Add support for 18.01.
Fixed bugs.
Improved some config as the result of Tetration experience.
Change-Id: Ie4b8d2d77ca3ad165675c42fa2d9a4798d871f9d
Signed-off-by: John DeNisco <jdenisco@cisco.com>
diff --git a/extras/vpp_config/vpplib/AutoConfig.py b/extras/vpp_config/vpplib/AutoConfig.py
index b995f41..36b6833 100644
--- a/extras/vpp_config/vpplib/AutoConfig.py
+++ b/extras/vpp_config/vpplib/AutoConfig.py
@@ -109,9 +109,8 @@
answer = raw_input("Please enter the netmask [n.n.n.n]: ")
plen = IPAddress(answer).netmask_bits()
return '{}/{}'.format(ipaddr, plen)
- except:
+ except None:
print "Please enter a valid IPv4 address."
- continue
@staticmethod
def _ask_user_range(question, first, last, default):
@@ -401,14 +400,15 @@
devices += ' num-tx-desc {}\n'.format(num_tx_desc)
devices += ' }'
- if total_mbufs is not 0:
+ # If the total mbufs is not 0 or less than the default, set num-bufs
+ logging.debug("Total mbufs: {}".format(total_mbufs))
+ if total_mbufs is not 0 and total_mbufs > 16384:
devices += '\n num-mbufs {}'.format(total_mbufs)
return devices
@staticmethod
- def _calc_vpp_workers(node, vpp_workers, numa_node,
- other_cpus_end, total_vpp_workers,
+ def _calc_vpp_workers(node, vpp_workers, numa_node, other_cpus_end, total_vpp_workers,
reserve_vpp_main_core):
"""
Calculate the VPP worker information
@@ -442,6 +442,7 @@
start += 1
workers_end = start + total_vpp_workers - 1
+
if workers_end <= end:
if reserve_vpp_main_core:
node['cpu']['vpp_main_core'] = start - 1
@@ -459,7 +460,7 @@
@staticmethod
def _calc_desc_and_queues(total_numa_nodes,
total_ports_per_numa,
- total_vpp_cpus,
+ total_rx_queues,
ports_per_numa_value):
"""
Calculate the number of descriptors and queues
@@ -467,26 +468,20 @@
:param total_numa_nodes: The total number of numa nodes
:param total_ports_per_numa: The total number of ports for this
numa node
- :param total_vpp_cpus: The total number of cpus to allocate for vpp
+ :param total_rx_queues: The total number of rx queues / port
:param ports_per_numa_value: The value from the ports_per_numa
dictionary
:type total_numa_nodes: int
:type total_ports_per_numa: int
- :type total_vpp_cpus: int
+ :type total_rx_queues: int
:type ports_per_numa_value: dict
:returns The total number of message buffers
- :returns: The total number of vpp workers
- :rtype: int
:rtype: int
"""
- # Get the total vpp workers
- total_vpp_workers = total_vpp_cpus
- ports_per_numa_value['total_vpp_workers'] = total_vpp_workers
-
# Get the number of rx queues
- rx_queues = max(1, total_vpp_workers)
- tx_queues = total_vpp_workers * total_numa_nodes + 1
+ rx_queues = max(1, total_rx_queues)
+ tx_queues = rx_queues * total_numa_nodes + 1
# Get the descriptor entries
desc_entries = 1024
@@ -496,7 +491,7 @@
total_ports_per_numa)
total_mbufs = total_mbufs
- return total_mbufs, total_vpp_workers
+ return total_mbufs
@staticmethod
def _create_ports_per_numa(node, interfaces):
@@ -542,8 +537,7 @@
# Get the number of cpus to skip, we never use the first cpu
other_cpus_start = 1
- other_cpus_end = other_cpus_start + \
- node['cpu']['total_other_cpus'] - 1
+ other_cpus_end = other_cpus_start + node['cpu']['total_other_cpus'] - 1
other_workers = None
if other_cpus_end is not 0:
other_workers = (other_cpus_start, other_cpus_end)
@@ -553,28 +547,29 @@
vpp_workers = []
reserve_vpp_main_core = node['cpu']['reserve_vpp_main_core']
total_vpp_cpus = node['cpu']['total_vpp_cpus']
+ total_rx_queues = node['cpu']['total_rx_queues']
# If total_vpp_cpus is 0 or is less than the numa nodes with ports
# then we shouldn't get workers
- total_with_main = total_vpp_cpus
+ total_workers_node = total_vpp_cpus / len(ports_per_numa)
+ total_main = 0
if reserve_vpp_main_core:
- total_with_main += 1
+ total_main = 1
total_mbufs = 0
- if total_with_main is not 0:
+ if total_main + total_workers_node is not 0:
for item in ports_per_numa.items():
numa_node = item[0]
value = item[1]
# Get the number of descriptors and queues
- mbufs, total_vpp_workers = self._calc_desc_and_queues(
- len(ports_per_numa),
- len(value['interfaces']), total_vpp_cpus, value)
+ mbufs = self._calc_desc_and_queues(len(ports_per_numa),
+ len(value['interfaces']), total_rx_queues, value)
total_mbufs += mbufs
# Get the VPP workers
- reserve_vpp_main_core = self._calc_vpp_workers(
- node, vpp_workers, numa_node, other_cpus_end,
- total_vpp_workers, reserve_vpp_main_core)
+ reserve_vpp_main_core = self._calc_vpp_workers(node, vpp_workers, numa_node,
+ other_cpus_end, total_workers_node,
+ reserve_vpp_main_core)
total_mbufs *= 2.5
total_mbufs = int(total_mbufs)
@@ -923,26 +918,23 @@
print "\nYour system has {} core(s) and {} Numa Nodes.". \
format(total_cpus, len(numa_nodes))
- print "To begin, we suggest not reserving any cores for VPP",
- print "or other processes."
- print "Then to improve performance try reserving cores as needed. "
-
- max_other_cores = total_cpus / 2
- question = '\nHow many core(s) do you want to reserve for processes \
-other than VPP? [0-{}][0]? '.format(str(max_other_cores))
- total_other_cpus = self._ask_user_range(question, 0, max_other_cores,
- 0)
- node['cpu']['total_other_cpus'] = total_other_cpus
+ print "To begin, we suggest not reserving any cores for VPP or other processes."
+ print "Then to improve performance start reserving cores and adding queues as needed. "
max_vpp_cpus = 4
total_vpp_cpus = 0
if max_vpp_cpus > 0:
- question = "How many core(s) shall we reserve for VPP workers[0-{}][0]? ". \
- format(max_vpp_cpus)
+ question = "\nHow many core(s) shall we reserve for VPP [0-{}][0]? ".format(max_vpp_cpus)
total_vpp_cpus = self._ask_user_range(question, 0, max_vpp_cpus, 0)
node['cpu']['total_vpp_cpus'] = total_vpp_cpus
- max_main_cpus = max_vpp_cpus - total_vpp_cpus
+ max_other_cores = (total_cpus - total_vpp_cpus) / 2
+ question = 'How many core(s) do you want to reserve for processes other than VPP? [0-{}][0]? '. \
+ format(str(max_other_cores))
+ total_other_cpus = self._ask_user_range(question, 0, max_other_cores, 0)
+ node['cpu']['total_other_cpus'] = total_other_cpus
+
+ max_main_cpus = max_vpp_cpus + 1 - total_vpp_cpus
reserve_vpp_main_core = False
if max_main_cpus > 0:
question = "Should we reserve 1 core for the VPP Main thread? "
@@ -953,6 +945,11 @@
node['cpu']['reserve_vpp_main_core'] = reserve_vpp_main_core
node['cpu']['vpp_main_core'] = 0
+ question = "How many RX queues per port shall we use for VPP [1-4][1]? ". \
+ format(max_vpp_cpus)
+ total_rx_queues = self._ask_user_range(question, 1, 4, 1)
+ node['cpu']['total_rx_queues'] = total_rx_queues
+
def modify_cpu(self, ask_questions=True):
"""
Modify the cpu configuration, asking for the user for the values.
@@ -1002,7 +999,7 @@
node['cpu']['cpus_per_node'] = cpus_per_node
# Ask the user some questions
- if ask_questions:
+ if ask_questions and total_cpus >= 8:
self._modify_cpu_questions(node, total_cpus, numa_nodes)
# Populate the interfaces with the numa node
@@ -1042,8 +1039,11 @@
question += " the OS [y/N]? "
answer = self._ask_user_yn(question, 'n')
if answer == 'y':
- driver = device['unused'][0]
- VppPCIUtil.bind_vpp_device(node, driver, dvid)
+ if 'unused' in device and len(device['unused']) != 0 and device['unused'][0] != '':
+ driver = device['unused'][0]
+ VppPCIUtil.bind_vpp_device(node, driver, dvid)
+ else:
+ logging.debug('Could not bind device {}'.format(dvid))
vppd[dvid] = device
for dit in vppd.items():
dvid = dit[0]
@@ -1071,6 +1071,12 @@
for dit in vppd.items():
dvid = dit[0]
device = dit[1]
+ if 'unused' in device and len(device['unused']) != 0 and device['unused'][0] != '':
+ driver = device['unused'][0]
+ logging.debug('Binding device {} to driver {}'.format(dvid, driver))
+ VppPCIUtil.bind_vpp_device(node, driver, dvid)
+ else:
+ logging.debug('Could not bind device {}'.format(dvid))
dpdk_devices[dvid] = device
del other_devices[dvid]
@@ -1146,6 +1152,12 @@
for dit in vppd.items():
dvid = dit[0]
device = dit[1]
+ if 'unused' in device and len(device['unused']) != 0 and device['unused'][0] != '':
+ driver = device['unused'][0]
+ logging.debug('Binding device {} to driver {}'.format(dvid, driver))
+ VppPCIUtil.bind_vpp_device(node, driver, dvid)
+ else:
+ logging.debug('Could not bind device {}'.format(dvid))
dpdk_devices[dvid] = device
del kernel_devices[dvid]
@@ -1169,8 +1181,12 @@
for dit in vppd.items():
dvid = dit[0]
device = dit[1]
- driver = device['unused'][0]
- VppPCIUtil.bind_vpp_device(node, driver, dvid)
+ if 'unused' in device and len(device['unused']) != 0 and device['unused'][0] != '':
+ driver = device['unused'][0]
+ logging.debug('Binding device {} to driver {}'.format(dvid, driver))
+ VppPCIUtil.bind_vpp_device(node, driver, dvid)
+ else:
+ logging.debug('Could not bind device {}'.format(dvid))
kernel_devices[dvid] = device
del dpdk_devices[dvid]
@@ -1216,8 +1232,7 @@
print "\nThere currently a total of {} huge pages.". \
format(total)
- question = \
- "How many huge pages do you want [{} - {}][{}]? ". \
+ question = "How many huge pages do you want [{} - {}][{}]? ". \
format(MIN_TOTAL_HUGE_PAGES, maxpages, MIN_TOTAL_HUGE_PAGES)
answer = self._ask_user_range(question, 1024, maxpages, 1024)
node['hugepages']['total'] = str(answer)
diff --git a/extras/vpp_config/vpplib/QemuUtils.py b/extras/vpp_config/vpplib/QemuUtils.py
index 37a13e2..a76b873 100644
--- a/extras/vpp_config/vpplib/QemuUtils.py
+++ b/extras/vpp_config/vpplib/QemuUtils.py
@@ -34,6 +34,7 @@
class QemuUtils(object):
"""QEMU utilities."""
+ # noinspection PyDictCreation
def __init__(self, qemu_id=1):
self._qemu_id = qemu_id
# Path to QEMU binary
diff --git a/extras/vpp_config/vpplib/VPPUtil.py b/extras/vpp_config/vpplib/VPPUtil.py
index 4551cf4..4ea9413 100644
--- a/extras/vpp_config/vpplib/VPPUtil.py
+++ b/extras/vpp_config/vpplib/VPPUtil.py
@@ -20,7 +20,8 @@
from collections import Counter
# VPP_VERSION = '1707'
-VPP_VERSION = '1710'
+# VPP_VERSION = '1710'
+VPP_VERSION = '1801'
class VPPUtil(object):
@@ -141,10 +142,9 @@
reps = 'deb [trusted=yes] https://nexus.fd.io/content/'
# When using a stable branch
- # reps += 'repositories/fd.io.stable.{}.ubuntu.{}.main/ ./\n' \
- # .format(fdio_release, ubuntu_version)
- reps += 'repositories/fd.io.ubuntu.{}.main/ ./\n' \
- .format(ubuntu_version)
+ # reps += 'repositories/fd.io.stable.{}.ubuntu.{}.main/ ./\n'.format(fdio_release, ubuntu_version)
+ # When using release
+ reps += 'repositories/fd.io.ubuntu.{}.main/ ./\n'.format(ubuntu_version)
cmd = 'echo "{0}" | sudo tee {1}'.format(reps, sfile)
(ret, stdout, stderr) = self.exec_command(cmd)
@@ -207,8 +207,8 @@
# When using stable
# reps += 'baseurl=https://nexus.fd.io/content/repositories/fd.io.stable.{}.{}/\n'.\
# format(fdio_release, centos_version)
- reps += 'baseurl=https://nexus.fd.io/content/repositories/fd.io.{}/\n'.\
- format(centos_version)
+ # When using release
+ reps += 'baseurl=https://nexus.fd.io/content/repositories/fd.io.{}/\n'.format(centos_version)
reps += 'enabled=1\n'
reps += 'gpgcheck=0'
@@ -661,6 +661,7 @@
format(cmd, node['host'],
stdout, stderr))
+ # noinspection RegExpRedundantEscape
@staticmethod
def status(node):
"""