def restore(xd, fd):
signature = read_exact(fd, len(SIGNATURE),
"not a valid guest state file: signature read")
if signature != SIGNATURE:
raise XendError("not a valid guest state file: found '%s'" %
signature)
l = read_exact(fd, sizeof_int,
"not a valid guest state file: config size read")
vmconfig_size = unpack("!i", l)[0]
vmconfig_buf = read_exact(fd, vmconfig_size,
"not a valid guest state file: config read")
p = sxp.Parser()
p.input(vmconfig_buf)
if not p.ready:
raise XendError("not a valid guest state file: config parse")
vmconfig = p.get_val()
dominfo = xd.restore_(vmconfig)
store_port = dominfo.getStorePort()
console_port = dominfo.getConsolePort()
assert store_port
assert console_port
try:
l = read_exact(fd, sizeof_unsigned_long,
"not a valid guest state file: pfn count read")
nr_pfns = unpack("L", l)[0] # native sizeof long
if nr_pfns > 16*1024*1024: # XXX
raise XendError(
"not a valid guest state file: pfn count out of range")
balloon.free(xc.pages_to_kib(nr_pfns))
cmd = map(str, [xen.util.auxbin.pathTo(XC_RESTORE),
xc.handle(), fd, dominfo.getDomid(), nr_pfns,
store_port, console_port])
log.debug("[xc_restore]: %s", string.join(cmd))
handler = RestoreInputHandler()
forkHelper(cmd, fd, handler.handler, True)
if handler.store_mfn is None or handler.console_mfn is None:
raise XendError('Could not read store/console MFN')
dominfo.unpause()
dominfo.completeRestore(handler.store_mfn, handler.console_mfn)
return dominfo
except:
dominfo.destroy()
raise
def decrease_untouched_memory(self, value):
if not self.is_enabled():
return
elif self.untouched_memory < value:
raise VmError(('I need %d KiB untouch mem, but only have %d KiB untouched mem in Memory Pool') %(value,self.reserve_memory))
else:
self.untouched_memory -= value
log.debug("MemoryPool: decrease_untouched_memory: untouched_memory %d KiB" %self.untouched_memory)
return
def increase_memory(self, value):
if not self.is_enabled():
return
else:
self.reserve_memory += value
if self.reserve_memory > self.default_reserved_memory:
raise VmError(('the maxsize of memory pool is %d KiB, but current is %d KiB') %(value,self.reserve_memory))
log.debug("MemoryPool: increase_memory:%d, reserved_memory %d KiB" %(value,self.reserve_memory))
return
def decrease_memory(self, value):
if not self.is_enabled() or value <= 4096: #4M for PV guest kernel and ramdisk unzip
return
elif self.reserve_memory < value:
raise VMError(('I need %d KiB, but only have %d KiB in Memory Pool') %(value,self.reserve_memory))
else:
self.reserve_memory -= value
log.debug("MemoryPool: decrease_memory: decrease: %d reserved_memory %d KiB" %(value,self.reserve_memory))
return
def increase_memory(self, value):
if not self.is_enabled():
return
else:
self.reserve_memory += value
if self.reserve_memory > self.default_reserved_memory:
raise VmError(
("the maxsize of memory pool is %d KiB, but current is %d KiB") % (value, self.reserve_memory)
)
log.debug("MemoryPool: increase_memory:%d, reserved_memory %d KiB" % (value, self.reserve_memory))
return
def add_usbdev(self, busid):
# if the adding usb device should be owned by usbback
# and is probed by other usb drivers, seize it!
bus, intf = busid.split(':')
buses = vusb_util.get_assigned_buses()
if str(bus) in buses:
if not vusb_util.usb_intf_is_binded(busid):
log.debug("add_usb(): %s is binded to other driver" % busid)
vusb_util.unbind_usb_device(bus)
vusb_util.bind_usb_device(bus)
return
def decrease_memory(self, value):
if not self.is_enabled() or value <= 4096: # 4M for PV guest kernel and ramdisk unzip
return
elif self.reserve_memory < value:
raise VMError(("I need %d KiB, but only have %d KiB in Memory Pool") % (value, self.reserve_memory))
else:
self.reserve_memory -= value
log.debug(
"MemoryPool: decrease_memory: decrease: %d reserved_memory %d KiB" % (value, self.reserve_memory)
)
return
def add_usbdev(self, busid):
# if the adding usb device should be owned by usbback
# and is probed by other usb drivers, seize it!
bus, intf = busid.split(':')
buses = vusb_util.get_assigned_buses()
if str(bus) in buses:
if not vusb_util.usb_intf_is_binded(busid):
log.debug("add_usb(): %s is binded to other driver" % busid)
vusb_util.unbind_usb_device(bus)
vusb_util.bind_usb_device(bus)
return
def decrease_untouched_memory(self, value):
if not self.is_enabled():
return
elif self.untouched_memory < value:
raise VmError(
("I need %d KiB untouch mem, but only have %d KiB untouched mem in Memory Pool")
% (value, self.reserve_memory)
)
else:
self.untouched_memory -= value
log.debug("MemoryPool: decrease_untouched_memory: untouched_memory %d KiB" % self.untouched_memory)
return
def save(fd, dominfo, live):
write_exact(fd, SIGNATURE, "could not write guest state file: signature")
config = sxp.to_string(dominfo.sxpr())
domain_name = dominfo.getName()
# Rename the domain temporarily, so that we don't get a name clash if this
# domain is migrating (live or non-live) to the local host. Doing such a
# thing is useful for debugging.
dominfo.setName('migrating-' + domain_name)
try:
write_exact(fd, pack("!i", len(config)),
"could not write guest state file: config len")
write_exact(fd, config, "could not write guest state file: config")
# xc_save takes three customization parameters: maxit, max_f, and
# flags the last controls whether or not save is 'live', while the
# first two further customize behaviour when 'live' save is
# enabled. Passing "0" simply uses the defaults compiled into
# libxenguest; see the comments and/or code in xc_linux_save() for
# more information.
cmd = [xen.util.auxbin.pathTo(XC_SAVE), str(xc.handle()), str(fd),
str(dominfo.getDomid()), "0", "0", str(int(live)) ]
log.debug("[xc_save]: %s", string.join(cmd))
def saveInputHandler(line, tochild):
log.debug("In saveInputHandler %s", line)
if line == "suspend":
log.debug("Suspending %d ...", dominfo.getDomid())
dominfo.shutdown('suspend')
dominfo.waitForShutdown()
log.info("Domain %d suspended.", dominfo.getDomid())
tochild.write("done\n")
tochild.flush()
log.debug('Written done')
forkHelper(cmd, fd, saveInputHandler, False)
dominfo.destroyDomain()
except Exception, exn:
log.exception("Save failed on domain %s (%d).", domain_name,
dominfo.getDomid())
try:
dominfo.setName(domain_name)
except:
log.exception("Failed to reset the migrating domain's name")
raise Exception, exn
def _init_PIFs(self):
# Initialise PIFs
# First configure ones off disk
saved_pifs = self.state_store.load_state('pif')
if saved_pifs:
for pif_uuid, pif in saved_pifs.items():
try:
XendPIF.recreate(pif, pif_uuid)
except CreateUnspecifiedAttributeError:
log.warn("Error recreating PIF %s", pif_uuid)
# Next discover any existing PIFs and check
# they are not already configured
configured_pifs = [XendAPIStore.get(
pif_uuid, "PIF")
.get_interface_name()
for pif_uuid in XendPIF.get_all()]
unconfigured_pifs = [(name, mtu, mac)
for name, mtu, mac in linux_get_phy_ifaces()
if name not in configured_pifs]
# Get a mapping from interface to bridge
if_to_br = dict([(i,b)
for (b,ifs) in Brctl.get_state().items()
for i in ifs])
for name, mtu, mac in unconfigured_pifs:
# Check PIF is on bridge
# if not, ignore
bridge_name = if_to_br.get(name, None)
if bridge_name is not None:
# Translate bridge name to network uuid
for network_uuid in XendNetwork.get_all():
network = XendAPIStore.get(
network_uuid, 'network')
if network.get_name_label() == bridge_name:
XendPIF.create_phy(network_uuid, name,
mac, mtu)
break
else:
log.debug("Cannot find network for bridge %s "
"when configuring PIF %s",
(bridge_name, name))
def init(self):
xoptions = XendOptions.instance()
self.default_reserved_memory = xoptions.get_reserved_memory() * 1024 * 1024 # KiB
if self.default_reserved_memory <= 0:
return
self.enable_memory_pool = 1
self.dom0_ballooning = xoptions.get_enable_dom0_ballooning()
if not self.dom0_ballooning:
return
self.reserve_memory = 0
self.untouched_memory = 0
# init reserved memory
# if not reserve_memory_size:
xc = xen.lowlevel.xc.xc()
physinfo = xc.physinfo()
total_mem = physinfo["total_memory"]
if total_mem < self.reserve_memory:
self.default_reserved_memory = total_mem
self.reserve_memory = self.default_reserved_memory
self.untouched_memory = self.default_reserved_memory
log.debug("MemoryPool: init reserved_memory %d KiB" % self.reserve_memory)
def init(self):
xoptions = XendOptions.instance()
self.default_reserved_memory = xoptions.get_reserved_memory() * 1024 * 1024 #KiB
if self.default_reserved_memory <= 0:
return
self.enable_memory_pool = 1
self.dom0_ballooning = xoptions.get_enable_dom0_ballooning()
if not self.dom0_ballooning:
return
self.reserve_memory = 0
self.untouched_memory = 0
#init reserved memory
#if not reserve_memory_size:
xc = xen.lowlevel.xc.xc()
physinfo = xc.physinfo()
total_mem = physinfo['total_memory']
if total_mem < self.reserve_memory:
self.default_reserved_memory = total_mem
self.reserve_memory = self.default_reserved_memory
self.untouched_memory = self.default_reserved_memory
log.debug("MemoryPool: init reserved_memory %d KiB" %self.reserve_memory)
def _init_PIFs(self):
# Initialise PIFs
# First configure ones off disk
saved_pifs = self.state_store.load_state('pif')
if saved_pifs:
for pif_uuid, pif in saved_pifs.items():
try:
XendPIF.recreate(pif, pif_uuid)
except CreateUnspecifiedAttributeError:
log.warn("Error recreating PIF %s", pif_uuid)
# Next discover any existing PIFs and check
# they are not already configured
configured_pifs = [
XendAPIStore.get(pif_uuid, "PIF").get_interface_name()
for pif_uuid in XendPIF.get_all()
]
unconfigured_pifs = [(name, mtu, mac)
for name, mtu, mac in linux_get_phy_ifaces()
if name not in configured_pifs]
# Get a mapping from interface to bridge
if_to_br = dict([(i, b) for (b, ifs) in Brctl.get_state().items()
for i in ifs])
for name, mtu, mac in unconfigured_pifs:
# Check PIF is on bridge
# if not, ignore
bridge_name = if_to_br.get(name, None)
if bridge_name is not None:
# Translate bridge name to network uuid
for network_uuid in XendNetwork.get_all():
network = XendAPIStore.get(network_uuid, 'network')
if network.get_name_label() == bridge_name:
XendPIF.create_phy(network_uuid, name, mac, mtu)
break
else:
log.debug(
"Cannot find network for bridge %s "
"when configuring PIF %s", (bridge_name, name))
def forkHelper(cmd, fd, inputHandler, closeToChild):
child = xPopen3(cmd, True, -1, [fd, xc.handle()])
if closeToChild:
child.tochild.close()
thread = threading.Thread(target = slurp, args = (child.childerr,))
thread.start()
try:
try:
while 1:
line = child.fromchild.readline()
if line == "":
break
else:
line = line.rstrip()
log.debug('%s', line)
inputHandler(line, child.tochild)
thread.join()
except IOError, exn:
raise XendError('Error reading from child process for %s: %s' %
(cmd, exn))
finally:
child.fromchild.close()
child.childerr.close()
if not closeToChild:
child.tochild.close()
status = child.wait()
if status >> 8 == 127:
raise XendError("%s failed: popen failed" % string.join(cmd))
elif status != 0:
raise XendError("%s failed" % string.join(cmd))
def saveInputHandler(line, tochild):
log.debug("In saveInputHandler %s", line)
if line == "suspend":
log.debug("Suspending %d ...", dominfo.getDomid())
dominfo.shutdown('suspend')
dominfo.waitForShutdown()
log.info("Domain %d suspended.", dominfo.getDomid())
tochild.write("done\n")
tochild.flush()
log.debug('Written done')
def _init_PSCSIs(self):
# Initialise PSCSIs and PSCSI_HBAs
saved_pscsis = self.state_store.load_state('pscsi')
saved_pscsi_table = {}
if saved_pscsis:
for pscsi_uuid, pscsi_record in saved_pscsis.items():
try:
saved_pscsi_table[pscsi_record['scsi_id']] = pscsi_uuid
except KeyError:
pass
saved_pscsi_HBAs = self.state_store.load_state('pscsi_HBA')
saved_pscsi_HBA_table = {}
if saved_pscsi_HBAs:
for pscsi_HBA_uuid, pscsi_HBA_record in saved_pscsi_HBAs.items():
try:
physical_host = int(pscsi_HBA_record['physical_host'])
saved_pscsi_HBA_table[physical_host] = pscsi_HBA_uuid
except (KeyError, ValueError):
pass
pscsi_table = {}
pscsi_HBA_table = {}
pscsi_records = []
for pscsi_mask in xendoptions().get_pscsi_device_mask():
pscsi_records += vscsi_util.get_all_scsi_devices(pscsi_mask)
log.debug("pscsi record count: %s" % len(pscsi_records))
for pscsi_record in pscsi_records:
scsi_id = pscsi_record['scsi_id']
if scsi_id:
saved_HBA_uuid = None
pscsi_uuid = saved_pscsi_table.get(scsi_id, None)
if pscsi_uuid is None:
pscsi_uuid = uuid.createString()
saved_pscsi_table[scsi_id] = pscsi_uuid
else:
try:
saved_HBA_uuid = saved_pscsis[pscsi_uuid].get('HBA', None)
except KeyError:
log.warn("Multi-path SCSI devices are not supported for XenAPI")
return
physical_host = int(pscsi_record['physical_HCTL'].split(':')[0])
if pscsi_HBA_table.has_key(physical_host):
pscsi_HBA_uuid = pscsi_HBA_table[physical_host]
elif saved_pscsi_HBA_table.has_key(physical_host):
pscsi_HBA_uuid = saved_pscsi_HBA_table[physical_host]
pscsi_HBA_table[physical_host] = pscsi_HBA_uuid
else:
pscsi_HBA_uuid = uuid.createString()
pscsi_HBA_table[physical_host] = pscsi_HBA_uuid
if saved_HBA_uuid is not None and \
saved_HBA_uuid != pscsi_HBA_uuid:
log.debug('The PSCSI(%s) host number was changed', scsi_id)
pscsi_record['HBA'] = pscsi_HBA_uuid
pscsi_table[pscsi_uuid] = pscsi_record
for pscsi_uuid, pscsi_record in pscsi_table.items():
XendPSCSI(pscsi_uuid, pscsi_record)
for physical_host, pscsi_HBA_uuid in pscsi_HBA_table.items():
XendPSCSI_HBA(pscsi_HBA_uuid, {'physical_host': physical_host})
def free(required):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. No matter where
# we expect the free memory to come from, therefore, we need to wait for
# it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_min_mem
# is 0, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
need_mem = (required + 1023) / 1024 + BALLOON_OUT_SLACK
xroot = XendRoot.instance()
xc = xen.lowlevel.xc.xc()
try:
dom0_min_mem = xroot.get_dom0_min_mem()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
last_new_alloc = None
while retries < RETRY_LIMIT:
free_mem = xc.physinfo()['free_memory']
if free_mem >= need_mem:
log.debug("Balloon: free %d; need %d; done.", free_mem,
need_mem)
return
if retries == 0:
log.debug("Balloon: free %d; need %d.", free_mem, need_mem)
if dom0_min_mem > 0:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem)
if (new_alloc >= dom0_min_mem and new_alloc != last_new_alloc):
log.debug("Balloon: setting dom0 target to %d.", new_alloc)
dom0 = XendDomain.instance().privilegedDomain()
dom0.setMemoryTarget(new_alloc)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning.
time.sleep(sleep_time)
retries += 1
sleep_time += SLEEP_TIME_GROWTH
# Not enough memory; diagnose the problem.
if dom0_min_mem == 0:
raise VmError(('Not enough free memory and dom0_min_mem is 0, so '
'I cannot release any more. I need %d MiB but '
'only have %d.') % (need_mem, free_mem))
elif new_alloc < dom0_min_mem:
raise VmError(
('I need %d MiB, but dom0_min_mem is %d and shrinking to '
'%d MiB would leave only %d MiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + dom0_alloc - dom0_min_mem))
else:
raise VmError('The privileged domain did not balloon!')
finally:
del xc
def get_PIF_refs(self):
log.debug(XendPIF.get_all())
return XendPIF.get_all()
def remove_network(self, interface):
# TODO
log.debug("remove_network(): Not implemented.")
def add_network(self, interface):
# TODO
log.debug("add_network(): Not implemented.")
def free(required):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. No matter where
# we expect the free memory to come from, therefore, we need to wait for
# it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_min_mem
# is 0, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
need_mem = (required + 1023) / 1024 + BALLOON_OUT_SLACK
xroot = XendRoot.instance()
xc = xen.lowlevel.xc.xc()
try:
dom0_min_mem = xroot.get_dom0_min_mem()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
last_new_alloc = None
while retries < RETRY_LIMIT:
free_mem = xc.physinfo()['free_memory']
if free_mem >= need_mem:
log.debug("Balloon: free %d; need %d; done.", free_mem,
need_mem)
return
if retries == 0:
log.debug("Balloon: free %d; need %d.", free_mem, need_mem)
if dom0_min_mem > 0:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem)
if (new_alloc >= dom0_min_mem and
new_alloc != last_new_alloc):
log.debug("Balloon: setting dom0 target to %d.",
new_alloc)
dom0 = XendDomain.instance().privilegedDomain()
dom0.setMemoryTarget(new_alloc)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning.
time.sleep(sleep_time)
retries += 1
sleep_time += SLEEP_TIME_GROWTH
# Not enough memory; diagnose the problem.
if dom0_min_mem == 0:
raise VmError(('Not enough free memory and dom0_min_mem is 0, so '
'I cannot release any more. I need %d MiB but '
'only have %d.') %
(need_mem, free_mem))
elif new_alloc < dom0_min_mem:
raise VmError(
('I need %d MiB, but dom0_min_mem is %d and shrinking to '
'%d MiB would leave only %d MiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + dom0_alloc - dom0_min_mem))
else:
raise VmError('The privileged domain did not balloon!')
finally:
del xc
def free(need_mem):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. Such memory
# does, however, need to be accounted for when calculating how much dom0
# needs to balloon. No matter where we expect the free memory to come
# from, we need to wait for it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_min_mem
# is 0, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
xoptions = XendOptions.instance()
dom0 = XendDomain.instance().privilegedDomain()
xc = xen.lowlevel.xc.xc()
dom0_start_alloc_mb = get_dom0_current_alloc() / 1024
try:
dom0_min_mem = xoptions.get_dom0_min_mem() * 1024
dom0_alloc = get_dom0_current_alloc()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
new_alloc = 0
last_new_alloc = None
last_free = None
rlimit = RETRY_LIMIT
# If unreasonable memory size is required, we give up waiting
# for ballooning or scrubbing, as if had retried.
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
total_mem = physinfo['total_memory']
if dom0_min_mem > 0:
max_free_mem = total_mem - dom0_min_mem
else:
max_free_mem = total_mem - dom0_alloc
if need_mem >= max_free_mem:
retries = rlimit
while retries < rlimit:
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
if free_mem >= need_mem:
log.debug("Balloon: %d KiB free; need %d; done.", free_mem,
need_mem)
return
if retries == 0:
rlimit += (
(need_mem - free_mem) / 1024 / 1024) * RETRY_LIMIT_INCR
log.debug(
"Balloon: %d KiB free; %d to scrub; need %d; retries: %d.",
free_mem, scrub_mem, need_mem, rlimit)
if dom0_min_mem > 0:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem - scrub_mem)
if free_mem + scrub_mem >= need_mem:
if last_new_alloc == None:
log.debug("Balloon: waiting on scrubbing")
last_new_alloc = dom0_alloc
else:
if (new_alloc >= dom0_min_mem
and new_alloc != last_new_alloc):
new_alloc_mb = new_alloc / 1024 # Round down
log.debug("Balloon: setting dom0 target to %d MiB.",
new_alloc_mb)
dom0.setMemoryTarget(new_alloc_mb)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning or scrubbing.
time.sleep(sleep_time)
if retries < 2 * RETRY_LIMIT:
sleep_time += SLEEP_TIME_GROWTH
if last_free != None and last_free >= free_mem + scrub_mem:
retries += 1
last_free = free_mem + scrub_mem
# Not enough memory; diagnose the problem.
if dom0_min_mem == 0:
raise VmError(('Not enough free memory and dom0_min_mem is 0, so '
'I cannot release any more. I need %d KiB but '
'only have %d.') % (need_mem, free_mem))
elif new_alloc < dom0_min_mem:
raise VmError(
('I need %d KiB, but dom0_min_mem is %d and shrinking to '
'%d KiB would leave only %d KiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + scrub_mem + dom0_alloc - dom0_min_mem))
else:
dom0.setMemoryTarget(dom0_start_alloc_mb)
raise VmError(('Not enough memory is available, and dom0 cannot'
' be shrunk any further'))
finally:
del xc
args = [blexec]
if kernel:
args.append("--kernel=%s" % kernel)
if ramdisk:
args.append("--ramdisk=%s" % ramdisk)
if kernel_args:
args.append("--args=%s" % kernel_args)
if quiet:
args.append("-q")
args.append("--output=%s" % fifo)
if blargs:
args.extend(shlex.split(blargs))
args.append(disk)
try:
log.debug("Launching bootloader as %s." % str(args))
env = os.environ.copy()
env['TERM'] = 'vt100'
oshelp.close_fds()
os.execvpe(args[0], args, env)
except OSError, e:
print e
pass
os._exit(1)
# record that this domain is bootloading
dom.bootloader_pid = child
# On Solaris, the master pty side does not have terminal semantics,
# so don't try to set any attributes, as it will fail.
if os.uname()[0] != 'SunOS':
def __init__(self):
"""Initalises the state of all host specific objects such as
* host
* host_CPU
* host_metrics
* PIF
* PIF_metrics
* network
* Storage Repository
* PPCI
"""
self.xc = xen.lowlevel.xc.xc()
self.state_store = XendStateStore(xendoptions().get_xend_state_path())
self.monitor = XendMonitor()
self.monitor.start()
# load host state from XML file
saved_host = self.state_store.load_state('host')
if saved_host and len(saved_host.keys()) == 1:
self.uuid = saved_host.keys()[0]
host = saved_host[self.uuid]
self.name = host.get('name_label', socket.gethostname())
self.desc = host.get('name_description', '')
self.host_metrics_uuid = host.get('metrics_uuid',
uuid.createString())
try:
self.other_config = eval(host['other_config'])
except:
self.other_config = {}
self.cpus = {}
else:
self.uuid = uuid.createString()
self.name = socket.gethostname()
self.desc = ''
self.other_config = {}
self.cpus = {}
self.host_metrics_uuid = uuid.createString()
# put some arbitrary params in other_config as this
# is directly exposed via XenAPI
self.other_config["xen_pagesize"] = self.xeninfo_dict()["xen_pagesize"]
self.other_config["platform_params"] = self.xeninfo_dict(
)["platform_params"]
# load CPU UUIDs
saved_cpus = self.state_store.load_state('cpu')
for cpu_uuid, cpu in saved_cpus.items():
self.cpus[cpu_uuid] = cpu
cpuinfo = osdep.get_cpuinfo()
physinfo = self.physinfo_dict()
cpu_count = physinfo['nr_cpus']
cpu_features = physinfo['hw_caps']
virt_caps = physinfo['virt_caps']
# If the number of CPUs don't match, we should just reinitialise
# the CPU UUIDs.
if cpu_count != len(self.cpus):
self.cpus = {}
for i in range(cpu_count):
u = uuid.createString()
self.cpus[u] = {'uuid': u, 'number': i}
for u in self.cpus.keys():
number = self.cpus[u]['number']
# We can run off the end of the cpuinfo list if domain0 does not
# have #vcpus == #pcpus. In that case we just replicate one that's
# in the hash table.
if not cpuinfo.has_key(number):
number = cpuinfo.keys()[0]
if arch.type == "x86":
self.cpus[u].update({
'host':
self.uuid,
'features':
cpu_features,
'virt_caps':
virt_caps,
'speed':
int(float(cpuinfo[number]['cpu MHz'])),
'vendor':
cpuinfo[number]['vendor_id'],
'modelname':
cpuinfo[number]['model name'],
'stepping':
cpuinfo[number]['stepping'],
'flags':
cpuinfo[number]['flags'],
})
elif arch.type == "ia64":
self.cpus[u].update({
'host':
self.uuid,
'features':
cpu_features,
'speed':
int(float(cpuinfo[number]['cpu MHz'])),
'vendor':
cpuinfo[number]['vendor'],
'modelname':
cpuinfo[number]['family'],
'stepping':
cpuinfo[number]['model'],
'flags':
cpuinfo[number]['features'],
})
else:
self.cpus[u].update({
'host': self.uuid,
'features': cpu_features,
})
self.srs = {}
# Initialise networks
# First configure ones off disk
saved_networks = self.state_store.load_state('network')
if saved_networks:
for net_uuid, network in saved_networks.items():
try:
XendNetwork.recreate(network, net_uuid)
except CreateUnspecifiedAttributeError:
log.warn("Error recreating network %s", net_uuid)
# Next discover any existing bridges and check
# they are not already configured
bridges = Brctl.get_state().keys()
configured_bridges = [
XendAPIStore.get(network_uuid, "network").get_name_label()
for network_uuid in XendNetwork.get_all()
]
unconfigured_bridges = [
bridge for bridge in bridges if bridge not in configured_bridges
]
for unconfigured_bridge in unconfigured_bridges:
XendNetwork.create_phy(unconfigured_bridge)
# Initialise PIFs
# First configure ones off disk
saved_pifs = self.state_store.load_state('pif')
if saved_pifs:
for pif_uuid, pif in saved_pifs.items():
try:
XendPIF.recreate(pif, pif_uuid)
except CreateUnspecifiedAttributeError:
log.warn("Error recreating PIF %s", pif_uuid)
# Next discover any existing PIFs and check
# they are not already configured
configured_pifs = [
XendAPIStore.get(pif_uuid, "PIF").get_interface_name()
for pif_uuid in XendPIF.get_all()
]
unconfigured_pifs = [(name, mtu, mac)
for name, mtu, mac in linux_get_phy_ifaces()
if name not in configured_pifs]
# Get a mapping from interface to bridge
if_to_br = dict([(i, b) for (b, ifs) in Brctl.get_state().items()
for i in ifs])
for name, mtu, mac in unconfigured_pifs:
# Check PIF is on bridge
# if not, ignore
bridge_name = if_to_br.get(name, None)
if bridge_name is not None:
# Translate bridge name to network uuid
for network_uuid in XendNetwork.get_all():
network = XendAPIStore.get(network_uuid, 'network')
if network.get_name_label() == bridge_name:
XendPIF.create_phy(network_uuid, name, mac, mtu)
break
else:
log.debug(
"Cannot find network for bridge %s "
"when configuring PIF %s", (bridge_name, name))
# initialise storage
saved_srs = self.state_store.load_state('sr')
if saved_srs:
for sr_uuid, sr_cfg in saved_srs.items():
if sr_cfg['type'] == 'qcow_file':
self.srs[sr_uuid] = XendQCoWStorageRepo(sr_uuid)
elif sr_cfg['type'] == 'local':
self.srs[sr_uuid] = XendLocalStorageRepo(sr_uuid)
# Create missing SRs if they don't exist
if not self.get_sr_by_type('local'):
image_sr_uuid = uuid.createString()
self.srs[image_sr_uuid] = XendLocalStorageRepo(image_sr_uuid)
if not self.get_sr_by_type('qcow_file'):
qcow_sr_uuid = uuid.createString()
self.srs[qcow_sr_uuid] = XendQCoWStorageRepo(qcow_sr_uuid)
saved_pbds = self.state_store.load_state('pbd')
if saved_pbds:
for pbd_uuid, pbd_cfg in saved_pbds.items():
try:
XendPBD.recreate(pbd_uuid, pbd_cfg)
except CreateUnspecifiedAttributeError:
log.warn("Error recreating PBD %s", pbd_uuid)
# Initialise PPCIs
saved_ppcis = self.state_store.load_state('ppci')
saved_ppci_table = {}
if saved_ppcis:
for ppci_uuid, ppci_record in saved_ppcis.items():
try:
saved_ppci_table[ppci_record['name']] = ppci_uuid
except KeyError:
pass
for pci_dev in PciUtil.get_all_pci_devices():
ppci_record = {
'domain': pci_dev.domain,
'bus': pci_dev.bus,
'slot': pci_dev.slot,
'func': pci_dev.func,
'vendor_id': pci_dev.vendor,
'vendor_name': pci_dev.vendorname,
'device_id': pci_dev.device,
'device_name': pci_dev.devicename,
'revision_id': pci_dev.revision,
'class_code': pci_dev.classcode,
'class_name': pci_dev.classname,
'subsystem_vendor_id': pci_dev.subvendor,
'subsystem_vendor_name': pci_dev.subvendorname,
'subsystem_id': pci_dev.subdevice,
'subsystem_name': pci_dev.subdevicename,
'driver': pci_dev.driver
}
# If saved uuid exists, use it. Otherwise create one.
ppci_uuid = saved_ppci_table.get(pci_dev.name, uuid.createString())
XendPPCI(ppci_uuid, ppci_record)
def free(need_mem, dominfo):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. Such memory
# does, however, need to be accounted for when calculating how much dom0
# needs to balloon. No matter where we expect the free memory to come
# from, we need to wait for it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_ballooning
# is False, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
xoptions = XendOptions.instance()
dom0 = XendDomain.instance().privilegedDomain()
xc = xen.lowlevel.xc.xc()
try:
dom0_min_mem = xoptions.get_dom0_min_mem() * 1024
dom0_ballooning = xoptions.get_enable_dom0_ballooning()
dom0_alloc = get_dom0_current_alloc()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
new_alloc = 0
last_new_alloc = None
last_free = None
rlimit = RETRY_LIMIT
# stop tmem from absorbing any more memory (must THAW when done!)
xc.tmem_control(0,TMEMC_FREEZE,-1, 0, 0, "")
# If unreasonable memory size is required, we give up waiting
# for ballooning or scrubbing, as if had retried.
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
total_mem = physinfo['total_memory']
if dom0_ballooning:
max_free_mem = total_mem - dom0_min_mem
else:
max_free_mem = total_mem - dom0_alloc
if need_mem >= max_free_mem:
retries = rlimit
freeable_mem = free_mem + scrub_mem
if freeable_mem < need_mem and need_mem < max_free_mem:
# flush memory from tmem to scrub_mem and reobtain physinfo
need_tmem_kb = need_mem - freeable_mem
tmem_kb = xc.tmem_control(0,TMEMC_FLUSH,-1, need_tmem_kb, 0, "")
log.debug("Balloon: tmem relinquished %d KiB of %d KiB requested.",
tmem_kb, need_tmem_kb)
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
# Check whethercurrent machine is a numa system and the new
# created hvm has all its vcpus in the same node, if all the
# conditions above are fit. We will wait until all the pages
# in scrub list are freed (if waiting time go beyond 20s,
# we will stop waiting it.)
if physinfo['nr_nodes'] > 1 and retries == 0:
oldnode = -1
waitscrub = 1
vcpus = dominfo.info['cpus'][0]
for vcpu in vcpus:
nodenum = 0
for node in physinfo['node_to_cpu']:
for cpu in node:
if vcpu == cpu:
if oldnode == -1:
oldnode = nodenum
elif oldnode != nodenum:
waitscrub = 0
nodenum = nodenum + 1
if waitscrub == 1 and scrub_mem > 0:
log.debug("wait for scrub %s", scrub_mem)
while scrub_mem > 0 and retries < rlimit:
time.sleep(sleep_time)
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
retries += 1
sleep_time += SLEEP_TIME_GROWTH
log.debug("scrub for %d times", retries)
retries = 0
sleep_time = SLEEP_TIME_GROWTH
while retries < rlimit:
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
if free_mem >= need_mem:
log.debug("Balloon: %d KiB free; need %d; done.",
free_mem, need_mem)
return
if retries == 0:
rlimit += ((need_mem - free_mem)/1024/1024) * RETRY_LIMIT_INCR
log.debug("Balloon: %d KiB free; %d to scrub; need %d; retries: %d.",
free_mem, scrub_mem, need_mem, rlimit)
if dom0_ballooning:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem - scrub_mem)
if free_mem + scrub_mem >= need_mem:
if last_new_alloc == None:
log.debug("Balloon: waiting on scrubbing")
last_new_alloc = dom0_alloc
else:
if (new_alloc >= dom0_min_mem and
new_alloc != last_new_alloc):
new_alloc_mb = new_alloc / 1024 # Round down
log.debug("Balloon: setting dom0 target to %d MiB.",
new_alloc_mb)
dom0.setMemoryTarget(new_alloc_mb)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning or scrubbing.
time.sleep(sleep_time)
if retries < 2 * RETRY_LIMIT:
sleep_time += SLEEP_TIME_GROWTH
if last_free != None and last_free >= free_mem + scrub_mem:
retries += 1
last_free = free_mem + scrub_mem
# Not enough memory; diagnose the problem.
if not dom0_ballooning:
raise VmError(('Not enough free memory and enable-dom0-ballooning '
'is False, so I cannot release any more. '
'I need %d KiB but only have %d.') %
(need_mem, free_mem))
elif new_alloc < dom0_min_mem:
raise VmError(
('I need %d KiB, but dom0_min_mem is %d and shrinking to '
'%d KiB would leave only %d KiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + scrub_mem + dom0_alloc - dom0_min_mem))
else:
dom0_start_alloc_mb = get_dom0_current_alloc() / 1024
dom0.setMemoryTarget(dom0_start_alloc_mb)
raise VmError(
('Not enough memory is available, and dom0 cannot'
' be shrunk any further'))
finally:
# allow tmem to accept pages again
xc.tmem_control(0,TMEMC_THAW,-1, 0, 0, "")
del xc
def _init_PSCSIs(self):
# Initialise PSCSIs and PSCSI_HBAs
saved_pscsis = self.state_store.load_state('pscsi')
saved_pscsi_table = {}
if saved_pscsis:
for pscsi_uuid, pscsi_record in saved_pscsis.items():
try:
saved_pscsi_table[pscsi_record['scsi_id']] = pscsi_uuid
except KeyError:
pass
saved_pscsi_HBAs = self.state_store.load_state('pscsi_HBA')
saved_pscsi_HBA_table = {}
if saved_pscsi_HBAs:
for pscsi_HBA_uuid, pscsi_HBA_record in saved_pscsi_HBAs.items():
try:
physical_host = int(pscsi_HBA_record['physical_host'])
saved_pscsi_HBA_table[physical_host] = pscsi_HBA_uuid
except (KeyError, ValueError):
pass
pscsi_table = {}
pscsi_HBA_table = {}
pscsi_records = []
for pscsi_mask in xendoptions().get_pscsi_device_mask():
pscsi_records += vscsi_util.get_all_scsi_devices(pscsi_mask)
log.debug("pscsi record count: %s" % len(pscsi_records))
for pscsi_record in pscsi_records:
scsi_id = pscsi_record['scsi_id']
if scsi_id:
saved_HBA_uuid = None
pscsi_uuid = saved_pscsi_table.get(scsi_id, None)
if pscsi_uuid is None:
pscsi_uuid = uuid.createString()
saved_pscsi_table[scsi_id] = pscsi_uuid
else:
try:
saved_HBA_uuid = saved_pscsis[pscsi_uuid].get(
'HBA', None)
except KeyError:
log.warn(
"Multi-path SCSI devices are not supported for XenAPI"
)
return
physical_host = int(
pscsi_record['physical_HCTL'].split(':')[0])
if pscsi_HBA_table.has_key(physical_host):
pscsi_HBA_uuid = pscsi_HBA_table[physical_host]
elif saved_pscsi_HBA_table.has_key(physical_host):
pscsi_HBA_uuid = saved_pscsi_HBA_table[physical_host]
pscsi_HBA_table[physical_host] = pscsi_HBA_uuid
else:
pscsi_HBA_uuid = uuid.createString()
pscsi_HBA_table[physical_host] = pscsi_HBA_uuid
if saved_HBA_uuid is not None and \
saved_HBA_uuid != pscsi_HBA_uuid:
log.debug('The PSCSI(%s) host number was changed', scsi_id)
pscsi_record['HBA'] = pscsi_HBA_uuid
pscsi_table[pscsi_uuid] = pscsi_record
for pscsi_uuid, pscsi_record in pscsi_table.items():
XendPSCSI(pscsi_uuid, pscsi_record)
for physical_host, pscsi_HBA_uuid in pscsi_HBA_table.items():
XendPSCSI_HBA(pscsi_HBA_uuid, {'physical_host': physical_host})
def add_network(self, interface):
# TODO
log.debug("add_network(): Not implemented.")
def get_PIF_refs(self):
log.debug(XendPIF.get_all())
return XendPIF.get_all()
def remove_usbdev(self, busid):
log.debug("remove_usbdev(): Not implemented.")
def free(need_mem, dominfo):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. Such memory
# does, however, need to be accounted for when calculating how much dom0
# needs to balloon. No matter where we expect the free memory to come
# from, we need to wait for it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_ballooning
# is False, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
xoptions = XendOptions.instance()
dom0 = XendDomain.instance().privilegedDomain()
xc = xen.lowlevel.xc.xc()
try:
dom0_min_mem = xoptions.get_dom0_min_mem() * 1024
dom0_ballooning = xoptions.get_enable_dom0_ballooning()
dom0_alloc = get_dom0_current_alloc()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
new_alloc = 0
last_new_alloc = None
last_free = None
rlimit = RETRY_LIMIT
# If unreasonable memory size is required, we give up waiting
# for ballooning or scrubbing, as if had retried.
physinfo = xc.physinfo()
free_mem = physinfo["free_memory"]
scrub_mem = physinfo["scrub_memory"]
total_mem = physinfo["total_memory"]
if dom0_ballooning:
max_free_mem = total_mem - dom0_min_mem
else:
max_free_mem = total_mem - dom0_alloc
if need_mem >= max_free_mem:
retries = rlimit
# Check whethercurrent machine is a numa system and the new
# created hvm has all its vcpus in the same node, if all the
# conditions above are fit. We will wait until all the pages
# in scrub list are freed (if waiting time go beyond 20s,
# we will stop waiting it.)
if physinfo["nr_nodes"] > 1 and retries == 0:
oldnode = -1
waitscrub = 1
vcpus = dominfo.info["cpus"][0]
for vcpu in vcpus:
nodenum = 0
for node in physinfo["node_to_cpu"]:
for cpu in node:
if vcpu == cpu:
if oldnode == -1:
oldnode = nodenum
elif oldnode != nodenum:
waitscrub = 0
nodenum = nodenum + 1
if waitscrub == 1 and scrub_mem > 0:
log.debug("wait for scrub %s", scrub_mem)
while scrub_mem > 0 and retries < rlimit:
time.sleep(sleep_time)
physinfo = xc.physinfo()
free_mem = physinfo["free_memory"]
scrub_mem = physinfo["scrub_memory"]
retries += 1
sleep_time += SLEEP_TIME_GROWTH
log.debug("scrub for %d times", retries)
retries = 0
sleep_time = SLEEP_TIME_GROWTH
while retries < rlimit:
physinfo = xc.physinfo()
free_mem = physinfo["free_memory"]
scrub_mem = physinfo["scrub_memory"]
if free_mem >= need_mem:
log.debug("Balloon: %d KiB free; need %d; done.", free_mem, need_mem)
return
if retries == 0:
rlimit += ((need_mem - free_mem) / 1024 / 1024) * RETRY_LIMIT_INCR
log.debug(
"Balloon: %d KiB free; %d to scrub; need %d; retries: %d.", free_mem, scrub_mem, need_mem, rlimit
)
if dom0_ballooning:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem - scrub_mem)
if free_mem + scrub_mem >= need_mem:
if last_new_alloc == None:
log.debug("Balloon: waiting on scrubbing")
last_new_alloc = dom0_alloc
else:
if new_alloc >= dom0_min_mem and new_alloc != last_new_alloc:
new_alloc_mb = new_alloc / 1024 # Round down
log.debug("Balloon: setting dom0 target to %d MiB.", new_alloc_mb)
dom0.setMemoryTarget(new_alloc_mb)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning or scrubbing.
time.sleep(sleep_time)
if retries < 2 * RETRY_LIMIT:
sleep_time += SLEEP_TIME_GROWTH
if last_free != None and last_free >= free_mem + scrub_mem:
retries += 1
last_free = free_mem + scrub_mem
# Not enough memory; diagnose the problem.
if not dom0_ballooning:
raise VmError(
(
"Not enough free memory and enable-dom0-ballooning "
"is False, so I cannot release any more. "
"I need %d KiB but only have %d."
)
% (need_mem, free_mem)
)
elif new_alloc < dom0_min_mem:
raise VmError(
("I need %d KiB, but dom0_min_mem is %d and shrinking to " "%d KiB would leave only %d KiB free.")
% (need_mem, dom0_min_mem, dom0_min_mem, free_mem + scrub_mem + dom0_alloc - dom0_min_mem)
)
else:
dom0_start_alloc_mb = get_dom0_current_alloc() / 1024
dom0.setMemoryTarget(dom0_start_alloc_mb)
raise VmError(("Not enough memory is available, and dom0 cannot" " be shrunk any further"))
finally:
del xc
def remove_network(self, interface):
# TODO
log.debug("remove_network(): Not implemented.")
args = [ blexec ]
if kernel:
args.append("--kernel=%s" % kernel)
if ramdisk:
args.append("--ramdisk=%s" % ramdisk)
if kernel_args:
args.append("--args=%s" % kernel_args)
if quiet:
args.append("-q")
args.append("--output=%s" % fifo)
if blargs:
args.extend(shlex.split(blargs))
args.append(disk)
try:
log.debug("Launching bootloader as %s." % str(args))
env = os.environ.copy()
env['TERM'] = 'vt100'
oshelp.close_fds()
os.execvpe(args[0], args, env)
except OSError, e:
print e
pass
os._exit(1)
# record that this domain is bootloading
dom.bootloader_pid = child
# On Solaris, the master pty side does not have terminal semantics,
# so don't try to set any attributes, as it will fail.
if os.uname()[0] != 'SunOS':
def remove_usbdev(self, busid):
log.debug("remove_usbdev(): Not implemented.")
def free(need_mem, dominfo):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. Such memory
# does, however, need to be accounted for when calculating how much dom0
# needs to balloon. No matter where we expect the free memory to come
# from, we need to wait for it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_ballooning
# is False, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
xoptions = XendOptions.instance()
dom0 = XendDomain.instance().privilegedDomain()
xc = xen.lowlevel.xc.xc()
memory_pool = MemoryPool.instance()
try:
dom0_min_mem = xoptions.get_dom0_min_mem() * 1024
dom0_ballooning = xoptions.get_enable_dom0_ballooning()
guest_size = 0
hvm = dominfo.info.is_hvm()
if memory_pool.is_enabled() and dominfo.domid:
if not hvm :
if need_mem <= 4 * 1024:
guest_size = 32
else:
guest_size = dominfo.image.getBitSize()
if guest_size == 32:
dom0_ballooning = 0
else: #no ballooning as memory pool enabled
dom0_ballooning = xoptions.get_enable_dom0_ballooning()
dom0_alloc = get_dom0_current_alloc()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
new_alloc = 0
last_new_alloc = None
last_free = None
rlimit = RETRY_LIMIT
mem_need_balloon = 0
left_memory_pool = 0
mem_target = 0
untouched_memory_pool = 0
real_need_mem = need_mem
# stop tmem from absorbing any more memory (must THAW when done!)
xc.tmem_control(0,TMEMC_FREEZE,-1, 0, 0, 0, "")
# If unreasonable memory size is required, we give up waiting
# for ballooning or scrubbing, as if had retried.
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
total_mem = physinfo['total_memory']
if memory_pool.is_enabled() and dominfo.domid:
if guest_size != 32 or hvm:
if need_mem > 4 * 1024:
dominfo.alloc_mem = need_mem
left_memory_pool = memory_pool.get_left_memory()
if need_mem > left_memory_pool:
dominfo.alloc_mem = 0
raise VmError(('Not enough free memory'
' so I cannot release any more. '
'I need %d KiB but only have %d in the pool.') %
(need_mem, memory_pool.get_left_memory()))
else:
untouched_memory_pool = memory_pool.get_untouched_memory()
if (left_memory_pool - untouched_memory_pool) > need_mem:
dom0_ballooning = 0
else:
mem_need_balloon = need_mem - left_memory_pool + untouched_memory_pool
need_mem = free_mem + scrub_mem + mem_need_balloon
if dom0_ballooning:
max_free_mem = total_mem - dom0_min_mem
else:
max_free_mem = total_mem - dom0_alloc
if need_mem >= max_free_mem:
retries = rlimit
freeable_mem = free_mem + scrub_mem
if freeable_mem < need_mem and need_mem < max_free_mem:
# flush memory from tmem to scrub_mem and reobtain physinfo
need_tmem_kb = need_mem - freeable_mem
tmem_kb = xc.tmem_control(0,TMEMC_FLUSH,-1, need_tmem_kb, 0, 0, "")
log.debug("Balloon: tmem relinquished %d KiB of %d KiB requested.",
tmem_kb, need_tmem_kb)
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
while retries < rlimit:
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
if free_mem >= need_mem:
if (guest_size != 32 or hvm) and dominfo.domid:
memory_pool.decrease_untouched_memory(mem_need_balloon)
memory_pool.decrease_memory(real_need_mem)
else:
log.debug("Balloon: %d KiB free; need %d; done.",
free_mem, need_mem)
return
if retries == 0:
rlimit += ((need_mem - free_mem)/1024/1024) * RETRY_LIMIT_INCR
log.debug("Balloon: %d KiB free; %d to scrub; need %d; retries: %d.",
free_mem, scrub_mem, need_mem, rlimit)
if dom0_ballooning:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem - scrub_mem)
if free_mem + scrub_mem >= need_mem:
if last_new_alloc == None:
log.debug("Balloon: waiting on scrubbing")
last_new_alloc = dom0_alloc
else:
if (new_alloc >= dom0_min_mem and
new_alloc != last_new_alloc):
new_alloc_mb = new_alloc / 1024 # Round down
log.debug("Balloon: setting dom0 target to %d MiB.",
new_alloc_mb)
dom0.setMemoryTarget(new_alloc_mb)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning or scrubbing.
time.sleep(sleep_time)
if retries < 2 * RETRY_LIMIT:
sleep_time += SLEEP_TIME_GROWTH
if last_free != None and last_free >= free_mem + scrub_mem:
retries += 1
last_free = free_mem + scrub_mem
# Not enough memory; diagnose the problem.
if not dom0_ballooning:
dominfo.alloc_mem = 0
raise VmError(('Not enough free memory and enable-dom0-ballooning '
'is False, so I cannot release any more. '
'I need %d KiB but only have %d.') %
(need_mem, free_mem))
elif new_alloc < dom0_min_mem:
dominfo.alloc_mem = 0
raise VmError(
('I need %d KiB, but dom0_min_mem is %d and shrinking to '
'%d KiB would leave only %d KiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + scrub_mem + dom0_alloc - dom0_min_mem))
else:
dom0_start_alloc_mb = get_dom0_current_alloc() / 1024
dom0.setMemoryTarget(dom0_start_alloc_mb)
dominfo.alloc_mem = 0
raise VmError(
('Not enough memory is available, and dom0 cannot'
' be shrunk any further'))
finally:
# allow tmem to accept pages again
xc.tmem_control(0,TMEMC_THAW,-1, 0, 0, 0, "")
del xc
def free(need_mem):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. Such memory
# does, however, need to be accounted for when calculating how much dom0
# needs to balloon. No matter where we expect the free memory to come
# from, we need to wait for it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_min_mem
# is 0, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
xoptions = XendOptions.instance()
dom0 = XendDomain.instance().privilegedDomain()
xc = xen.lowlevel.xc.xc()
dom0_start_alloc_mb = get_dom0_current_alloc() / 1024
try:
dom0_min_mem = xoptions.get_dom0_min_mem() * 1024
dom0_alloc = get_dom0_current_alloc()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
new_alloc = 0
last_new_alloc = None
last_free = None
rlimit = RETRY_LIMIT
# If unreasonable memory size is required, we give up waiting
# for ballooning or scrubbing, as if had retried.
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
total_mem = physinfo['total_memory']
if dom0_min_mem > 0:
max_free_mem = total_mem - dom0_min_mem
else:
max_free_mem = total_mem - dom0_alloc
if need_mem >= max_free_mem:
retries = rlimit
while retries < rlimit:
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
if free_mem >= need_mem:
log.debug("Balloon: %d KiB free; need %d; done.",
free_mem, need_mem)
return
if retries == 0:
rlimit += ((need_mem - free_mem)/1024/1024) * RETRY_LIMIT_INCR
log.debug("Balloon: %d KiB free; %d to scrub; need %d; retries: %d.",
free_mem, scrub_mem, need_mem, rlimit)
if dom0_min_mem > 0:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem - scrub_mem)
if free_mem + scrub_mem >= need_mem:
if last_new_alloc == None:
log.debug("Balloon: waiting on scrubbing")
last_new_alloc = dom0_alloc
else:
if (new_alloc >= dom0_min_mem and
new_alloc != last_new_alloc):
new_alloc_mb = new_alloc / 1024 # Round down
log.debug("Balloon: setting dom0 target to %d MiB.",
new_alloc_mb)
dom0.setMemoryTarget(new_alloc_mb)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning or scrubbing.
time.sleep(sleep_time)
if retries < 2 * RETRY_LIMIT:
sleep_time += SLEEP_TIME_GROWTH
if last_free != None and last_free >= free_mem + scrub_mem:
retries += 1
last_free = free_mem + scrub_mem
# Not enough memory; diagnose the problem.
if dom0_min_mem == 0:
raise VmError(('Not enough free memory and dom0_min_mem is 0, so '
'I cannot release any more. I need %d KiB but '
'only have %d.') %
(need_mem, free_mem))
elif new_alloc < dom0_min_mem:
raise VmError(
('I need %d KiB, but dom0_min_mem is %d and shrinking to '
'%d KiB would leave only %d KiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + scrub_mem + dom0_alloc - dom0_min_mem))
else:
dom0.setMemoryTarget(dom0_start_alloc_mb)
raise VmError(
('Not enough memory is available, and dom0 cannot'
' be shrunk any further'))
finally:
del xc
def free(need_mem, dominfo):
"""Balloon out memory from the privileged domain so that there is the
specified required amount (in KiB) free.
"""
# We check whether there is enough free memory, and if not, instruct dom0
# to balloon out to free some up. Memory freed by a destroyed domain may
# not appear in the free_memory field immediately, because it needs to be
# scrubbed before it can be released to the free list, which is done
# asynchronously by Xen; ballooning is asynchronous also. Such memory
# does, however, need to be accounted for when calculating how much dom0
# needs to balloon. No matter where we expect the free memory to come
# from, we need to wait for it to become available.
#
# We are not allowed to balloon below dom0_min_mem, or if dom0_ballooning
# is False, we cannot balloon at all. Memory can still become available
# through a rebooting domain, however.
#
# Eventually, we time out (presumably because there really isn't enough
# free memory).
#
# We don't want to set the memory target (triggering a watch) when that
# has already been done, but we do want to respond to changing memory
# usage, so we recheck the required alloc each time around the loop, but
# track the last used value so that we don't trigger too many watches.
xoptions = XendOptions.instance()
dom0 = XendDomain.instance().privilegedDomain()
xc = xen.lowlevel.xc.xc()
memory_pool = MemoryPool.instance()
try:
dom0_min_mem = xoptions.get_dom0_min_mem() * 1024
dom0_ballooning = xoptions.get_enable_dom0_ballooning()
guest_size = 0
hvm = dominfo.info.is_hvm()
if memory_pool.is_enabled() and dominfo.domid:
if not hvm:
if need_mem <= 4 * 1024:
guest_size = 32
else:
guest_size = dominfo.image.getBitSize()
if guest_size == 32:
dom0_ballooning = 0
else: #no ballooning as memory pool enabled
dom0_ballooning = xoptions.get_enable_dom0_ballooning()
dom0_alloc = get_dom0_current_alloc()
retries = 0
sleep_time = SLEEP_TIME_GROWTH
new_alloc = 0
last_new_alloc = None
last_free = None
rlimit = RETRY_LIMIT
mem_need_balloon = 0
left_memory_pool = 0
mem_target = 0
untouched_memory_pool = 0
real_need_mem = need_mem
# stop tmem from absorbing any more memory (must THAW when done!)
xc.tmem_control(0, TMEMC_FREEZE, -1, 0, 0, 0, "")
# If unreasonable memory size is required, we give up waiting
# for ballooning or scrubbing, as if had retried.
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
total_mem = physinfo['total_memory']
if memory_pool.is_enabled() and dominfo.domid:
if guest_size != 32 or hvm:
if need_mem > 4 * 1024:
dominfo.alloc_mem = need_mem
left_memory_pool = memory_pool.get_left_memory()
if need_mem > left_memory_pool:
dominfo.alloc_mem = 0
raise VmError(
('Not enough free memory'
' so I cannot release any more. '
'I need %d KiB but only have %d in the pool.') %
(need_mem, memory_pool.get_left_memory()))
else:
untouched_memory_pool = memory_pool.get_untouched_memory()
if (left_memory_pool - untouched_memory_pool) > need_mem:
dom0_ballooning = 0
else:
mem_need_balloon = need_mem - left_memory_pool + untouched_memory_pool
need_mem = free_mem + scrub_mem + mem_need_balloon
if dom0_ballooning:
max_free_mem = total_mem - dom0_min_mem
else:
max_free_mem = total_mem - dom0_alloc
if need_mem >= max_free_mem:
retries = rlimit
freeable_mem = free_mem + scrub_mem
if freeable_mem < need_mem and need_mem < max_free_mem:
# flush memory from tmem to scrub_mem and reobtain physinfo
need_tmem_kb = need_mem - freeable_mem
tmem_kb = xc.tmem_control(0, TMEMC_FLUSH, -1, need_tmem_kb, 0, 0,
"")
log.debug("Balloon: tmem relinquished %d KiB of %d KiB requested.",
tmem_kb, need_tmem_kb)
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
while retries < rlimit:
physinfo = xc.physinfo()
free_mem = physinfo['free_memory']
scrub_mem = physinfo['scrub_memory']
if free_mem >= need_mem:
if (guest_size != 32 or hvm) and dominfo.domid:
memory_pool.decrease_untouched_memory(mem_need_balloon)
memory_pool.decrease_memory(real_need_mem)
else:
log.debug("Balloon: %d KiB free; need %d; done.", free_mem,
need_mem)
return
if retries == 0:
rlimit += (
(need_mem - free_mem) / 1024 / 1024) * RETRY_LIMIT_INCR
log.debug(
"Balloon: %d KiB free; %d to scrub; need %d; retries: %d.",
free_mem, scrub_mem, need_mem, rlimit)
if dom0_ballooning:
dom0_alloc = get_dom0_current_alloc()
new_alloc = dom0_alloc - (need_mem - free_mem - scrub_mem)
if free_mem + scrub_mem >= need_mem:
if last_new_alloc == None:
log.debug("Balloon: waiting on scrubbing")
last_new_alloc = dom0_alloc
else:
if (new_alloc >= dom0_min_mem
and new_alloc != last_new_alloc):
new_alloc_mb = new_alloc / 1024 # Round down
log.debug("Balloon: setting dom0 target to %d MiB.",
new_alloc_mb)
dom0.setMemoryTarget(new_alloc_mb)
last_new_alloc = new_alloc
# Continue to retry, waiting for ballooning or scrubbing.
time.sleep(sleep_time)
if retries < 2 * RETRY_LIMIT:
sleep_time += SLEEP_TIME_GROWTH
if last_free != None and last_free >= free_mem + scrub_mem:
retries += 1
last_free = free_mem + scrub_mem
# Not enough memory; diagnose the problem.
if not dom0_ballooning:
dominfo.alloc_mem = 0
raise VmError(
('Not enough free memory and enable-dom0-ballooning '
'is False, so I cannot release any more. '
'I need %d KiB but only have %d.') % (need_mem, free_mem))
elif new_alloc < dom0_min_mem:
dominfo.alloc_mem = 0
raise VmError(
('I need %d KiB, but dom0_min_mem is %d and shrinking to '
'%d KiB would leave only %d KiB free.') %
(need_mem, dom0_min_mem, dom0_min_mem,
free_mem + scrub_mem + dom0_alloc - dom0_min_mem))
else:
dom0_start_alloc_mb = get_dom0_current_alloc() / 1024
dom0.setMemoryTarget(dom0_start_alloc_mb)
dominfo.alloc_mem = 0
raise VmError(('Not enough memory is available, and dom0 cannot'
' be shrunk any further'))
finally:
# allow tmem to accept pages again
xc.tmem_control(0, TMEMC_THAW, -1, 0, 0, 0, "")
del xc
