[Qemu-devel] [PATCH 4/5] Add qemu_mbind interface for pinning memory to host node

[Wanlong Gao <gaowanlong@cn.fujitsu.com>]

Add qemu_mbind() interface for pinning memory to host node
manually. Use the mbind() syscall wrapper which defined
in libnuma.

Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
---
 configure            | 18 ++++++++++++++++++
 include/qemu/osdep.h | 26 ++++++++++++++++++++++++++
 util/osdep.c         | 15 +++++++++++++++
 3 files changed, 59 insertions(+)

diff --git a/configure b/configure
index 5ae7e4a..5364d01 100755
--- a/configure
+++ b/configure
@@ -3141,6 +3141,20 @@ if compile_prog "" "" ; then
 fi
 
 ##########################################
+# check if we have mbind
+
+mbind=no
+cat > $TMPC << EOF
+#include <numaif.h>
+int main(void) { return mbind(0, 0, MPOL_BIND, 0, 0, 0); }
+EOF
+if compile_prog "" "-lnuma"; then
+    mbind=yes
+    LIBS="-lnuma $LIBS"
+    libs_qga="-lnuma $libs_qga"
+fi
+
+##########################################
 # check if we have usable SIGEV_THREAD_ID
 
 sigev_thread_id=no
@@ -3560,6 +3574,7 @@ echo "preadv support    $preadv"
 echo "fdatasync         $fdatasync"
 echo "madvise           $madvise"
 echo "posix_madvise     $posix_madvise"
+echo "mbind             $mbind"
 echo "sigev_thread_id   $sigev_thread_id"
 echo "uuid support      $uuid"
 echo "libcap-ng support $cap_ng"
@@ -3875,6 +3890,9 @@ fi
 if test "$posix_madvise" = "yes" ; then
   echo "CONFIG_POSIX_MADVISE=y" >> $config_host_mak
 fi
+if test "$mbind" = "yes"; then
+  echo "CONFIG_MBIND=y" >> $config_host_mak
+fi
 if test "$sigev_thread_id" = "yes" ; then
   echo "CONFIG_SIGEV_THREAD_ID=y" >> $config_host_mak
 fi
diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h
index 57d7b1f..82a790e 100644
--- a/include/qemu/osdep.h
+++ b/include/qemu/osdep.h
@@ -152,6 +152,32 @@ int qemu_madvise(void *addr, size_t len, int advice);
 int qemu_open(const char *name, int flags, ...);
 int qemu_close(int fd);
 
+#define QEMU_MPOL_INVALID -1
+
+#if defined(CONFIG_MBIND)
+#include <numaif.h>
+/* Policies */
+#define QEMU_MPOL_DEFAULT MPOL_DEFAULT
+#define QEMU_MPOL_PREFERRED MPOL_PREFERRED
+#define QEMU_MPOL_BIND MPOL_BIND
+#define QEMU_MPOL_INTERLEAVE MPOL_INTERLEAVE
+/* Flags for qemu_mbind */
+#define QEMU_MPOL_MF_STRICT MPOL_MF_STRICT
+#define QEMU_MPOL_MF_MOVE MPOL_MF_MOVE
+#define QEMU_MPOL_MF_MOVE_ALL MPOL_MF_MOVE_ALL
+#else
+#define QEMU_MPOL_DEFAULT QEMU_MPOL_INVALID
+#define QEMU_MPOL_PREFERRED QEMU_MPOL_INVALID
+#define QEMU_MPOL_BIND QEMU_MPOL_INVALID
+#define QEMU_MPOL_INTERLEAVE QEMU_MPOL_INVALID
+#define QEMU_MPOL_MF_STRICT QEMU_MPOL_INVALID
+#define QEMU_MPOL_MF_MOVE QEMU_MPOL_INVALID
+#define QEMU_MPOL_MF_MOVE_ALL QEMU_MPOL_INVALID
+#endif
+int qemu_mbind(void *addr, unsigned long len, int mode,
+               unsigned long *nodemask, unsigned long maxnode,
+               unsigned flags);
+
 #if defined(__HAIKU__) && defined(__i386__)
 #define FMT_pid "%ld"
 #elif defined(WIN64)
diff --git a/util/osdep.c b/util/osdep.c
index 685c8ae..70f33c7 100644
--- a/util/osdep.c
+++ b/util/osdep.c
@@ -37,6 +37,10 @@
 #include <sys/mman.h>
 #endif
 
+#if defined(CONFIG_MBIND)
+#include <numaif.h>
+#endif
+
 #ifdef CONFIG_SOLARIS
 #include <sys/types.h>
 #include <sys/statvfs.h>
@@ -472,3 +476,14 @@ writev(int fd, const struct iovec *iov, int iov_cnt)
     return readv_writev(fd, iov, iov_cnt, true);
 }
 #endif
+
+int qemu_mbind(void *addr, unsigned long len, int mode,
+               unsigned long *nodemask, unsigned long maxnode,
+               unsigned flags)
+{
+#if defined(CONFIG_MBIND)
+    return mbind(addr, len, mode, nodemask, maxnode, flags);
+#else
+    return 0;
+#endif
+}
-- 
1.8.3.rc2.10.g0c2b1cf