[PATCH 0 of 3] Enable domain checkpointing via xm save --checkpoint

[PATCH 0 of 3] Enable domain checkpointing via xm save --checkpoint

[Brendan Cully]

The attached patches create a --checkpoint flag for xm save:

xc_domain_resume lets the caller alert the guest that it is being
resumed. Callers should not do this unless they know the guest
supports the operation.

When xend wants to resume a domain, it checks for the suspend_cancel
feature in the elfnotes it has recorded from domain build or
resume. If the flag is present, it simply lets the guest handle
resuming, otherwise it tears down and rebuilds the domain and lets the
guest act as if it is resuming in a new domain.

A new suspend_cancel feature is added to the kernel 'features'
elfnote.

[PATCH 1 of 3] Let xend choose to do hard or soft domain resumption depending on

[Brendan Cully]

# HG changeset patch
# User Brendan Cully <brendan@cs.ubc.ca>
# Date 1172543764 28800
# Node ID b2edb7f3393255f7b51af7fe688de687233b525e
# Parent  e7b2a282c9e70fbadd38ccc7cc7d1fcb6a32b886
Let xend choose to do hard or soft domain resumption depending on
whether the domain advertises support for soft resumption in its
elfnotes.

Signed-off-by: Brendan Cully <brendan@kublai.com>

diff -r e7b2a282c9e7 -r b2edb7f33932 tools/libxc/xc_resume.c
--- a/tools/libxc/xc_resume.c	Mon Feb 26 17:20:36 2007 +0000
+++ b/tools/libxc/xc_resume.c	Mon Feb 26 18:36:04 2007 -0800
@@ -169,13 +169,9 @@ static int xc_domain_resume_any(int xc_h
  * (2) should be used only for guests which cannot handle the special
  * new return code. (1) is always safe (but slower).
  */
-int xc_domain_resume(int xc_handle, uint32_t domid)
+int xc_domain_resume(int xc_handle, uint32_t domid, int fast)
 {
-    /*
-     * XXX: Implement a way to select between options (1) and (2).
-     * Or expose the options as two different methods to Python.
-     */
-    return (0
+    return (fast
             ? xc_domain_resume_cooperative(xc_handle, domid)
             : xc_domain_resume_any(xc_handle, domid));
 }
diff -r e7b2a282c9e7 -r b2edb7f33932 tools/libxc/xenctrl.h
--- a/tools/libxc/xenctrl.h	Mon Feb 26 17:20:36 2007 +0000
+++ b/tools/libxc/xenctrl.h	Mon Feb 26 18:36:04 2007 -0800
@@ -243,10 +243,12 @@ int xc_domain_destroy(int xc_handle,
  *
  * @parm xc_handle a handle to an open hypervisor interface
  * @parm domid the domain id to resume
+ * @parm fast use cooperative resume (guest must support this)
  * return 0 on success, -1 on failure
  */
 int xc_domain_resume(int xc_handle,
-                      uint32_t domid);
+		     uint32_t domid,
+		     int fast);
 
 /**
  * This function will shutdown a domain. This is intended for use in
diff -r e7b2a282c9e7 -r b2edb7f33932 tools/python/xen/lowlevel/xc/xc.c
--- a/tools/python/xen/lowlevel/xc/xc.c	Mon Feb 26 17:20:36 2007 +0000
+++ b/tools/python/xen/lowlevel/xc/xc.c	Mon Feb 26 18:36:04 2007 -0800
@@ -178,7 +178,17 @@ static PyObject *pyxc_domain_shutdown(Xc
 
 static PyObject *pyxc_domain_resume(XcObject *self, PyObject *args)
 {
-    return dom_op(self, args, xc_domain_resume);
+    uint32_t dom;
+    int fast;
+
+    if (!PyArg_ParseTuple(args, "ii", &dom, &fast))
+        return NULL;
+
+    if (xc_domain_resume(self->xc_handle, dom, fast) != 0)
+        return pyxc_error_to_exception();
+
+    Py_INCREF(zero);
+    return zero;
 }
 
 static PyObject *pyxc_vcpu_setaffinity(XcObject *self,
@@ -1124,7 +1134,8 @@ static PyMethodDef pyxc_methods[] = {
       (PyCFunction)pyxc_domain_resume,
       METH_VARARGS, "\n"
       "Resume execution of a suspended domain.\n"
-      " dom [int]: Identifier of domain to be resumed.\n\n"
+      " dom [int]: Identifier of domain to be resumed.\n"
+      " fast [int]: Use cooperative resume.\n\n"
       "Returns: [int] 0 on success; -1 on error.\n" },
 
     { "domain_shutdown", 
diff -r e7b2a282c9e7 -r b2edb7f33932 tools/python/xen/xend/XendCheckpoint.py
--- a/tools/python/xen/xend/XendCheckpoint.py	Mon Feb 26 17:20:36 2007 +0000
+++ b/tools/python/xen/xend/XendCheckpoint.py	Mon Feb 26 18:36:04 2007 -0800
@@ -137,23 +137,6 @@ def save(fd, dominfo, network, live, dst
         log.exception("Save failed on domain %s (%s).", domain_name,
                       dominfo.getDomid())
 
-        dominfo._releaseDevices()
-        dominfo.testDeviceComplete()
-        dominfo.testvifsComplete()
-        log.debug("XendCheckpoint.save: devices released")
-
-        dominfo._resetChannels()
-
-        dominfo._removeDom('control/shutdown')
-        dominfo._removeDom('device-misc/vif/nextDeviceID')
-
-        dominfo._createChannels()
-        dominfo._introduceDomain()
-        dominfo._storeDomDetails()
-
-        dominfo._createDevices()
-        log.debug("XendCheckpoint.save: devices created")
-
         dominfo.resumeDomain()
         log.debug("XendCheckpoint.save: resumeDomain")
 
diff -r e7b2a282c9e7 -r b2edb7f33932 tools/python/xen/xend/XendDomainInfo.py
--- a/tools/python/xen/xend/XendDomainInfo.py	Mon Feb 26 17:20:36 2007 +0000
+++ b/tools/python/xen/xend/XendDomainInfo.py	Mon Feb 26 18:36:04 2007 -0800
@@ -1662,10 +1662,31 @@ class XendDomainInfo:
     def resumeDomain(self):
         log.debug("XendDomainInfo.resumeDomain(%s)", str(self.domid))
 
+        if self.domid is None:
+            return
         try:
-            if self.domid is not None:
-                xc.domain_resume(self.domid)
-                ResumeDomain(self.domid)
+            # could also fetch a parsed key from xenstore
+            fast = 'suspend_cancel' in self.info.get_notes().get('FEATURES', '').split('|')
+            if not fast:
+                self._releaseDevices()
+                self.testDeviceComplete()
+                self.testvifsComplete()
+                log.debug("XendDomainInfo.resumeDomain: devices released")
+
+                self._resetChannels()
+
+                self._removeDom('control/shutdown')
+                self._removeDom('device-misc/vif/nextDeviceID')
+
+                self._createChannels()
+                self._introduceDomain()
+                self._storeDomDetails()
+
+                self._createDevices()
+                log.debug("XendDomainInfo.resumeDomain: devices created")
+
+            xc.domain_resume(self.domid, fast)
+            ResumeDomain(self.domid)
         except:
             log.exception("XendDomainInfo.resume: xc.domain_resume failed on domain %s." % (str(self.domid)))

[PATCH 2 of 3] Add xm save -c/--checkpoint option

[Brendan Cully]

# HG changeset patch
# User Brendan Cully <brendan@cs.ubc.ca>
# Date 1172543764 28800
# Node ID b207ed258dcf3a6aee865ca82bc170f68e589342
# Parent  b2edb7f3393255f7b51af7fe688de687233b525e
Add xm save -c/--checkpoint option

xm save --checkpoint leaves the domain running after creating the
snapshot.

Signed-off-by: Brendan Cully <brendan@cs.ubc.ca>

diff -r b2edb7f33932 -r b207ed258dcf tools/python/xen/xend/XendCheckpoint.py
--- a/tools/python/xen/xend/XendCheckpoint.py	Mon Feb 26 18:36:04 2007 -0800
+++ b/tools/python/xen/xend/XendCheckpoint.py	Mon Feb 26 18:36:04 2007 -0800
@@ -54,7 +54,7 @@ def read_exact(fd, size, errmsg):
     return buf
 
 
-def save(fd, dominfo, network, live, dst):
+def save(fd, dominfo, network, live, dst, checkpoint=False):
     write_exact(fd, SIGNATURE, "could not write guest state file: signature")
 
     config = sxp.to_string(dominfo.sxpr())
@@ -121,9 +121,11 @@ def save(fd, dominfo, network, live, dst
             os.close(qemu_fd)
             os.remove("/tmp/xen.qemu-dm.%d" % dominfo.getDomid())
 
-        dominfo.destroyDomain()
-        dominfo.testDeviceComplete()
-
+        if checkpoint:
+            dominfo.resumeDomain()
+        else:
+            dominfo.destroyDomain()
+            dominfo.testDeviceComplete()
         try:
             dominfo.setName(domain_name)
         except VmError:
diff -r b2edb7f33932 -r b207ed258dcf tools/python/xen/xend/XendDomain.py
--- a/tools/python/xen/xend/XendDomain.py	Mon Feb 26 18:36:04 2007 -0800
+++ b/tools/python/xen/xend/XendDomain.py	Mon Feb 26 18:36:04 2007 -0800
@@ -1172,7 +1172,7 @@ class XendDomain:
         XendCheckpoint.save(sock.fileno(), dominfo, True, live, dst)
         sock.close()
 
-    def domain_save(self, domid, dst):
+    def domain_save(self, domid, dst, checkpoint):
         """Start saving a domain to file.
 
         @param domid: Domain ID or Name
@@ -1196,8 +1196,8 @@ class XendDomain:
                 oflags |= os.O_LARGEFILE
             fd = os.open(dst, oflags)
             try:
-                # For now we don't support 'live checkpoint' 
-                XendCheckpoint.save(fd, dominfo, False, False, dst)
+                XendCheckpoint.save(fd, dominfo, False, False, dst,
+                                    checkpoint=checkpoint)
             finally:
                 os.close(fd)
         except OSError, ex:
diff -r b2edb7f33932 -r b207ed258dcf tools/python/xen/xm/main.py
--- a/tools/python/xen/xm/main.py	Mon Feb 26 18:36:04 2007 -0800
+++ b/tools/python/xen/xm/main.py	Mon Feb 26 18:36:04 2007 -0800
@@ -102,7 +102,7 @@ SUBCOMMAND_HELP = {
     'reboot'      : ('<Domain> [-wa]', 'Reboot a domain.'),
     'restore'     : ('<CheckpointFile> [-p]',
                      'Restore a domain from a saved state.'),
-    'save'        : ('<Domain> <CheckpointFile>',
+    'save'        : ('[-c] <Domain> <CheckpointFile>',
                      'Save a domain state to restore later.'),
     'shutdown'    : ('<Domain> [-waRH]', 'Shutdown a domain.'),
     'top'         : ('', 'Monitor a host and the domains in real time.'),
@@ -232,6 +232,9 @@ SUBCOMMAND_OPTIONS = {
     'resume': (
       ('-p', '--paused', 'Do not unpause domain after resuming it'),
     ),
+    'save': (
+      ('-c', '--checkpoint', 'Leave domain running after creating snapshot'),
+    ),
    'restore': (
       ('-p', '--paused', 'Do not unpause domain after restoring it'),
     ),
@@ -604,21 +607,37 @@ def xm_shell(args):
 #########################################################################
 
 def xm_save(args):
-    arg_check(args, "save", 2)
-
-    try:
-        dominfo = parse_doms_info(server.xend.domain(args[0]))
+    arg_check(args, "save", 2, 3)
+
+    try:
+        (options, params) = getopt.gnu_getopt(args, 'c', ['checkpoint'])
+    except getopt.GetoptError, opterr:
+        err(opterr)
+        sys.exit(1)
+
+    checkpoint = False
+    for (k, v) in options:
+        if k in ['-c', '--checkpoint']:
+            checkpoint = True
+
+    if len(params) != 2:
+        err("Wrong number of parameters")
+        usage('save')
+        sys.exit(1)
+
+    try:
+        dominfo = parse_doms_info(server.xend.domain(params[0]))
     except xmlrpclib.Fault, ex:
         raise ex
     
     domid = dominfo['domid']
-    savefile = os.path.abspath(args[1])
+    savefile = os.path.abspath(params[1])
 
     if not os.access(os.path.dirname(savefile), os.W_OK):
         err("xm save: Unable to create file %s" % savefile)
         sys.exit(1)
     
-    server.xend.domain.save(domid, savefile)
+    server.xend.domain.save(domid, savefile, checkpoint)
     
 def xm_restore(args):
     arg_check(args, "restore", 1, 2)

[PATCH 3 of 3] Add suspend_cancel flag to linux elf notes

[Brendan Cully]

# HG changeset patch
# User Brendan Cully <brendan@cs.ubc.ca>
# Date 1172543764 28800
# Node ID 728b370eaca01f5d7dd3e1ec59b34a0860372645
# Parent  b207ed258dcf3a6aee865ca82bc170f68e589342
Add suspend_cancel flag to linux elf notes.

Signed-off-by: Brendan Cully <brendan@cs.ubc.ca>

diff -r b207ed258dcf -r 728b370eaca0 linux-2.6-xen-sparse/arch/i386/kernel/head-xen.S
--- a/linux-2.6-xen-sparse/arch/i386/kernel/head-xen.S	Mon Feb 26 18:36:04 2007 -0800
+++ b/linux-2.6-xen-sparse/arch/i386/kernel/head-xen.S	Mon Feb 26 18:36:04 2007 -0800
@@ -175,6 +175,7 @@ ENTRY(cpu_gdt_table)
 	.ascii	         "|auto_translated_physmap"
 	.ascii	         "|pae_pgdir_above_4gb"
 	.ascii	         "|supervisor_mode_kernel"
+	.ascii		 "|suspend_cancel"
 #ifdef CONFIG_X86_PAE
 	.ascii	",PAE=yes[extended-cr3]"
 #else
@@ -197,7 +198,7 @@ ENTRY(cpu_gdt_table)
 	ELFNOTE(Xen, XEN_ELFNOTE_ENTRY,          .long,  startup_32)
 	ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, .long,  hypercall_page)
 	ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW,   .long,  HYPERVISOR_VIRT_START)
-	ELFNOTE(Xen, XEN_ELFNOTE_FEATURES,       .asciz, "writable_page_tables|writable_descriptor_tables|auto_translated_physmap|pae_pgdir_above_4gb|supervisor_mode_kernel")
+	ELFNOTE(Xen, XEN_ELFNOTE_FEATURES,       .asciz, "writable_page_tables|writable_descriptor_tables|auto_translated_physmap|pae_pgdir_above_4gb|supervisor_mode_kernel|suspend_cancel")
 #ifdef CONFIG_X86_PAE
 	ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE,       .asciz, "yes")
 	ELFNOTE(Xen, XEN_ELFNOTE_L1_MFN_VALID,   .quad,  _PAGE_PRESENT,_PAGE_PRESENT)
diff -r b207ed258dcf -r 728b370eaca0 xen/common/libelf/libelf-dominfo.c
--- a/xen/common/libelf/libelf-dominfo.c	Mon Feb 26 18:36:04 2007 -0800
+++ b/xen/common/libelf/libelf-dominfo.c	Mon Feb 26 18:36:04 2007 -0800
@@ -12,7 +12,8 @@ const char *elf_xen_feature_names[] = {
     [XENFEAT_writable_descriptor_tables] = "writable_descriptor_tables",
     [XENFEAT_auto_translated_physmap] = "auto_translated_physmap",
     [XENFEAT_supervisor_mode_kernel] = "supervisor_mode_kernel",
-    [XENFEAT_pae_pgdir_above_4gb] = "pae_pgdir_above_4gb"
+    [XENFEAT_pae_pgdir_above_4gb] = "pae_pgdir_above_4gb",
+    [XENFEAT_suspend_cancel] = "suspend_cancel"
 };
 const int elf_xen_features =
 sizeof(elf_xen_feature_names) / sizeof(elf_xen_feature_names[0]);
diff -r b207ed258dcf -r 728b370eaca0 xen/include/public/features.h
--- a/xen/include/public/features.h	Mon Feb 26 18:36:04 2007 -0800
+++ b/xen/include/public/features.h	Mon Feb 26 18:36:04 2007 -0800
@@ -56,6 +56,12 @@
  */
 #define XENFEAT_pae_pgdir_above_4gb        4
 
+/*
+ * If set, the guest is checkpointable. If the suspend hypercall returns
+ * 1, the domain will undo its suspend prep work and continue.
+ */
+#define XENFEAT_suspend_cancel             5
+
 #define XENFEAT_NR_SUBMAPS 1
 
 #endif /* __XEN_PUBLIC_FEATURES_H__ */

Re: [PATCH 0 of 3] Enable domain checkpointing via xm save --checkpoint

[Keir Fraser]

Basically okay, but please create a new elfnote rather than adding a
feature. The feature stuff has kind of got a bit confused, but is meant to
indicate features provided by Xen to a particular guest. The feature maps in
the guest header indicate which Xen features it requires and/or supports.
This new flag has nothing to do with the hypervisor itself so I think it
belongs in its own elfnote (XEN_ELFNOTE_SUSPEND_CANCEL?). This can either be
a boolean (so no elfnote description bytes) or perhaps we should map it to a
long with value 1 for now (so we can add further save/restore revisions
later if we like).

Oh, also remember to add the elfnote to the x86/64 head-xen.S.

 -- Keir

On 27/2/07 02:36, "Brendan Cully" <brendan@cs.ubc.ca> wrote:

> The attached patches create a --checkpoint flag for xm save:
> 
> xc_domain_resume lets the caller alert the guest that it is being
> resumed. Callers should not do this unless they know the guest
> supports the operation.
> 
> When xend wants to resume a domain, it checks for the suspend_cancel
> feature in the elfnotes it has recorded from domain build or
> resume. If the flag is present, it simply lets the guest handle
> resuming, otherwise it tears down and rebuilds the domain and lets the
> guest act as if it is resuming in a new domain.
> 
> A new suspend_cancel feature is added to the kernel 'features'
> elfnote.
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Re: [PATCH 0 of 3] Enable domain checkpointing via xm save --checkpoint

[Brendan Cully]

On Tuesday, 27 February 2007 at 16:10, Keir Fraser wrote:
> Basically okay, but please create a new elfnote rather than adding a
> feature. The feature stuff has kind of got a bit confused, but is meant to
> indicate features provided by Xen to a particular guest. The feature maps in
> the guest header indicate which Xen features it requires and/or supports.
> This new flag has nothing to do with the hypervisor itself so I think it
> belongs in its own elfnote (XEN_ELFNOTE_SUSPEND_CANCEL?). This can either be
> a boolean (so no elfnote description bytes) or perhaps we should map it to a
> long with value 1 for now (so we can add further save/restore revisions
> later if we like).

Ok, I've resubmitted with this approach, making the note a long.

> Oh, also remember to add the elfnote to the x86/64 head-xen.S.

Done, thanks for the heads up.