Skip to contentSkip to conversation

Fix auto-ballooning of dom0 for HVM domains

PATCH In review Charles Coffing 1 participant 2 messages May 18, 2006

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix auto-ballooning of dom0 for HVM domains

Hi,

I've been trying to make the auto-ballooning of domain 0 work reliably for HVM domains.

Patch #1 is a simple bug fix, in preparation for patch #2. Patch #2 tries to calculate how much memory is needed for HVM domains (XenSource bug #521) and is certainly open for discussion. Patches apply cleanly to both 3.0-testing and unstable.

Patch #1 (xen-hvm-auto-balloon.diff):
When a domain (whether para- or fully-virtualized) reports how much overhead memory it requires (via getDomainMemory in image.py), all such memory was immediately allocated to the domain itself. This is certainly incorrect for HVM domains, since additional increase_reservation calls are made later in qemu. Since all ballooned memory is already taken, qemu will fail. The fix is to treat the requested memory size and the overhead size as separate values. The requested memory size is immediately allocated to the new domain; the overhead is left unallocated for whatever else might need it later.

Patch #2 (xen-get-dom-memory.diff):
This patch calculates the overhead needed for HVM domains. If HVM is supported by the hardware, I add a little ballooning overhead to paravirtualized VMs also, to avoid low-memory situations. (There are various unchecked alloc_domheap_pages calls in shadow*.c that I am trying to avoid tripping over for now...) The values in this patch work fine on 32 bit; I may update them later based on feedback and/or testing on 64 bit.

Thanks,
Chuck

tools/python/xen/xend/XendDomainInfo.py +1 -1
Index: xen-3.0-testing/tools/python/xen/xend/XendDomainInfo.py
===================================================================
@@ -1247,7 +1247,7 @@ class XendDomainInfo:
             m = self.image.getDomainMemory(self.info['memory'] * 1024)
             balloon.free(m)
             xc.domain_setmaxmem(self.domid, m)
-            xc.domain_memory_increase_reservation(self.domid, m, 0, 0)
+            xc.domain_memory_increase_reservation(self.domid, self.info['memory'] * 1024, 0, 0)
 
             self.createChannels()
 
tools/python/xen/xend/image.py +16 -8
Index: xen-3.0-testing/tools/python/xen/xend/image.py
===================================================================
@@ -19,6 +19,7 @@
 
 import os, string
 import re
+import math
 
 import xen.lowlevel.xc
 from xen.xend import sxp
@@ -143,11 +144,13 @@ class ImageHandler:
                           % (self.ostype, self.vm.getDomid(), str(result)))
 
 
-    def getDomainMemory(self, mem):
+    def getDomainMemory(self, mem_kb):
         """@return The memory required, in KiB, by the domain to store the
-        given amount, also in KiB.  This is normally just mem, but HVM domains
-        have overheads to account for."""
-        return mem
+        given amount, also in KiB.  This is normally just mem, but if HVM is
+        supported, keep a little extra free."""
+        if 'hvm' in xc.xeninfo()['xen_caps']:
+            mem_kb += 4*1024;
+        return mem_kb
 
     def buildDomain(self):
         """Build the domain. Define in subclass."""
@@ -379,15 +382,20 @@ class HVMImageHandler(ImageHandler):
         os.waitpid(self.pid, 0)
         self.pid = 0
 
-    def getDomainMemory(self, mem):
+    def getDomainMemory(self, mem_kb):
         """@see ImageHandler.getDomainMemory"""
-        page_kb = 4
-        extra_pages = 0
         if os.uname()[4] == 'ia64':
             page_kb = 16
             # ROM size for guest firmware, ioreq page and xenstore page
             extra_pages = 1024 + 2
-        return mem + extra_pages * page_kb
+        else:
+            page_kb = 4
+            # This was derived emperically:
+            #   2.4 MB overhead per 1024 MB RAM + 8 MB constant
+            #   + 4 to avoid low-memory condition
+            extra_mb = (2.4/1024) * (mem_kb/1024.0) + 12;
+            extra_pages = int( math.ceil( extra_mb*1024 / page_kb ))
+        return mem_kb + extra_pages * page_kb
 
     def register_shutdown_watch(self):
         """ add xen store watch on control/shutdown """

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

#1 lore raw

Sorry, I forgot the Signed-off-by line. Consider this my sign-off on the previous email:

Quoted 1 line
Quoted 1 line
Quoted 1 line

On Thu, May 18, 2006 at 11:04 AM, in message

<[email protected]>,

Quoted 3 lines · Charles Coffing

Hi,

I've been trying to make the auto- ballooning of domain 0 work

reliably

Quoted 4 lines

for HVM domains.

Patch #1 is a simple bug fix, in preparation for patch #2. Patch #2 tries to calculate how much memory is needed for HVM domains

(XenSource

Quoted 1 line

bug #521) and is certainly open for discussion. Patches apply

cleanly

Quoted 4 lines

to both 3.0- testing and unstable.

Patch #1 (xen- hvm- auto- balloon.diff):
When a domain (whether para- or fully- virtualized) reports how

much

Quoted 1 line

overhead memory it requires (via getDomainMemory in image.py), all

such

Quoted 3 lines

memory was immediately allocated to the domain itself. This is certainly incorrect for HVM domains, since additional increase_reservation calls are made later in qemu. Since all

ballooned

Quoted 3 lines

memory is already taken, qemu will fail. The fix is to treat the requested memory size and the overhead size as separate values. The requested memory size is immediately allocated to the new domain;

the

Quoted 4 lines

overhead is left unallocated for whatever else might need it later.

Patch #2 (xen- get- dom- memory.diff):
This patch calculates the overhead needed for HVM domains. If HVM

is

Quoted 2 lines

supported by the hardware, I add a little ballooning overhead to paravirtualized VMs also, to avoid low- memory situations. (There

are

Quoted 2 lines

various unchecked alloc_domheap_pages calls in shadow*.c that I am trying to avoid tripping over for now...) The values in this patch

work

Quoted 1 line

fine on 32 bit; I may update them later based on feedback and/or

testing

Quoted 4 lines

on 64 bit.

Thanks,
Chuck