max_vcpu_id+1)])->processor,
+            extent_order, memflags )) == NULL) ) 
         {
             DPRINTK("Could not allocate order=%d extent: "
                     "id=%d memflags=%x (%ld of %d)\n",
@@ -91,6 +99,12 @@
     unsigned long i, j;
     xen_pfn_t gpfn;
     xen_pfn_t mfn;
+    int max_vcpu_id = 0;
+    struct vcpu *v;
+
+    for_each_vcpu (d, v) 
+        if ( v->vcpu_id > max_vcpu_id )
+            max_vcpu_id = v->vcpu_id;
 
     if ( !guest_handle_okay(extent_list, nr_extents) )
         return 0;
@@ -110,8 +124,11 @@
         if ( unlikely(__copy_from_guest_offset(&gpfn, extent_list, i, 1)) )
             goto out;
 
-        if ( unlikely((page = alloc_domheap_pages(
-            d, extent_order, memflags)) == NULL) )
+        /* spread each allocation across the total number of 
+         * vcpus allocated to this domain */
+        if ( unlikely((page = __alloc_domheap_pages( d, 
+            (d->vcpu[i % (max_vcpu_id+1)])->processor,
+            extent_order, memflags )) == NULL) ) 
         {
             DPRINTK("Could not allocate order=%d extent: "
                     "id=%d memflags=%x (%ld of %d)\n",
@@ -293,10 +310,11 @@
     unsigned long in_chunk_order, out_chunk_order;
     xen_pfn_t     gpfn, gmfn, mfn;
     unsigned long i, j, k;
-    unsigned int  memflags = 0;
+    unsigned int  memflags = 0, max_vcpu_id = 0;
     long          rc = 0;
     struct domain *d;
     struct page_info *page;
+    struct vcpu *v;
 
     if ( copy_from_guest(&exch, arg, 1) )
         return -EFAULT;
@@ -367,6 +385,11 @@
     }
     d = current->domain;
 
+    /* calc max_vcpu_id */
+    for_each_vcpu (d, v) 
+        if ( v->vcpu_id > max_vcpu_id )
+            max_vcpu_id = v->vcpu_id;
+
     for ( i = 0; i < (exch.in.nr_extents >> in_chunk_order); i++ )
     {
         if ( hypercall_preempt_check() )
@@ -412,8 +435,11 @@
         /* Allocate a chunk's worth of anonymous output pages. */
         for ( j = 0; j < (1UL << out_chunk_order); j++ )
         {
-            page = alloc_domheap_pages(
-                NULL, exch.out.extent_order, memflags);
+            /* spread each allocation across the total number of 
+             * vcpus allocated to this domain */
+            page = __alloc_domheap_pages( NULL, 
+                  (d->vcpu[j % (max_vcpu_id+1)])->processor,
+                  exch.out.extent_order, memflags);
             if ( unlikely(page == NULL) )
             {
                 rc = -ENOMEM;
ryanh@us.ibm.com