[Xen-devel] [PATCH 2 of 4] tools: Save superpages in the same batch, to make detection easier

[George Dunlap <george.dunlap@xxxxxxxxxxxxx>]

On the first time through (when pfns are mostly allocated on
the receiving side), try to keep superpages together in the same
batch by ending a batch early if we see the first page of a
potential superpage and there isn't enough room in the batch
for a full superpage.

Signed-off-by: George Dunlap <george.dunlap@xxxxxxxxxxxxx>

diff -r 57558d87761f -r a629b41a8d1f tools/libxc/xc_domain_save.c
--- a/tools/libxc/xc_domain_save.c      Mon May 16 11:50:46 2011 +0100
+++ b/tools/libxc/xc_domain_save.c      Mon May 16 11:50:46 2011 +0100
@@ -82,6 +82,11 @@ struct outbuf {
      ((mfn_to_pfn(_mfn) < (dinfo->p2m_size)) &&   \
       (pfn_to_mfn(mfn_to_pfn(_mfn)) == (_mfn))))
 
+#define SUPERPAGE_PFN_SHIFT  9
+#define SUPERPAGE_NR_PFNS    (1UL << SUPERPAGE_PFN_SHIFT)
+
+#define SUPER_PAGE_START(pfn)    (((pfn) & (SUPERPAGE_NR_PFNS-1)) == 0 )
+
 /*
 ** During (live) save/migrate, we maintain a number of bitmaps to track
 ** which pages we have to send, to fixup, and to skip.
@@ -906,6 +911,7 @@ int xc_domain_save(xc_interface *xch, in
     int rc = 1, frc, i, j, last_iter = 0, iter = 0;
     int live  = (flags & XCFLAGS_LIVE);
     int debug = (flags & XCFLAGS_DEBUG);
+    int superpages = !!hvm;
     int race = 0, sent_last_iter, skip_this_iter = 0;
     unsigned int sent_this_iter = 0;
     int tmem_saved = 0;
@@ -1262,6 +1268,12 @@ int xc_domain_save(xc_interface *xch, in
                            (test_bit(n, to_fix)  && last_iter)) )
                         continue;
 
+                    /* First time through, try to keep superpages in the same 
batch */
+                    if ( superpages && iter == 1
+                         && SUPER_PAGE_START(n)
+                         && batch + SUPERPAGE_NR_PFNS > MAX_BATCH_SIZE )
+                        break;
+
                     /*
                     ** we get here if:
                     **  1. page is marked to_send & hasn't already been 
re-dirtied

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