Currently, Controllers and CMSes that integrate with Open vSwitch
after following the IntegrationGuide.md, expect to start VMs after
the underlying network infrastructure is ready. I have been asked
a few times on how to handle the same with containers. This commit
provides documentation on one way to achieve it.
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
This test covers inclusion of public header files and basic usage
of the API such as vlog.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Also moves definitions for struct vconn and pvconn to the public
header. The provider interface is kept private.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
A new function vlog_insert_module() is introduced to avoid using
list_insert() from the vlog.h header.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Expose the struct ovs_list definition in <openvswitch/list.h>. Keep the
list access API private for now.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
struct list is a common name and can't be used in public headers.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Insted of exposing the full sat-math.h API, only the macros
used in headers is exposed through <openvswitch/util.h>
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
This picks the ovs_mutex and ovsthread_once API from lib/ovs-thread.h
and exposes it in <openvswitch/thread.h>.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
The following macros are renamed to avoid conflicts with other headers:
* WARN_UNUSED_RESULT to OVS_WARN_UNUSED_RESULT
* PRINTF_FORMAT to OVS_PRINTF_FORMAT
* NO_RETURN to OVS_NO_RETURN
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
This commit fixes unwanted whitespace in the ovs
bash completion script output.
Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>
internal table.
In current recirculation implementation, the flow misses (with
'recirc_id' set) are always looked up on the receiving bridge's
internal flow table. However, the bond port may actually reside
on another bridge which gets connected to the receiving bridge
via patch port. Since the recirculation rules are pushed to the
other bridge's internal table, the flow lookup on the receiving
bridge will match nothing but the drop rule, causing unexpected
packet drops.
This commit fixes the above bug via keeping lookup the misses
(with 'recirc_id' set) in default table (table 0) and processing
it until reaching the bridge that owns the bond port. Then,
the misses can hit the post recirculation flows as expected.
VMware-BZ: 1362178
Reported-by: Ansis Atteka <aatteka@nicira.com>
Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Andy Zhou <azhou@nicira.com>
This command is useful at least for testing.
Example output:
% ovs-appctl ovs/route/lookup '10.0.0.1'
gateway 172.17.0.254
dev wm0
%
Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
When using interfaces.d/<foobar>, interfaces are not picked up.
Let ifquery figure out the format of the interfaces files for us.
Signed-off-by: Jan Vansteenkiste <jan@vstone.eu>
Signed-off-by: Ben Pfaff <blp@nicira.com>
When the Packet-In message is triggered by a group action, OFPR_GROUP is
passed internally as the reason. The wire_reason() function converts the
reason to OFPR_ACTION if the wire protocol is earlier than OF1.4.
The wire_reason() function also converts other unsupported reasons
(i.e., OFPR_ACTION_SET and OFPR_PACKET_OUT) to OFPR_ACTION if it detects
a wire protocol earlier than OF1.4.
By default reason code OFPR_GROUP for Packet-In will be enabled for
async messages as in ofconn_flush(). Upon a connection being established
with a controller, the protocol version is checked and OFPR_GROUP will
be disabled in async config if the protocol is lower than OF1.4. Any
controller running OF1.4+ is still be able to enable OFPR_GROUP at its
will without being affected by this check.
The patch also includes tests cases for both OF1.3 and OF1.4 to ensure
proper reason code is given for packet-in message triggered by group
action.
Signed-off-by: Shu Shen <shu.shen@radisys.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Per review comment, in this patch, we update the return values of
NlFillOvsMsg() and NlFillOvsHdr() from NTSTATUS to BOOLEAN to make them
consistent with the Nl* functions.
Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Eitan Eliahu <eliahue@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
If ovs-vswitchd is killed at the right time, then a probe flow may be
left in the datapath. This commit adds the DPIF_FP_MODIFY flag to
feature probes so that re-creating the same probe flow will not cause an
error.
These flow_put flags were previously changed in commit a7d1bbdcfe
("ofproto-dpif: Use DPIF_FP_CREATE but not DPIF_FP_MODIFY."), despite
the commit message addressing a different case.
Suggested-by: Ben Pfaff <blp@nicira.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Various functions in ofproto-dpif and dpif-netlink detect support for
features in very similar ways. Refactor their common code to a single
function.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
The UFID parameter to dpif_flow_get() is optional, but the current
implementation dereferences it to initialize part of the output flow.
This field is filled in by the dpif implementation, so don't initialize
it here.
This does not fix any existing bug because every caller currently passes
in a UFID. The next patch will introduce the first call to
dpif_flow_get() that doesn't provide a UFID, which would break without
this change.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Refactor ovs-router so that it can work with non-Linux platforms
at least in some extent, using the existing route-table code as
a fallback. Known restriction: for such platforms, "ovs/router/show"
command does not show "Cached" kernel routes.
Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
The cache here doesn't work anymore as the recent
commit b772066ffd
("route-table: Remove Unregister.")
made this function called before daemonizing,
thus with a different pid.
Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Acked-by: Ben Pfaff <blp@nicira.com>
nla_is_last() is not available in 3.18, it's only in net-next.
Convert to grep based to check to account for distribution backports.
Fixes: 684b5f ("datapath: Rename last_action() as nla_is_last() and move to netlink.h")
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Commit 64bb477 "dpif: Minimize memory copy for revalidation." introduced
a bug where the corner case of ukey creation (in revalidator threads)
could result in access to uninitialized memory when deleting flows from
the datapath. This could result in OVS aborting or deadlock. Fix it.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Change a test so that the result will be the same in both
little-endian and big-endian systems by editing the test case so that
only one bit differs.
Reported-by: Mijo Safradin <mijo@linux.vnet.ibm.com>
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Preparation for supporting ONFACT_ET_COPY_FIELD.
ONF-JIRA: EXT-320
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
Since my original prototype, the oxm_id_len field was removed and
replaced by 2 bytes of padding.
ONF-JIRA: EXT-320
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
On current master, the 'struct tnl_port' in tunnel module will be
recreated whenever the global connectivity sequence number changes
(e.g. when adding unrelated flow). This is unnecessary and could
cause drop of tunnel packet if a lookup happens between the removal
and recreate.
This commit fixes the above issue by only checking the netdev's own
sequence number.
Found by code inspection.
Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>
The standard mode for printing flows doesn't always provide the full
range of information that is available, particularly with the UFID
changes. Turn on more detail.
Suggested-by: Justin Pettit <jpettit@nicira.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Gurucharan Shetty <gshetty@nicira.com>
When doing test builds, build against the following kernels:
- KERNEL=3.17.4
- KERNEL=3.16.7
- KERNEL=3.14.25
- KERNEL=3.12.33
- KERNEL=3.10.61
- KERNEL=3.4.104
- KERNEL=2.6.32.64
When doing pure kernel test builds, skip compilation of user space bits.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
An examination of the source code and the tests shows that the FAQ claimed
more for version 2.3 than was really in there.
Reported-by: null pointer <null.pointer.boom@gmail.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Moving the functions that build netlink messages to Netlink.c from
Vport.c
Signed-off-by: Nithin Raju <nithin@vmware.com>
Acked-by: Eitan Eliahu <eliahue@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Commit 7eb4b1f1d7 ("ofp-actions: Support OF1.5 (draft) masked
Set-Field, merge with reg_load.") introduced a bug in that a set_field
action that set an entire field would be translated incorrectly to
reg_load, if the field being set only occupied a portion of the bytes that
it contains. For example, an MPLS label is 20 bits but has a 4-byte field,
which meant that a set_field would get translated into a reg_load that
wrote all 32 bits; in turn, the receiver of that reg_load would reject it
because it was attempting to set invalid bits (the top 12 bits).
This commit fixes the problem by omitting invalid bits when encoding a
reg_load action.
Reported-by: Pravin Shelar <pshelar@nicira.com>
Tested-by: Pravin Shelar <pshelar@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Commit 526a7c85d1 "util: Add be32_prefix_mask()." added an include for
byte-order.h into util.h, which could cause link failures if users of
libopenvswitch defined their own version of htonll(). Change the
include, as only htonl() is needed and arpa/inet.h provides this.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
Automake sets $V to tell the compiler whether to print verbose messages
as it compiles or not. Add support for this variable in cccl, allowing
more quiet build output on windows if the build is configured with
--silent or the developer runs make V=0.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Occasionally, the testsuite would send the OFPT_SET_ASYNC before
ovs-vswitchd got a chance to send its OFPT_ROLE_REPLY message, causing a
reordering of the testsuite output and a false positive. Give the test
script something extra to do so this is less likely to happen.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
