[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] [PATCH 6/6] xen, tools: Add new hypercalls/tools to examine heap



Introduce two dom0_ops and a userspace tool used to probe Xen's heap
availability and examine which nodes a domain's memory resides within.

availheap takes a zone and a node value and returns the current number
of free pages in that area of the heap.  Using -1 as wildcards, one can
probe total free pages in a zone, or node.

getnodestat examines the page ownership of a domain.  By walking the
domain's page list we can determine from which nodes a domain's memory
hails.

(bebop) tmp # xen_numastat -H -d -1
HEAPINFO: TOTAL_FREE_PAGES: 578057
HEAPINFO: NR_ZONES: 3
HEAPINFO: NR_NODES: 2
HEAPINFO: ZONE_XEN: NODE0: FREE_PAGES: 3507
HEAPINFO: ZONE_XEN: NODE1: FREE_PAGES: 0
HEAPINFO: ZONE_DOM: NODE0: FREE_PAGES: 0
HEAPINFO: ZONE_DOM: NODE1: FREE_PAGES: 107094
HEAPINFO: ZONE_DMA: NODE0: FREE_PAGES: 467456
HEAPINFO: ZONE_DMA: NODE1: FREE_PAGES: 0
DOM0: NODE0: PAGES: 43773
DOM0: NODE1: PAGES: 21768
DOM8: NODE0: PAGES: 5
DOM8: NODE1: PAGES: 262144

-- 
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
(512) 838-9253   T/L: 678-9253
ryanh@xxxxxxxxxx


diffstat output:
 b/tools/xen_numastat/Makefile       |   35 +++++
 b/tools/xen_numastat/xen_numastat.1 |   22 +++
 b/tools/xen_numastat/xen_numastat.c |  215 ++++++++++++++++++++++++++++++++++++
 tools/Makefile                      |    1 
 tools/libxc/xc_domain.c             |   45 +++++++
 tools/libxc/xenctrl.h               |   15 ++
 xen/common/dom0_ops.c               |   72 ++++++++++++
 xen/common/page_alloc.c             |    5 
 xen/include/public/dom0_ops.h       |   22 +++
 xen/include/xen/mm.h                |    5 
 10 files changed, 432 insertions(+), 5 deletions(-)

Signed-off-by: Ryan Harper <ryanh@xxxxxxxxxx>
---
diff -r 4d529088b115 tools/Makefile
--- a/tools/Makefile    Mon Jul 31 10:52:38 2006 -0500
+++ b/tools/Makefile    Mon Jul 31 10:53:35 2006 -0500
@@ -13,6 +13,7 @@ SUBDIRS-y += console
 SUBDIRS-y += console
 SUBDIRS-y += xenmon
 SUBDIRS-y += guest-headers
+SUBDIRS-y += xen_numastat
 SUBDIRS-$(VTPM_TOOLS) += vtpm_manager
 SUBDIRS-$(VTPM_TOOLS) += vtpm
 SUBDIRS-y += xenstat
diff -r 4d529088b115 tools/libxc/xc_domain.c
--- a/tools/libxc/xc_domain.c   Mon Jul 31 10:52:38 2006 -0500
+++ b/tools/libxc/xc_domain.c   Mon Jul 31 10:53:35 2006 -0500
@@ -524,6 +524,51 @@ int xc_domain_iomem_permission(int xc_ha
     op.u.iomem_permission.allow_access = allow_access;
 
     return do_dom0_op(xc_handle, &op);
+}
+
+int xc_availheap(int xc_handle,
+                 int zone,
+                 int node,
+                 uint32_t *nr_zones,
+                 uint32_t *nr_nodes,
+                 uint64_t *pages)
+{
+    DECLARE_DOM0_OP;
+    int rc = 0;
+
+    op.cmd = DOM0_AVAILHEAP;
+    op.u.availheap.zone = zone;
+    op.u.availheap.node = node;
+
+    rc = do_dom0_op(xc_handle, &op);
+    if ( rc >= 0 ) {
+        if (nr_zones)
+            *nr_zones = op.u.availheap.nr_zones;
+        if (nr_nodes)
+            *nr_nodes = op.u.availheap.nr_nodes;
+        *pages = op.u.availheap.pages;
+    }
+
+    return rc;
+}
+
+int xc_domain_getnodestat(int xc_handle,
+                          uint32_t domid,
+                          uint32_t node,
+                          uint64_t *pages)
+{
+    DECLARE_DOM0_OP;
+    int rc = 0;
+
+    op.cmd = DOM0_GETDOMNODESTAT;
+    op.u.getdomnodestat.domain = (domid_t) domid;
+    op.u.getdomnodestat.node   = node;
+
+    rc = do_dom0_op(xc_handle, &op);
+    if ( rc >= 0 )
+        *pages = op.u.getdomnodestat.pages;
+
+    return rc;
 }
 
 /*
diff -r 4d529088b115 tools/libxc/xenctrl.h
--- a/tools/libxc/xenctrl.h     Mon Jul 31 10:52:38 2006 -0500
+++ b/tools/libxc/xenctrl.h     Mon Jul 31 10:53:35 2006 -0500
@@ -541,6 +541,21 @@ int xc_get_pfn_type_batch(int xc_handle,
 /* Get current total pages allocated to a domain. */
 long xc_get_tot_pages(int xc_handle, uint32_t domid);
 
+/**
+ * This function retrieves the the number of pages in the
+ * specified domain that are on the specified node.
+ *
+ * @parm xc_handle a handle to an open hypervisor interface
+ * @parm domid the domain to query, -1 for per-node free list
+ * @parm node the node to query
+ * @parm *pages caller variable to put output
+ * @return 0 on success, <0 on failure.
+ */
+int xc_domain_getnodestat(int xc_handle, uint32_t domid, uint32_t node,
+                          uint64_t *pages);
+
+int xc_availheap(int xc_handle, int zone, int node,
+                 uint32_t *nr_zones, uint32_t *nr_nodes, uint64_t *pages);
 
 /*
  * Trace Buffer Operations
diff -r 4d529088b115 xen/common/dom0_ops.c
--- a/xen/common/dom0_ops.c     Mon Jul 31 10:52:38 2006 -0500
+++ b/xen/common/dom0_ops.c     Mon Jul 31 10:53:35 2006 -0500
@@ -22,6 +22,7 @@
 #include <public/dom0_ops.h>
 #include <public/sched_ctl.h>
 #include <acm/acm_hooks.h>
+#include <asm/numa.h>
 
 extern long arch_do_dom0_op(
     struct dom0_op *op, XEN_GUEST_HANDLE(dom0_op_t) u_dom0_op);
@@ -712,6 +713,77 @@ long do_dom0_op(XEN_GUEST_HANDLE(dom0_op
     }
     break;
 
+    case DOM0_AVAILHEAP:
+    {
+        ret = -EINVAL;
+        if ( op->u.availheap.node >= num_online_nodes() )
+            break;
+        if ( op->u.availheap.zone >= NR_ZONES )
+            break;
+
+        /* indicate the number of zones/nodes queried.
+         * NB: -1 is wild card for all zones/nodes */
+        ( op->u.availheap.zone < 0 ) ?
+            (op->u.availheap.nr_zones=NR_ZONES) :
+            (op->u.availheap.nr_zones=1);
+
+        ( op->u.availheap.node < 0 ) ?
+            (op->u.availheap.nr_nodes=num_online_nodes()) :
+            (op->u.availheap.nr_nodes=1);
+
+        op->u.availheap.pages =
+            avail_heap_pages(op->u.availheap.zone, op->u.availheap.node);
+
+        if ( copy_to_guest(u_dom0_op, op, 1) )
+            ret = -EFAULT;
+        else
+            ret = 0;
+    }
+    break;
+
+    case DOM0_GETDOMNODESTAT:
+    {
+        struct domain *d;
+        struct page_info *pg;
+
+        ret = -EINVAL;
+        if ( op->u.getdomnodestat.node >= num_online_nodes() )
+            break;
+
+        ret = -ESRCH;
+        d = find_domain_by_id(op->u.getdomnodestat.domain);
+
+        if ( d != NULL )
+        {
+            /* clear out pages count */
+            op->u.getdomnodestat.pages = 0;
+
+            /* walk domain's page list and count pages on node */
+            spin_lock(&d->page_alloc_lock);
+            list_for_each_entry(pg, &d->page_list, list)
+            {
+                if ( phys_to_nid(page_to_maddr(pg)) ==
+                     (int) op->u.getdomnodestat.node )
+                    op->u.getdomnodestat.pages++;
+            }
+            list_for_each_entry(pg, &d->xenpage_list, list)
+            {
+                if ( phys_to_nid(page_to_maddr(pg)) ==
+                     (int) op->u.getdomnodestat.node )
+                    op->u.getdomnodestat.pages++;
+            }
+            spin_unlock(&d->page_alloc_lock);
+
+            put_domain(d);
+
+            if ( copy_to_guest(u_dom0_op, op, 1) )
+                ret = -EFAULT;
+            else
+                ret = 0;
+        }
+    }
+    break;
+
     default:
         ret = arch_do_dom0_op(op, u_dom0_op);
         break;
diff -r 4d529088b115 xen/common/page_alloc.c
--- a/xen/common/page_alloc.c   Mon Jul 31 10:52:38 2006 -0500
+++ b/xen/common/page_alloc.c   Mon Jul 31 10:53:35 2006 -0500
@@ -241,11 +241,6 @@ unsigned long alloc_boot_pages(unsigned 
  * BINARY BUDDY ALLOCATOR
  */
 
-#define MEMZONE_XEN 0
-#define MEMZONE_DOM 1
-#define MEMZONE_DMADOM 2
-#define NR_ZONES    3
-
 #define pfn_dom_zone_type(_pfn)                                 \
     (((_pfn) <= MAX_DMADOM_PFN) ? MEMZONE_DMADOM : MEMZONE_DOM)
 
diff -r 4d529088b115 xen/include/public/dom0_ops.h
--- a/xen/include/public/dom0_ops.h     Mon Jul 31 10:52:38 2006 -0500
+++ b/xen/include/public/dom0_ops.h     Mon Jul 31 10:53:35 2006 -0500
@@ -543,6 +543,26 @@ struct dom0_settimeoffset {
 };
 typedef struct dom0_settimeoffset dom0_settimeoffset_t;
 DEFINE_XEN_GUEST_HANDLE(dom0_settimeoffset_t);
+
+#define DOM0_GETDOMNODESTAT   51
+typedef struct dom0_getdomnodestat {
+    int domain;           /* domain to query, -1 to query per-node free list */
+    int32_t node;         /* node to query */
+    uint64_t pages;       /* number of pages on node */
+} dom0_getdomnodestat_t;
+DEFINE_XEN_GUEST_HANDLE(dom0_getdomnodestat_t);
+
+#define DOM0_AVAILHEAP        52
+typedef struct dom0_availheap {
+    /* in  */
+    int zone;             /* query available pages in zone, -1 for all */
+    int node;             /* query available pages in node, -1 for all */
+    /* out */
+    uint32_t nr_zones;    /* number of zones queried */
+    uint32_t nr_nodes;    /* number of nodes queried */
+    uint64_t pages;
+} dom0_availheap_t;
+DEFINE_XEN_GUEST_HANDLE(dom0_availheap_t);
 
 struct dom0_op {
     uint32_t cmd;
@@ -587,6 +607,8 @@ struct dom0_op {
         struct dom0_hypercall_init    hypercall_init;
         struct dom0_domain_setup      domain_setup;
         struct dom0_settimeoffset     settimeoffset;
+        struct dom0_getdomnodestat    getdomnodestat;
+        struct dom0_availheap         availheap;
         uint8_t                       pad[128];
     } u;
 };
diff -r 4d529088b115 xen/include/xen/mm.h
--- a/xen/include/xen/mm.h      Mon Jul 31 10:52:38 2006 -0500
+++ b/xen/include/xen/mm.h      Mon Jul 31 10:53:35 2006 -0500
@@ -32,6 +32,11 @@
 #include <xen/types.h>
 #include <xen/list.h>
 #include <xen/spinlock.h>
+
+#define MEMZONE_XEN 0
+#define MEMZONE_DOM 1
+#define MEMZONE_DMADOM 2
+#define NR_ZONES    3
 
 struct domain;
 struct page_info;
diff -r 4d529088b115 tools/xen_numastat/Makefile
--- /dev/null   Thu Jan 01 00:00:00 1970 +0000
+++ b/tools/xen_numastat/Makefile       Mon Jul 31 10:53:35 2006 -0500
@@ -0,0 +1,35 @@
+INSTALL                = install
+INSTALL_PROG   = $(INSTALL) -m0755
+INSTALL_DIR    = $(INSTALL) -d -m0755
+INSTALL_DATA   = $(INSTALL) -m0644
+
+XEN_ROOT=../..
+include $(XEN_ROOT)/tools/Rules.mk
+
+CFLAGS  += -Wall -Werror -g
+
+CFLAGS  += -I $(XEN_XC)
+CFLAGS  += -I $(XEN_LIBXC)
+
+HDRS     = $(wildcard *.h)
+OBJS     = $(patsubst %.c,%.o,$(wildcard *.c))
+
+BIN      = xen_numastat
+MAN1     = $(wildcard *.1)
+LIBBIN   = 
+
+all: build
+build: $(BIN)
+
+install: build
+       [ -d $(DESTDIR)/usr/bin ] || $(INSTALL_DIR) $(DESTDIR)/usr/bin
+       $(INSTALL_PROG) $(BIN) $(DESTDIR)/usr/bin
+       [ -d $(DESTDIR)/usr/share/man/man1 ] || \
+               $(INSTALL_DIR) $(DESTDIR)/usr/share/man/man1
+       $(INSTALL_DATA) $(MAN1) $(DESTDIR)/usr/share/man/man1
+
+clean:
+       $(RM) *.a *.so *.o $(BIN)
+
+%: %.c $(HDRS) Makefile
+       $(CC) $(CFLAGS) -o $@ $< -L$(XEN_LIBXC) -lxenctrl
diff -r 4d529088b115 tools/xen_numastat/xen_numastat.1
--- /dev/null   Thu Jan 01 00:00:00 1970 +0000
+++ b/tools/xen_numastat/xen_numastat.1 Mon Jul 31 10:53:35 2006 -0500
@@ -0,0 +1,22 @@
+.\" DO NOT MODIFY THIS FILE!  It was generated by help2man 1.35.
+.TH XEN_NUMASTAT "1" "May 2006" "Usage: xen_numastat [OPTION]" "User Commands"
+.SH NAME
+xen_numastat - Displays NUMA statistics about machine and domains
+.SH SYNOPSIS
+.B xen_numastat
+[\fIOPTIONS\fR]
+.SH DESCRIPTION
+Displays NUMA statistics about machine and domains
+.TP
+\fB\-h\fR, \fB\-\-help\fR
+display this help and exit
+.TP
+\fB\-H\fR, \fB\-\-heapinfo\fR
+display info about Xen HEAP
+.TP
+\fB\-d\fR, \fB\-\-domid\fR <D>
+display NUMA info about domain D
+.TP
+\fB\-p\fR, \fB\-\-physinfo\fR
+display NUMA info about this machine
+.PP
diff -r 4d529088b115 tools/xen_numastat/xen_numastat.c
--- /dev/null   Thu Jan 01 00:00:00 1970 +0000
+++ b/tools/xen_numastat/xen_numastat.c Mon Jul 31 10:53:35 2006 -0500
@@ -0,0 +1,215 @@
+/*\
+ *  Copyright (C) International Business Machines  Corp., 2006
+ *  Author(s): Ryan Harper <ryanh@xxxxxxxxxx>
+ *
+ *  Xen Domain NUMA statistics tool
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; under version 2 of the License.
+ * 
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.  * 
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+\*/
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <err.h>
+#include <errno.h>
+#include <xenctrl.h>
+#include <getopt.h>
+#include <inttypes.h>
+
+#define DOMAIN 0
+#define PHYSICAL 1
+/* this is a bit arbitrary, might be nice to query from xen */
+#define MAX_DOMAINS 256
+
+/* taken from xen/include/xen/mm.h */
+#define MEMZONE_XEN 0
+#define MEMZONE_DOM 1
+#define MEMZONE_DMADOM 2
+
+static void usage(const char *program) {
+       printf("Usage: %s [OPTION]\n"
+              "Displays NUMA statistics about machine and domains \n"
+              "\n"
+              "  -h, --help       display this help and exit\n"
+              "  -H, --heapinfo   display info about Xen HEAP\n"
+              "  -d, --domid <D>  display NUMA info about domain D\n"
+              "  -p, --physinfo   display NUMA info about this machine\n"
+              , program);
+}
+
+int str_to_domid(const char* domidstr) {
+   char *end = NULL;
+   int domid;
+   
+   domid = strtol(domidstr, &end, 10);
+   if (end && *end) {
+      fprintf(stderr, "Invalid DOMID `%s'\n", domidstr);
+      fprintf(stderr, "*end -> %c\n", *end);
+      exit(EINVAL);
+   }
+
+   return domid;
+}
+
+void display_dominfo(int xc_handle, int domid, int nr_nodes)
+{
+   xc_domaininfo_t *dominfo =NULL;
+   uint64_t pages;
+   unsigned int i, j, num_domains = 1;
+
+   /* malloc a large buffer for info */
+   if ( !(dominfo = malloc(MAX_DOMAINS * sizeof(xc_domaininfo_t))) )
+       err(errno, "failed to malloc memory");
+       
+   /* get actual number of domains */
+   num_domains = xc_domain_getinfolist(xc_handle, 0, MAX_DOMAINS, dominfo);
+
+   for (i=0; i<num_domains; i++) {
+      /* show all domains, or only requested domain */
+      if (domid == -1 || domid == dominfo[i].domain) {
+          /* print domain per-node info */
+         for (j=0; j < nr_nodes; j++) {
+            if (xc_domain_getnodestat(xc_handle, 
+                                      dominfo[i].domain, j, &pages) < 0)
+               err(errno, "xc_domain_getnodestat()");
+            printf("DOM%d: NODE%d: PAGES: %"PRIu64"\n", 
+                   dominfo[i].domain, j, pages);
+         }
+      }
+   }
+
+   free(dominfo);
+}
+
+void display_heapinfo(int xc_handle)
+{
+   int i,j;
+   uint32_t nr_zones, nr_nodes;
+   uint64_t pages;
+   char zone[8];
+
+   /* get zones/nodes and total free pages */
+   if( xc_availheap(xc_handle, -1, -1, &nr_zones, &nr_nodes, &pages) )
+      err(errno, "xc_availheap()");
+
+   printf("HEAPINFO: TOTAL_FREE_PAGES: %"PRIu64"\n", pages);
+   printf("HEAPINFO: NR_ZONES: %d\n", nr_zones);
+   printf("HEAPINFO: NR_NODES: %d\n", nr_nodes);
+
+   for (i=0; i<nr_zones; i++) {
+      switch(i) {
+          case MEMZONE_XEN:
+              sprintf(zone, "XEN");
+              break;
+          case MEMZONE_DOM:
+              sprintf(zone, "DOM");
+              break;
+          case MEMZONE_DMADOM:
+              sprintf(zone, "DMA");
+              break;
+      }
+      for (j=0; j<nr_nodes; j++) {
+         if( xc_availheap(xc_handle, i, j, NULL, NULL, &pages) )
+            err(errno, "xc_availheap()");
+         printf("HEAPINFO: ZONE_%s: NODE%d: FREE_PAGES: %"PRIu64"\n", 
+                zone, j, pages);
+      }
+   }
+}
+
+void display_physinfo(int xc_handle, xc_physinfo_t *info) 
+{
+
+   printf("PHYSINFO: NR_NODES: %d\n", info->nr_nodes);
+
+}
+
+int main(int argc, char * argv[])
+{
+    
+       char *sopt = "d:Hhp";
+       int ch;
+       int opt_ind=0, heapinfo=0, dominfo=0, physinfo=0, domid=-1;
+   int xc_handle, rc;
+   xc_physinfo_t info;
+   xc_memory_chunk_t *chunks;
+   xc_node_to_cpu_t *map;
+       struct option lopt[] = {
+      { "domid",    1, 0, 'd' },
+      { "heapinfo", 0, 0, 'H' },
+               { "help",     0, 0, 'h' },
+      { "physinfo", 0, 0, 'p' },
+               { 0, 0, 0, 0 }
+       };
+
+       while((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
+               switch(ch) {
+      case 'd':
+         /* convert domid */
+         domid = str_to_domid(optarg);
+         dominfo = 1;
+         break;
+      case 'H':
+         heapinfo = 1;
+         break;
+               case 'h':
+                       usage(argv[0]);
+                       exit(0);
+                       break;
+      case 'p':
+         physinfo = 1;
+         break;
+      }
+       }
+
+   /* display help if no options are passed */
+   if (dominfo == 0 && physinfo == 0 && heapinfo == 0) {
+       usage(argv[0]);
+       exit(0);
+   }
+   
+   /* set up */
+   xc_handle = xc_interface_open();
+   if (xc_handle == -1)
+      err(errno, "xc_interface_open()");
+
+   /* make space for mem chunks */
+   chunks = (xc_memory_chunk_t *)
+      malloc( sizeof(xc_memory_chunk_t) * PUBLIC_MAXCHUNKS );
+   set_xen_guest_handle(info.memory_chunks, chunks);
+
+   /* make space for node_to_cpu mapping */
+   map = (xc_node_to_cpu_t *)
+      malloc( sizeof(xc_node_to_cpu_t) * PUBLIC_MAX_NUMNODES );
+   set_xen_guest_handle(info.node_to_cpu, map);
+
+   /* get the machine physinfo */
+   rc = xc_physinfo(xc_handle, &info);
+   if (rc < 0)
+      err(errno, "xc_physinfo()");
+
+   if (physinfo > 0)
+      display_physinfo(xc_handle, &info);
+   if (heapinfo > 0)
+      display_heapinfo(xc_handle);
+   if (dominfo > 0)
+      display_dominfo(xc_handle, domid, info.nr_nodes);
+
+   /* free malloc'd memory */
+   free(chunks);
+   free(map);
+
+   xc_interface_close(xc_handle);
+
+   return 0;
+}

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.