From: Bill Burns <bburns@redhat.com>
Date: Thu, 20 Dec 2007 14:45:58 -0500
Subject: [xen] provide NUMA memory usage information
Message-id: 476AC676.1060208@redhat.com
O-Subject: Re: [RHEL5.2 PATCH]: 0/2: NUMA xen tools and hypervisor fixes
Bugzilla: 235850

This patch addresses:

 BZ 235850: LTC33766: 200939: Xen Hypervisor - Provide NUMA memory usage information

This adds a new hypervisor sysctl hypercall 'XEN_SYSCTL_availheap' which
returns the amount of memory available in each NUMA node on a system.
This is a simple re-diff of the upstream change:

  changeset:   15534:29761c9b9105
  user:        kfraser@localhost.localdomain
  date:        Wed Jul 11 13:03:57 2007 +0100
  description:
  Add new domctl hypercall to expose current heap values.  This
  functionality is needed for probing how much memory is available in a
  given node prior to VM creation.

Since this is a completely new hypercall there is no ABI risk

 common/page_alloc.c     |   22 +++++++++++++++++-----
 common/sysctl.c         |   15 +++++++++++++++
 include/public/sysctl.h |   13 +++++++++++++
 include/xen/mm.h        |    2 ++
 4 files changed, 47 insertions(+), 5 deletions(-)

Dan.
--
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=|

Acked-by: Jarod Wilson <jwilson@redhat.com>

diff --git a/common/page_alloc.c b/common/page_alloc.c
index 4cbece4..81baf22 100644
--- a/common/page_alloc.c
+++ b/common/page_alloc.c
@@ -927,6 +927,23 @@ void free_domheap_pages(struct page_info *pg, unsigned int order)
         put_domain(d);
 }
 
+unsigned long avail_domheap_pages_region(
+    unsigned int node, unsigned int min_width, unsigned int max_width)
+{
+    int zone_lo, zone_hi;
+
+    zone_lo = min_width ? (min_width - (PAGE_SHIFT + 1)) : (MEMZONE_XEN + 1);
+    zone_lo = max_t(int, MEMZONE_XEN + 1, zone_lo);
+    zone_lo = min_t(int, NR_ZONES - 1, zone_lo);
+
+    zone_hi = max_width ? (max_width - (PAGE_SHIFT + 1)) : (NR_ZONES - 1);
+    zone_hi = max_t(int, MEMZONE_XEN + 1, zone_hi);
+    zone_hi = min_t(int, NR_ZONES - 1, zone_hi);
+
+    return avail_heap_pages(zone_lo, zone_hi, node);
+}
+
+
 
 unsigned long avail_domheap_pages(void)
 {
@@ -948,11 +965,6 @@ unsigned long avail_domheap_pages(void)
     return avail_nrm + avail_dma;
 }
 
-unsigned long avail_nodeheap_pages(int node)
-{
-    return avail_heap_pages(0, NR_ZONES - 1, node);
-}
-
 static void pagealloc_keyhandler(unsigned char key)
 {
     unsigned int zone = MEMZONE_XEN;
diff --git a/common/sysctl.c b/common/sysctl.c
index b839d7f..6fabf1d 100644
--- a/common/sysctl.c
+++ b/common/sysctl.c
@@ -21,6 +21,9 @@
 #include <xen/keyhandler.h>
 #include <asm/current.h>
 #include <public/sysctl.h>
+#include <asm/numa.h>
+#include <xen/nodemask.h>
+
 
 extern long arch_do_sysctl(
     struct xen_sysctl *op, XEN_GUEST_HANDLE(xen_sysctl_t) u_sysctl);
@@ -112,6 +115,18 @@ long do_sysctl(XEN_GUEST_HANDLE(xen_sysctl_t) u_sysctl)
     }
     break;
 
+    case XEN_SYSCTL_availheap:
+    {
+        op->u.availheap.avail_bytes = avail_domheap_pages_region(
+            op->u.availheap.node,
+            op->u.availheap.min_bitwidth,
+            op->u.availheap.max_bitwidth);
+        op->u.availheap.avail_bytes <<= PAGE_SHIFT;
+
+        ret = copy_to_guest(u_sysctl, op, 1) ? -EFAULT : 0;
+    }
+    break;
+
 #ifdef PERF_COUNTERS
     case XEN_SYSCTL_perfc_op:
     {
diff --git a/include/public/sysctl.h b/include/public/sysctl.h
index d51052a..11f6db5 100644
--- a/include/public/sysctl.h
+++ b/include/public/sysctl.h
@@ -167,6 +167,18 @@ struct xen_sysctl_getcpuinfo {
 typedef struct xen_sysctl_getcpuinfo xen_sysctl_getcpuinfo_t;
 DEFINE_XEN_GUEST_HANDLE(xen_sysctl_getcpuinfo_t); 
 
+#define XEN_SYSCTL_availheap         9
+struct xen_sysctl_availheap {
+    /* IN variables. */
+    uint32_t min_bitwidth;  /* Smallest address width (zero if don't care). */
+    uint32_t max_bitwidth;  /* Largest address width (zero if don't care). */
+    int32_t  node;          /* NUMA node of interest (-1 for all nodes). */
+    /* OUT variables. */
+    uint64_t avail_bytes;   /* Bytes available in the specified region. */
+};
+typedef struct xen_sysctl_availheap xen_sysctl_availheap_t;
+DEFINE_XEN_GUEST_HANDLE(xen_sysctl_availheap_t);
+
 struct xen_sysctl {
     uint32_t cmd;
     uint32_t interface_version; /* XEN_SYSCTL_INTERFACE_VERSION */
@@ -178,6 +190,7 @@ struct xen_sysctl {
         struct xen_sysctl_perfc_op          perfc_op;
         struct xen_sysctl_getdomaininfolist getdomaininfolist;
         struct xen_sysctl_debug_keys        debug_keys;
+        struct xen_sysctl_availheap         availheap;
         struct xen_sysctl_getcpuinfo        getcpuinfo;
         uint8_t                             pad[128];
     } u;
diff --git a/include/xen/mm.h b/include/xen/mm.h
index a94d4ef..a5183fd 100644
--- a/include/xen/mm.h
+++ b/include/xen/mm.h
@@ -61,6 +61,8 @@ struct page_info *__alloc_domheap_pages(
     struct domain *d, unsigned int cpu, unsigned int order, 
     unsigned int memflags);
 void free_domheap_pages(struct page_info *pg, unsigned int order);
+unsigned long avail_domheap_pages_region(
+    unsigned int node, unsigned int min_width, unsigned int max_width);
 unsigned long avail_domheap_pages(void);
 #define alloc_domheap_page(d) (alloc_domheap_pages(d,0,0))
 #define free_domheap_page(p)  (free_domheap_pages(p,0))