dpif-linux.c makes a nl_lookup_genl_mcgroup(OVS_VPORT_FAMILY) that is not
implemented yet on Windows yet. Multicast group is used currently to
subscribe to events related to VPORTs. Whether the exact same mechanism
would be used is unclear yet.
In the interim, we can implement code to support nl_l
While we work out whether nl_sock_join_mcgroup() will be the mechanism
to support VPORT events, it is easy to add support for
nl_lookup_genl_mcgroup() and make progress on the other commands.
In this patch, we implement support for nl_lookup_genl_mcgroup() only
for the VPORT family though, which i
Hi Alin,
Thanks for working on it! I am almost done reviewing the change. I have a
few comments below.
>ovs/ovs-issues#15
>
>All NET_BUFFERs of a NET_BUFFER_LIST must go through
>the pipeline: extract, find flow, execute. Previously,
>only the first NET_BUFFER of a NET_BUFFER_LIST was going
>thro
2014-08-22 5:56 GMT-07:00 Jiri Pirko :
> Fri, Aug 22, 2014 at 02:42:04PM CEST, j...@mojatatu.com wrote:
>>On 08/21/14 13:05, Florian Fainelli wrote:
>>>2014-08-21 9:18 GMT-07:00 Jiri Pirko :
The goal of this is to provide a possibility to suport various switch
chips. Drivers should implemen
On Fri, Aug 15, 2014 at 10:28 AM, Ben Pfaff wrote:
>
>
> diff --git a/NEWS b/NEWS
> index 1ea7bda..656f750 100644
> --- a/NEWS
> +++ b/NEWS
> @@ -17,6 +17,9 @@ Post-v2.3.0
> * OpenFlow 1.5 (draft) Copy-Field action is now supported.
> * OpenFlow 1.3+ table features requests are now su
On Aug 22, 2014, at 12:14 PM, John Fastabend wrote:
> On 08/22/2014 05:56 AM, Jiri Pirko wrote:
>> Fri, Aug 22, 2014 at 02:42:04PM CEST, j...@mojatatu.com wrote:
>>> On 08/21/14 13:05, Florian Fainelli wrote:
2014-08-21 9:18 GMT-07:00 Jiri Pirko :
> The goal of this is to provide a poss
On Aug 22, 2014, at 12:39 PM, John Fastabend wrote:
> On 08/21/2014 09:19 AM, Jiri Pirko wrote:
>> Benefit from the possibility to work with flows in switch devices and
>> use the swdev api to offload flow datapath.
>
> we should add a description here on the strategy being used.
>
> If I read
MAC learning, like flow entry learning via the "learn" action, should only
happen if a packet was actually received.
Signed-off-by: Ben Pfaff
---
ofproto/ofproto-dpif-xlate.c |6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofpro
xlate_push_stats() mostly does nothing if 'stats->n_packets' is 0. This
commit allows it to skip more complicated internal logic in that case.
The one case I see in which xlate_push_stats() does do something if
'stats->n_packets' is 0 is in the case of a cached fin_timeout translation.
That trans
Both existing callers calculated 'may_learn' as 'stats->n_packets > 0', so
it was redundant. Because xlate_push_stats() is now entirely a no-op if
'stats->n_packets' is 0, we can now delete the tests entirely from the
cases that previously only ran if 'may_learn'.
Signed-off-by: Ben Pfaff
---
o
OpenFlow packet and byte counters have always been something of an
approximation in Open vSwitch. First, they are updated only periodically.
Second, they can be misattributed because statistics collection does a
retranslation and gives the statistics to whichever OpenFlow flow or flows
they happen
Series pushed, thanks for the reviews!
Jarno
On Aug 22, 2014, at 11:15 AM, Ben Pfaff wrote:
> On Fri, Aug 15, 2014 at 04:38:46PM -0700, Jarno Rajahalme wrote:
>> "xlate_receive" did not tell much about what it is used for. We have
>> two users of it that only want the ofproto and the OF port
On Aug 22, 2014, at 11:15 AM, Ben Pfaff wrote:
> On Fri, Aug 15, 2014 at 04:38:46PM -0700, Jarno Rajahalme wrote:
>> "xlate_receive" did not tell much about what it is used for. We have
>> two users of it that only want the ofproto and the OF port number, use the
>> new xlate_lookup_ofproto() f
On Aug 22, 2014, at 1:58 PM, Jarno Rajahalme wrote:
> The default memory order for atomic operations is
> memory_order_seq_cst, which synchronizes the whole system, i.e., locks
> the memory bus, which is very expensive. memory_order_relaxed, on the
> other hand, does not impose any kind of memo
Neither 'miss_config', 'n_missed', nor 'n_matched' is used to
synchronize the state of any other variable, so we can use relaxed
atomic operations on them.
Signed-off-by: Jarno Rajahalme
---
ofproto/ofproto-dpif.c | 10 ++
ofproto/ofproto.c | 12 +++-
2 files changed, 13
Signed-off-by: Jarno Rajahalme
---
lib/timeval.c |6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/timeval.c b/lib/timeval.c
index 66732b6..9d90f1b 100644
--- a/lib/timeval.c
+++ b/lib/timeval.c
@@ -147,7 +147,7 @@ time_timespec__(struct clock *c, struct timespec *t
'miimon_cnt' and the actual device miimon configuration is only
loosely coupled, so we can use the relaxed atomic_count for it.
Signed-off-by: Jarno Rajahalme
---
lib/netdev-linux.c | 21 +
1 file changed, 9 insertions(+), 12 deletions(-)
diff --git a/lib/netdev-linux.c b/
'dump->status' does not syncronize the state of any other variable, so
we can use relaxed atomics on it.
Signed-off-by: Jarno Rajahalme
---
lib/dpif-linux.c |7 ---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c
index e5067ba..2c387ed 10
'netflow_count' and the existence of actual netflow objects is not
tightly synchronized, so we can use the relaxed atomic_count for it.
Signed-off-by: Jarno Rajahalme
---
ofproto/netflow.c | 19 ---
1 file changed, 8 insertions(+), 11 deletions(-)
diff --git a/ofproto/netflow.
All access to struct netdev_registered_class ref_cnt member was done
with netdev_class_mutex held, so it does not need to be an atomic
variable.
Signed-off-by: Jarno Rajahalme
---
lib/netdev.c | 24 ++--
1 file changed, 10 insertions(+), 14 deletions(-)
diff --git a/lib/ne
Neither 'enable_megaflows', 'udpif->flow_limit', 'udpif->n_flows', nor
'udpif->n_flows_timestamp' are used to synchronize the state of any
other variables, so we can use relaxed atomic operations to access
them.
Move the atomic read operation of 'enable_megaflows' outside the loop
in handle_upcall
Even though there is no need to optimize netdev-dummy, it might be
good to do this right, in case it serves as an inspiration for
something else later.
Signed-off-by: Jarno Rajahalme
---
lib/netdev-dummy.c |4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/lib/netdev-dumm
Signed-off-by: Jarno Rajahalme
---
lib/ovs-atomic.h | 57 ++
1 file changed, 57 insertions(+)
diff --git a/lib/ovs-atomic.h b/lib/ovs-atomic.h
index b1c7b22..0f52c5f 100644
--- a/lib/ovs-atomic.h
+++ b/lib/ovs-atomic.h
@@ -412,6 +412,63 @@ ty
The atomics here do not synchronize the state of any other variables,
so we can use relaxed atomics.
cfm_should_process_flow() is rearranged to set the megaflow mask bits
only if necessary, and to avoid the atomic operation on non-BFD
packets.
Signed-off-by: Jarno Rajahalme
---
lib/cfm.c | 42
ovs_refcount_unref() needs to syncronize with the other instances of
itself rather than with ovs_refcount_ref().
Signed-off-by: Jarno Rajahalme
---
lib/ovs-atomic.h |8 +---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/lib/ovs-atomic.h b/lib/ovs-atomic.h
index 8e83411..9
barrier->count is used as a simple counter and is not expected the
synchronize the state of any other variable, so we can use atomic_count,
which uses relaxed atomics.
Ditto for the 'next_id' within ovsthread_wrapper().
Signed-off-by: Jarno Rajahalme
---
lib/ovs-thread.c | 10 +-
lib/
Avoiding the atomic read may help if a function using
ovsthread_once_start() is ever called in a loop, as the new
'maybe_not_done' can be kept in a register. The atomic read will
still be done as long as 'maybe_not_done' is true. Since
'maybe_not_done' is not an atomic variable, whis may happen
i
Trivial ID counters do not synchronize anything, therefore can use
atomic_count.
Signed-off-by: Jarno Rajahalme
---
lib/jsonrpc.c |4 ++--
lib/ofp-msgs.c |6 ++
lib/ofp-parse.c |4 ++--
3 files changed, 6 insertions(+), 8 deletions(-)
diff --git a/lib/jsonrpc.c b/lib/jsonrpc.
The default memory order for atomic operations is
memory_order_seq_cst, which synchronizes the whole system, i.e., locks
the memory bus, which is very expensive. memory_order_relaxed, on the
other hand, does not impose any kind of memory barrier (not even a
compiler barrier), which allows for bett
Otherwise the dereference operator could target a portion of a ternary
expression, for example.
Also minor style fixes.
Signed-off-by: Jarno Rajahalme
---
lib/ovs-atomic-gcc4+.h |8
lib/ovs-atomic-i586.h | 32
lib/ovs-atomic-x86_64.h | 14 +++
When an atomic variable is not serving to synchronize threads about
the state of other (atomic or non-atomic) variables, no memory barrier
is needed with the atomic operation. However, the default memory
order for an atomic operation is memory_order_seq_cst, which always
causes a system-wide locki
The atomics here do not synchronize the state of any other variables,
so we can use atomic_count and relaxed atomics.
bfd_should_process_flow() is rearranged to set the megaflow mask bits
only if necessary, and to avoid the atomic operation on non-BFD
packets.
Signed-off-by: Jarno Rajahalme
---
Signed-off-by: Jarno Rajahalme
---
lib/dpif-netdev.c |4 ++--
lib/ovs-thread.h |2 +-
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index 7401293..17da3dd 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -617,7 +617,7 @@ dp_
Hi Saurabh,
I took i am fine with the change as an ack.
I'll wait for explicit ack next time.
Regads,
Ankur
From: Saurabh Shah
Sent: Friday, August 22, 2014 1:35 PM
To: Ankur Sharma; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH 1/3 v2] datapath-window
Hey Ankur,
Something to be careful of in future. You should wait for the reviewers to
give an actual 'Acked-by' for a patch. If its a private Ack, it would be a
good idea to mention it and then put an Ack for them. For example, Nithin
& Samuel reviewed your V1 patch and gave comments, but that doe
On Thu, Aug 21, 2014 at 05:42:01PM +1200, Joe Stringer wrote:
> Future patches will make use of the 'struct dump_op' in a broader sense,
> so this patch renames it to make things a bit clearer.
>
> Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
___
d
On Thu, Aug 21, 2014 at 05:41:59PM +1200, Joe Stringer wrote:
> Currently, ukeys are protected during revalidator_sweep__() as only one
> thread accesses the ukey at a time. This is ensured using barriers:
> all revalidators will be in the GC phase, so they will only access their
> own ukey collect
On Thu, Aug 21, 2014 at 05:42:00PM +1200, Joe Stringer wrote:
> An upcoming patch will change the access patterns for ukey maps to
> increase the number of writers, and shift write-access from revalidator
> threads to upcall handler threads. As such, it no longer makes sense to
> tie these maps to
On 08/21/2014 09:19 AM, Jiri Pirko wrote:
Benefit from the possibility to work with flows in switch devices and
use the swdev api to offload flow datapath.
we should add a description here on the strategy being used.
If I read this correctly this will try to add any flow to the
hardware along
From: Andreea-Cristina Bernat
Date: Sun, 17 Aug 2014 16:29:43 +0300
> The "rcu_dereference()" call is used directly in a condition.
> Since its return value is never dereferenced it is recommended to use
> "rcu_access_pointer()" instead of "rcu_dereference()".
> Therefore, this patch makes the re
On 08/22/2014 05:56 AM, Jiri Pirko wrote:
Fri, Aug 22, 2014 at 02:42:04PM CEST, j...@mojatatu.com wrote:
On 08/21/14 13:05, Florian Fainelli wrote:
2014-08-21 9:18 GMT-07:00 Jiri Pirko :
The goal of this is to provide a possibility to suport various switch
chips. Drivers should implement relev
On 08/21/2014 09:18 AM, Jiri Pirko wrote:
The netdevice represents a port in a switch, it will expose
IFLA_PHYS_SWITCH_ID value via rtnl. Two netdevices with the same value
belong to one physical switch.
Signed-off-by: Jiri Pirko
What is the relation between phys_port_id and phys_switch_id?
On Thu, Aug 21, 2014 at 05:41:58PM +1200, Joe Stringer wrote:
> Signed-off-by: Joe Stringer
As of this patch, there isn't an advantage to using a cmap yet, right?
Here, I'd be inclined to add the ovsrcu_quiesce() call outside the
'if' statement, unconditionally, because otherwise it's just a mat
On Thu, Aug 21, 2014 at 05:41:57PM +1200, Joe Stringer wrote:
> Signed-off-by: Joe Stringer
My reading of the code is that this is redundant because
dpif_operate() always fills in the flow.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org
On Thu, Aug 21, 2014 at 05:41:56PM +1200, Joe Stringer wrote:
> These tests had the potential to fail due to statistics not updating
> before the test script retrieves them. Fix them by waiting until the
> next revalidation cycle.
>
> Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
From: Jiri Benc
Date: Thu, 21 Aug 2014 21:33:44 +0200
> When there are multiple vlan headers present in a received frame, the first
> one is put into vlan_tci and protocol is set to ETH_P_8021Q. Anything in the
> skb beyond the VLAN TPID may be still non-linear, including the inner TCI
> and ethe
On Fri, Aug 15, 2014 at 04:38:49PM -0700, Jarno Rajahalme wrote:
> Seems better as we no longer modify the flow.
>
> Signed-off-by: Jarno Rajahalme
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/de
On Fri, Aug 15, 2014 at 04:38:48PM -0700, Jarno Rajahalme wrote:
> xlate configuration (via xcfg) is already holding the references, so
> the callers of xlate_receive() can safely use the returned pointers
> upto the point when they quiesce. If the objects are needed for
> longer term, the callers
On Fri, Aug 15, 2014 at 04:38:47PM -0700, Jarno Rajahalme wrote:
> The netflow member was not used for anything, so it can be removed.
>
> Signed-off-by: Jarno Rajahalme
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.o
On Fri, Aug 15, 2014 at 04:38:46PM -0700, Jarno Rajahalme wrote:
> "xlate_receive" did not tell much about what it is used for. We have
> two users of it that only want the ofproto and the OF port number, use the
> new xlate_lookup_ofproto() for those.
>
> Fix the comments of xlate_receive() as w
Applied, thanks!
On Wed, Aug 20, 2014 at 01:38:45PM +, Saurabh Shah wrote:
>
> Acked-by: Saurabh Shah
>
> From: Nithin Raju
> Date: Thursday, August 14, 2014 at 9:02 PM
> To: "dev@openvswitch.org"
> Subject: [ovs-dev] [PATCH v3] datapath-windows: Use
> OVS_DEFAULT_PORT_NO instead of
On Fri, Aug 22, 2014 at 10:00:29AM -0700, Ankur Sharma wrote:
> In this patch we remove reference to OvsNetlink.h.
> Since we do not refer to lib/netlink-protocol.h anymore,
> hence removed the WIN_DP based check as well.
>
> Signed-off-by: Ankur Sharma
> Reported-at: https://github.com/openvswit
On Fri, Aug 22, 2014 at 10:00:13AM -0700, Ankur Sharma wrote:
> In this patch we incorporate the usage of netlink message
> and the parsing APIs that were added in previous commit.
>
> Signed-off-by: Ankur Sharma
> Reported-at: https://github.com/openvswitch/ovs-issues/issues/18
> Acked-by: Nithi
On Thu, 2014-08-21 at 18:18 +0200, Jiri Pirko wrote:
> The goal of this is to provide a possibility to suport various switch
> chips. Drivers should implement relevant ndos to do so. Now there is a
> couple of ndos defines:
> - for getting physical switch id is in place.
> - for work with flows.
>
On Thu, 2014-08-21 at 18:19 +0200, Jiri Pirko wrote:
[...]
> --- a/net/dsa/slave.c
> +++ b/net/dsa/slave.c
> @@ -171,6 +171,19 @@ static int dsa_slave_ioctl(struct net_device *dev,
> struct ifreq *ifr, int cmd)
> return -EOPNOTSUPP;
> }
>
> +static int dsa_slave_swdev_get_id(struct net_de
On Fri, Aug 22, 2014 at 01:16:44AM -0700, Nithin Raju wrote:
> In this patch, we add support for querying the genl family id for any
> family supported by the OVS kernel datapath. On platforms that support
> netlink natively, the operating system assigns a family ID, and the
> OS netlink infrastruc
In this change we introduce Netlink.c, Netlink.h and NetlinkProto.h
in datapath-windows. These files will provide netlink message
data structures and parsing APIs.
Changes are on similar lines to userspace netlink code.
We have mapped the userspace APIs and data structures to
kernel coding conve
On Fri, Aug 22, 2014 at 04:27:01PM +, Samuel Ghinet wrote:
> [QUOTE]
> OVS 1.9.3 needed a trip to userspace for every new microflow. But OVS
> 1.11 introduced megaflows, so that in many cases that is no longer
> necessary.
> [/QUOTE]
> Yeah, the microflow is the kernel flow I was trying to say
In this patch we remove reference to OvsNetlink.h.
Since we do not refer to lib/netlink-protocol.h anymore,
hence removed the WIN_DP based check as well.
Signed-off-by: Ankur Sharma
Reported-at: https://github.com/openvswitch/ovs-issues/issues/18
Acked-by: Nithin Raju
---
datapath-windows/autom
In this patch we incorporate the usage of netlink message
and the parsing APIs that were added in previous commit.
Signed-off-by: Ankur Sharma
Reported-at: https://github.com/openvswitch/ovs-issues/issues/18
Acked-by: Nithin Raju
---
datapath-windows/automake.mk | 5 ++-
datapath-win
[QUOTE]
OVS 1.9.3 needed a trip to userspace for every new microflow. But OVS
1.11 introduced megaflows, so that in many cases that is no longer
necessary.
[/QUOTE]
Yeah, the microflow is the kernel flow I was trying to say.
We can get to have quite many microflows created, for all kinds of cases
Nithin,
Regarding
"is it required to update many of the functions to process an NB instead of an
NBL"
I could try to preserve NBL wherever it was a NBL, even though the
functionality requires an NB only.
The worry I have here is, having the parameter "NET_BUFFER_LIST" gives no
assurance that it
Hi Sam,
Thanks for the clarification about the need to update the pipeline to handle
NBs. I'll review the patch and get back. I don't think it is necessary to spin
up a v2 patch. You can do that with any review comments I have.
I'll get back early next week. I'm on vacation today. Just wanted to
On Fri, Aug 22, 2014 at 03:24:47PM +, Samuel Ghinet wrote:
> What I was testing against was ovs 1.9.3 and ovs 1.11
OVS 1.9.3 needed a trip to userspace for every new microflow. But OVS
1.11 introduced megaflows, so that in many cases that is no longer
necessary.
> The fact is that, even for
On Fri, Aug 22, 2014 at 09:19:41PM +0900, Simon Horman wrote:
> I have been working with Netronome on examining the possibilities of
> providing (richer) load balancing facilities in Open vSwitch.
>
> It seems to us that the current select group provides for some load
> balancing functionality. An
What I was testing against was ovs 1.9.3 and ovs 1.11
The fact is that, even for the latest version of ovs, there is no kernel port
type = NORMAL or FLOOD (see enum ovs_vport_type). OFPP_NORMAL and OFPP_FLOOD
exist in userspace only.
Therefore, you cannot have exactly the same handling of a flow
On Fri, Aug 22, 2014 at 05:17:32PM +0530, Saloni Jain wrote:
> Kindly validate following approach for implementation of eviction
> mechanism and vacancy events as specified in openflow specification
> 1.4. According to specification eviction and vacancy events are set
> through table_mod messages.
Ankur,
About 1: From my side, it's ok.
And if you / we will do it in a later commit, perhaps you / we can rename the
userspace enum as well? (enum: nl_attr_type -> nl_attr_policy_type)
Sam
From: Ankur Sharma [ankursha...@vmware.com]
Sent: Friday, August 2
This patch implements the vhost-net offload API. It adds support for
a new port type to userspace datapath called dpdkvhost. This allows KVM
(QEMU) to offload the servicing of virtio-net devices to it's associated
dpdkvhost port. Instructions for use are in INSTALL.DPDK.
This has been tested on I
Hi Jarno,
There's been a few iterations. The previous 64-bit version was using a
32-bit hash combined with a thread ID.
The 128-bit hash was based on a suggestion to have sufficient bits to
randomly generate the identifiers, however I've since shifted to using the
hash of the flow key---This allo
Fri, Aug 22, 2014 at 02:42:04PM CEST, j...@mojatatu.com wrote:
>On 08/21/14 13:05, Florian Fainelli wrote:
>>2014-08-21 9:18 GMT-07:00 Jiri Pirko :
>>>The goal of this is to provide a possibility to suport various switch
>>>chips. Drivers should implement relevant ndos to do so. Now there is a
>>>c
On 08/21/14 13:05, Florian Fainelli wrote:
2014-08-21 9:18 GMT-07:00 Jiri Pirko :
The goal of this is to provide a possibility to suport various switch
chips. Drivers should implement relevant ndos to do so. Now there is a
couple of ndos defines:
- for getting physical switch id is in place.
- f
Hi,
I have been working with Netronome on examining the possibilities of
providing (richer) load balancing facilities in Open vSwitch.
It seems to us that the current select group provides for some load
balancing functionality. And that in particular the way that it is
implemented in Open vSwitch
Thanks Ben for your reply.
Dear Ben/Team,
Kindly validate following approach for implementation of eviction mechanism and
vacancy events as specified in openflow specification 1.4.
According to specification eviction and vacancy events are set through
table_mod messages.
Based on this scene
From: Florian Fainelli
> 2014-08-21 10:06 GMT-07:00 Jiri Pirko :
> > Thu, Aug 21, 2014 at 06:56:13PM CEST, f.faine...@gmail.com wrote:
> >>2014-08-21 9:19 GMT-07:00 Jiri Pirko :
> >>> Signed-off-by: Jiri Pirko
> >>> ---
> >>> net/dsa/Kconfig | 2 +-
> >>> net/dsa/slave.c | 16
>
We need a version number for the datapath interface extensions that
we can leverage at some point.
Signed-off-by: Nithin Raju
---
datapath-windows/include/OvsDpInterfaceExt.h |3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/datapath-windows/include/OvsDpInterfaceExt.h
In this patch, we add support for querying the genl family id for any
family supported by the OVS kernel datapath. On platforms that support
netlink natively, the operating system assigns a family ID, and the
OS netlink infrastructure supports querying the family ID by name.
In case of Windows, si
77 matches
Mail list logo
