Currently, Ryu will send the OFPGroupMod message with
"command_bucket_id=0" by the default if using the OpenFlow version 1.5.
OTOH, Open vSwitch expects "command_bucket_id=OFPG_BUCKET_ALL" if
"command" is other than OFPGC_INSERT_BUCKET or OFPGC_REMOVE_BUCKET,
otherwise OVS will rejects the OFPGroupMod message.
This patch changes the default value of "command_bucket_id" to
OFPG_BUCKET_ALL as OVS expects.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, the matches of flows in the learning switch examples are based
on the destination and the in_port. In a multi-OF-switch environment this
could cause issues when doing calls from a host on a first switch to
multiple hosts on another switch, namely that new destinations aren't
added as a flow because the source is already known on the second switch
(when doing multiple calls to different hosts on another switch).
This patch fixes this issue by adding the eth_src to the match field.
Reported-by: Jerico Moeyersons <jerico09@hotmail.com>
In some test cases in test_icmpv6.py, @raises() decorator does not catch
the correct Exception, because "pack" function cannot be resolved and it
need to be specified as "struct.pack".
This patch fixes to this unresolved references and make it more strictly
matching of Exception class (e.g., AssertionError).
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Currently, ICMPv4 packet instance without its payload, it will raise
TypeError when calculating packet length, because the default value of
"data" field (payload data) is None and len() can not be applied against
None type object.
This patch changes the default value of "data" to the empty binary
value and confirms the value type of "data" field before serialization.
Also, this patch applies the same improvements to ICMPv6 packet library.
Reported-by: William Fisher <william.w.fisher@gmail.com>
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
As a default, netaddr.valid_ipv4() uses the "aton" rule for its validation,
so valid_ipv4('2') and valid_ipv4('1.2.3') returns True.
It may cause unexpected behaviors.
This commit replaces netaddr.valid_ipv4() to ryu.lib.ip.valid_ipv4(),
which is wrapper of netaddr.valid_ipv4(), and use "pton" rule in the method.
So, address representation like '2' or '1.2.3' will be rejected.
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Currently, 'is_connected' attribute in SshServer is refered
before it is created, in Python 2.7.
This raises an Exception and the SSH operation won't work
anymore after this happens.
This commit ensures that 'is_connected' is created before
it is refered.
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, ryu-manager uses distutils.version.LooseVersion for
"--zapi-frr-version" to parse the given version sting.
With custom type class for oslo_config.cfg.Opt, oslo_config might access
__class__ attribute for equal comparison.
But in case on Python 2, LooseVersion does not have __class__ attribute
and it causes AttributeError. (This error is not always reproduced)
This patch injects required attribute into LooseVersion and avoids this
problem.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
To connect OVSDB Manager library to OVS, some additional configurations
on OVS is required, but currently no description about these
configurations.
This patch adds descriptions for OVS configurations and includes some
improvements of sample application.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, with ryu.app.ofctl.api.send_msg(),
a BarrierRequest will send with every messages.
So, using send_msg() with a BarrierRequest will
cause duplex transmission of BarrierRequests.
This commit enables to send a single BarrierRequest
if send_msg() is used with a BarrierRequest
by avoiding sending an additional BarrierRequest.
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch enables to support Zebra messages of FRRouting version 3.0
and introduces a flag to switch FRRouting version.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, rest_qos.py will raise AttributeError when deleting OVSDB
server address because rest_qos.py will try to split the given address
string but the address is None when deleting.
This patch checks if the given address is None or not before the string
manipulation and fixes this problem.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, BGPSpeaker will listen the incoming connection with the
unspecified IP addresses ('0.0.0.0' and '::').
This patch enables to specify the local listen address list for starting
BGPSpeaker instance.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, when new arguments are introduced into BGPSpeaker,
ryu.application.RyuBGPSpeaker need modifications to support the
introduced arguments.
This patch enables to pass arguments to BGPSpeaker without some
modifications on RyuBGPSpeaker.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch enables to initiate OpenFlow connection from controller side
by using "--ofp-switch-address-list" and "--ofp-switch-connect-interval"
options.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Note: The binary representation is not extracted from the real packets
and based on the implemetation, so might be required to be fixup with
the real packets.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch implements the following messages on FRRouting:
- FRR_ZEBRA_MPLS_LABELS_ADD
- FRR_ZEBRA_MPLS_LABELS_DELETE
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch implements the following messages on FRRouting:
- FRR_ZEBRA_INTERFACE_ENABLE_RADV
- FRR_ZEBRA_INTERFACE_DISABLE_RADV
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch implements the following messages on FRRouting:
- ZEBRA_INTERFACE_NBR_ADDRESS_ADD
- ZEBRA_INTERFACE_NBR_ADDRESS_DELETE
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Note: This patch does not support all messages defined in FRRouting,
and some of message are left on TODO.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
The ZEBRA_IPV4_IMPORT_LOOKUP/ZEBRA_IPV6_IMPORT_LOOKUP has asymmetric
structures at when sent to Zebra daemon and when sent from Zebra daemon.
Especially, "prefix_len" field is inserted before "prefix" filed when
sent to Zebra daemon.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch fixes to parse the missing field of the number of unreserved
bandwidth classes in Interface Link Parameters structure.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
In some systems pip is not installed and the parse_requirements() function
is not needed. So only import pip if needed. Otherwise this fails with:
from ryu.base import app_manager
File "/usr/lib/python2.7/site-packages/ryu/base/app_manager.py", line 34, in <module>
from ryu import utils
File "/usr/lib/python2.7/site-packages/ryu/utils.py", line 23, in <module>
from pip import req as pip_req
ImportError: No module named pip
Signed-off-by: Thomas Bechtold <tbechtold@suse.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
If remove port data from self.port_state first,
self.get_port method can't get the port data,
this problem will cause link can't immediately be removed.
This patch remove port data after get_port and link delete.
Signed-off-by: Yu Ren <j6y4u4xup6@gmail.com>
Signed-off-by: mgrex97 <j6y4u4xup6@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, a thread for processing outgoing message will be remained in
"RpcSession" activity even if RPC session is closed by the remote peer,
and garbages on memory will grow.
This patch fixes to close "RpcSession" activity when RPC session
closing.
Signed-off-by: IWASE Yusuke <iwase.yusuke0@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Currently, Ryu expects that the IPv4 peer address format is
"4 bytes with IPv4 address, followed by 12 bytes padding".
However RFC 7854 says the format is
"12 bytes padding, followed by 4 bytes with IPv4 address".
This patch fixes it to correctly parse/serialize peer address.
Reported-by: Henk Smit <hhwsmit@xs4all.nl>
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
This patch enables Ryu to listen ovsdb connections on IPv6 Address.
Note that you need to make a config file like belows:
[ovsdb]
address=::
Signed-off-by: Dingyuan Hu <hdyvip@gmail.com>
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
About the 'length' field in ICMPv6 options,
RFC4861 says that "The value 0 is invalid".
This patch adds assertions to raise a exception
for such invalid ICMPv6 packets.
Reported-by: William Fisher <william.w.fisher@gmail.com>
Reported-by: Shivaram Mysore <shivaram.mysore@gmail.com>
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
The length of nd_option is units of 8 octets.
Currently, for nd_options which is undefined in the ICMPv6 packet library,
the parser assumes that the length is [8 * length - 2].
It causes fails while parsing these options.
This patch fixes it to parse such options correctly.
Reported-by: William Fisher <william.w.fisher@gmail.com>
Reported-by: Shivaram Mysore <shivaram.mysore@gmail.com>
Signed-off-by: Satoshi Fujimoto <satoshi.fujimoto7@gmail.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
