On Fri, Feb 19, 2016 at 02:35:40PM -0800, Jarno Rajahalme wrote:
> With the comments below:
>
> Acked-by: Jarno Rajahalme
Thanks for the review.
> > +/* Properties for NXAST_CONTROLLER2. */
> > +enum nx_action_controller2_prop_type {
> > +NXAC2PT_MAX_LEN,/* ovs_be16 max length t
With small comments below:
Acked-by: Jarno Rajahalme
> On Feb 19, 2016, at 12:34 AM, Ben Pfaff wrote:
>
> One purpose of OpenFlow packet-in messages is to allow a controller to
> interpose on the path of a packet through the flow tables. If, for
> example, the controller needs to modify a pac
Acked-by: Jarno Rajahalme
> On Feb 19, 2016, at 12:34 AM, Ben Pfaff wrote:
>
> Signed-off-by: Ben Pfaff
> ---
> ovn/controller/pinctrl.h | 7 +++
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/ovn/controller/pinctrl.h b/ovn/controller/pinctrl.h
> index 65d5dfe..fd279ff
On Fri, Feb 19, 2016 at 02:47:59PM -0800, Jarno Rajahalme wrote:
> With one comment below:
>
> Acked-by: Jarno Rajahalme
Thanks for the review.
> > On Feb 19, 2016, at 12:34 AM, Ben Pfaff wrote:
> >
> > It hadn't occurred to me before that any special support was actually
> > necessary or use
This macro is not required as we drop support for unsupported
kernel versions.
Signed-off-by: Pravin B Shelar
---
acinclude.m4 | 9 -
datapath/compat.h| 5 -
datapath/linux/compat/include/linux/netfilter_ip
Currently OVS out of tree datapath supports a large number of kernel
versions. From 2.6.32 to 4.3 and various distribution-specific
kernels. But at this point major features are only available on more
recent kernels. For example, stateful services are only available
starting in kernel 3.10 and STT
On Fri, Feb 19, 2016 at 03:43:35PM -0800, Jarno Rajahalme wrote:
> With small comments below:
>
> Acked-by: Jarno Rajahalme
Thanks for the review!
I fixed the typos you pointed out.
> > + * The controller might change the pipeline configuration concurrently with
> > + * steps 2 through 4. For
Hi Sorin,
Thanks for the patch. It needs to be updated to make it thread-safe. The
DeferredActionQueue needs a RW lock for synchronization.
I also ran into memory leak while testing out the patch (the kernel
couldn¹t be uninstalled without a reboot).
If am able to get to it, I will send out a pat
On Fri, Feb 19, 2016 at 03:00:47PM -0800, Joe Stringer wrote:
> On 19 February 2016 at 00:34, Ben Pfaff wrote:
> > An upcoming commit will add another case where it's desirable to ensure
> > that a variable-length array is aligned on an 8-byte boundary. This macro
> > makes that a little easier.
I’ve reviewed the series unto this point. Maybe someone else will review the
OVN patches. I did note, however, that the last patch does not help the test
failure:
Before the last patch:
# put_arp
-actions=push:NXM_NX_REG1[],push:NXM_OF_ETH_SRC[],push:NXM_NX_ARP_SHA[],push:NXM_OF_ARP_SPA[],pop:
> On Feb 19, 2016, at 3:56 PM, Ben Pfaff wrote:
>
> On Fri, Feb 19, 2016 at 02:47:59PM -0800, Jarno Rajahalme wrote:
>> With one comment below:
>>
>> Acked-by: Jarno Rajahalme
>
> Thanks for the review.
>
>>> On Feb 19, 2016, at 12:34 AM, Ben Pfaff wrote:
>>>
>>> It hadn't occurred to me b
Oh, that's a completely different test failure. I'm sure it's less
mysterious. Probably I introduced it somewhere in revision today.
I see it now too, I'll fix it.
The test failure I had a problem with is:
1751: ovn -- 3 HVs, 3 LS, 3 lports/LS, 1 LR
On Fri, Feb 19, 2016 at 04:17:42PM -080
On Fri, Feb 19, 2016 at 04:24:30PM -0800, Ben Pfaff wrote:
> Oh, that's a completely different test failure. I'm sure it's less
> mysterious. Probably I introduced it somewhere in revision today.
>
> I see it now too, I'll fix it.
It was really that I forgot to update the test after changing ho
This series is also available at:
https://github.com/blp/ovs-reviews/tree/l3-4
g...@github.com:blp/ovs-reviews.git l3-4
This is the fifth revision of my "closures" series.
The first version was:
http://openvswitch.org/pipermail/dev/2016-January/064607.html
The second version, unf
An upcoming commit will use this as a building block in adding ARP support
to the OVN L3 logical router implementation.
Signed-off-by: Ben Pfaff
---
lib/ofp-actions.h | 30 +++
lib/packets.c | 38 +
lib/packets.h | 3 +-
ovn/contr
This was more or less implemented inside lflow.c until now, but some
upcoming code that shouldn't be in that file needs to use it too.
This also adds a second index on lports, so that lports can be looked up
based on the logical datapath tunnel key and the logical port tunnel key.
An upcoming comm
I don't understand why, but without this change, the test in the previous
commit does not pass.
Signed-off-by: Ben Pfaff
---
ovn/controller/lflow.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/ovn/controller/lflow.c b/ovn/controller/lflow.c
index f16c5c9..a18c760 100644
--
This is sufficient support that an L3 logical router can now transmit
packets to VMs (and other destinations) without having to know the
IP-to-MAC binding in advance. The details are carefully documented in all
of the appropriate places.
There are several important caveats that need to be fixed b
An simap is convenient but it isn't very flexible. If the client wants to
keep extra data with each node then it has to build a second parallel data
structure. A callback function is kind of a pain for the clients from the
point of view of having to write it and deal with auxiliary data, etc., bu
On Mon, Feb 08, 2016 at 06:54:54PM -0800, Andy Zhou wrote:
> On Mon, Feb 8, 2016 at 4:52 PM, Ben Pfaff wrote:
>
> > Some of our data structures derived from hmap use the same member names.
> > This means it's possible to confuse them in iteration, e.g. to iterate a
> > shash with SIMAP_FOR_EACH.
On Tue, Feb 16, 2016 at 08:38:52PM +, Zoltán Balogh wrote:
> I agree, it's better to use the fixed value and not to depend on the 'stat'.
> That's why the 'if' condition checks if 'stat' is available. But your
> suggested patch is simpler and better.
I think I like simple.
Thanks for all t
96
p{margin:10px 0;padding:0;}
table{border-collapse:collapse;}
h1,h2,h3,h4,h5,h6{display:block;margin:0;padding:0;}
img,a img{border:0;height:auto;outline:none;text-decoration:none;}
body,#bodyTable,#bodyCell{height:100%;margin:0;padding:0;width:100%;}
#outlook a{padding:0;}
img{-ms
Hi Jarno. Among the oldest patches in patchwork are a couple from Simon
that add some mew miniflow related functions:
https://patchwork.ozlabs.org/patch/570523/
https://patchwork.ozlabs.org/patch/570522/
I think you're the best one to review these, will you have a look?
Thanks,
On Mon, Jan 25, 2016 at 10:43:16AM -0500, Aaron Conole wrote:
> Mauricio Vasquez B writes:
> > A ring name length of 10 characters is not enough for dpdkr ports
> > starting from dpdkr10, then it is increased to RTE_RING_NAMESIZE
> > characters.
> >
> > Signed-off-by: Mauricio Vasquez B
> >
> >
Daniele, are you the right person to review and apply this pair of
patches?
Thanks,
Ben.
On Mon, Jan 25, 2016 at 02:55:57PM -0500, Mauricio Vasquez B wrote:
> Current implementation of dpdk_dev_parse_name does not perform a robust
> error handling, port names as "dpdkr", "dpdkr1x", "dpdkr 5" are
On Fri, Jan 29, 2016 at 09:27:33PM +0530, Numan Siddique wrote:
> If a logical port has two ipv4 addresses and one ipv6 address
> it will be stored as ["MAC IPv41 IPv42 IPv61"] instead of
> ["MAC IPv41", "MAC IPv42", "MAC IPv61"].
>
> Signed-off-by: Numan Siddique
This appears at first to add a
On Fri, Jan 29, 2016 at 09:28:21PM +0530, Numan Siddique wrote:
> From: Ben Pfaff
>
> Signed-off-by: Numan Siddique
It's kind of odd to add this separate from implementing it. Usually, we
add code and its documentation in the same commit.
___
dev mai
On Fri, Jan 29, 2016 at 09:28:54PM +0530, Numan Siddique wrote:
> For every port security defined for a logical port, add following lflows
> in "ls_in_port_sec" and "ls_out_port_sec" stage
>- A priority 90 flow to allow ipv4 traffic for known ip addresses
> and (broadcast ip - for ingress,
On Fri, Jan 29, 2016 at 09:29:22PM +0530, Numan Siddique wrote:
> For each lport, adds a priority 90 lflow in ls_in_port_sec and ls_out_port_sec
> stages to allow ipv6 traffic for
> - known ipv6 addresses
> - link local address of the lport
> - ip6 packet with ip6.src = :: and
> - ip6.dst=ff00:
On 20.02.2016 00:03, Daniele Di Proietto wrote:
> Would it be safe to assume that the enabled queues are sequential?
> In this case we could just play with 'real_n_txq' instead of keeping
> a mapping and the patch would be simpler. I'm not sure if that's a
> reasonable assumption though.
I thought
On Fri, Feb 19, 2016 at 02:30:31PM -0500, Lance Richardson wrote:
> Add make targets for Fedora and RHEL7 RPMs, update INSTALL.Fedora.md
> to document their use
>
> Added distribution tarball and rpm build directory to .gitignore.
>
> Signed-off-by: Lance Richardson
This seems reasonable to me.
On Tue, Feb 16, 2016 at 02:53:39PM -0800, Jesse Gross wrote:
> On Fri, Feb 12, 2016 at 11:25 AM, Simon Horman
> wrote:
> > If an skb was not MPLS initially then it may be GSO and in that case if it
> > became MPLS then GSO can't be performed because both MPLS and tunnels make
> > use of the inner_
On Tue, Feb 16, 2016 at 11:15:20PM +0100, Jiri Benc wrote:
> Sorry for the late answer, was busy with a conference and internal
> meetings in the past two weeks.
>
> On Tue, 2 Feb 2016 17:15:15 -0800, Jesse Gross wrote:
> > I think this sounds like a good idea if we can find a way to do it
> > cle
On Mon, Dec 21, 2015 at 03:25:42PM -0800, Justin Pettit wrote:
>
> > On Dec 8, 2015, at 5:08 PM, Ben Pfaff wrote:
> >
> > The existing wording was very specific to the actual operation being
> > performed. While this is nice for users, it becomes difficult to maintain
> > as more and more opera
On Mon, Dec 21, 2015 at 03:26:51PM -0800, Justin Pettit wrote:
>
> > On Dec 8, 2015, at 5:08 PM, Ben Pfaff wrote:
> >
> > It seems easier to understand if all of the tests for a given action
> > are grouped together.
> >
> > Signed-off-by: Ben Pfaff
>
> Acked-by: Justin Pettit
Thanks, appli
An upcoming commit will introduce another member that has a length, and
it seems weird that bare 'len' would be one or the other.
Signed-off-by: Ben Pfaff
---
lib/learning-switch.c| 2 +-
lib/ofp-print.c | 10 +-
lib/ofp-util.c | 42 +++
It's not const in all cases so it doesn't entirely make sense to mark
it const here.
Signed-off-by: Ben Pfaff
---
lib/ofp-util.c| 2 +-
lib/ofp-util.h| 2 +-
ofproto/connmgr.c | 2 +-
3 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/ofp-util.c b/lib/ofp-util.c
index 23f
A number of times I've looked at code and thought that it would be easier
to understand if I could write an initializer instead of
ofpbuf_use_const(). This commit adds a function for that purpose and
adapts a lot of code to use it, in the places where I thought it made
the code better.
In theory
This series is also available at:
https://github.com/blp/ovs-reviews/tree/l3-4
g...@github.com:blp/ovs-reviews.git l3-4
This is the fourth revision of my "closures" series.
The first version was:
http://openvswitch.org/pipermail/dev/2016-January/064607.html
The second version, un
Signed-off-by: Ben Pfaff
---
NEWS | 1 +
include/openflow/nicira-ext.h | 43 -
lib/ofp-msgs.h| 3 +
lib/ofp-util.c| 214 ++
ofproto/connmgr.c | 4 +-
ofproto/ofproto.c
It hadn't occurred to me before that any special support was actually
necessary or useful for nested properties, but the functions introduced in
this commit are nice wrappers to deal with the extra 4-byte padding that
ensures that the nested properties begin on 8-byte boundaries just like
the outer
Signed-off-by: Ben Pfaff
---
ovn/controller/pinctrl.h | 7 +++
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/ovn/controller/pinctrl.h b/ovn/controller/pinctrl.h
index 65d5dfe..fd279ff 100644
--- a/ovn/controller/pinctrl.h
+++ b/ovn/controller/pinctrl.h
@@ -1,4 +1,3 @@
-
/* Co
An upcoming commit will use this as a building block in adding ARP support
to the OVN L3 logical router implementation.
Signed-off-by: Ben Pfaff
Conflicts:
tests/ovn.at
---
lib/ofp-actions.h | 30 +++
lib/packets.c | 38 +
lib/packets.h
An upcoming commit will add another case where it's desirable to ensure
that a variable-length array is aligned on an 8-byte boundary. This macro
makes that a little easier.
Signed-off-by: Ben Pfaff
CC: Joe Stringer
---
lib/ofp-actions.h | 43 ++-
1 file
One purpose of OpenFlow packet-in messages is to allow a controller to
interpose on the path of a packet through the flow tables. If, for
example, the controller needs to modify a packet in some way that the
switch doesn't directly support, the controller should be able to
program the switch to se
An simap is convenient but it isn't very flexible. If the client wants to
keep extra data with each node then it has to build a second parallel data
structure. A callback function is kind of a pain for the clients from the
point of view of having to write it and deal with auxiliary data, etc., bu
Signed-off-by: Ben Pfaff
---
NEWS | 3 +-
include/openflow/nicira-ext.h | 1 +
lib/ofp-actions.c | 168 ++
lib/ofp-actions.h | 9 ++-
lib/ofp-print.c | 11 +++
lib/ofp-util.c
This was more or less implemented inside lflow.c until now, but some
upcoming code that shouldn't be in that file needs to use it too.
This also adds a second index on lports, so that lports can be looked up
based on the logical datapath tunnel key and the logical port tunnel key.
An upcoming comm
I don't understand why, but without this change, the test in the previous
commit does not pass.
Signed-off-by: Ben Pfaff
---
ovn/controller/lflow.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/ovn/controller/lflow.c b/ovn/controller/lflow.c
index f16c5c9..a18c760 100644
--
This is sufficient support that an L3 logical router can now transmit
packets to VMs (and other destinations) without having to know the
IP-to-MAC binding in advance. The details are carefully documented in all
of the appropriate places.
There are several important caveats that need to be fixed b
> On Wed, 17 Feb 2016 14:40:26 +
> "Stokes, Ian" wrote:
>
> > > -Original Message-
> > > From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Stokes,
> > > Ian
> > > Sent: Thursday, February 11, 2016 2:42 PM
> > > To: Flavio Leitner
> > > Cc: dev@openvswitch.org
> > > Subject:
Hi Joe,
Thanks for pointing out my mistake. Please find below updated one.
m 69e63a45e2773c124deb885bbc3d5deb3e032126 Mon Sep 17 00:00:00 2001
From: soumyadeep chowdhury
Date: Fri, 19 Feb 2016 07:39:34 -0500
Subject: [PATCH 4/4] Write the functions for dump-conntrack and
flush-conntrack
Signe
Current mbuf initialization relies on magic numbers and does not
accomodate mbufs of different sizes.
Resolve this issue by ensuring that mbufs are always aligned to a 1k
boundary (a typical DPDK NIC Rx buffer alignment).
Signed-off-by: Mark Kavanagh
---
lib/netdev-dpdk.c | 87 +
This patchset enables DPDK ports to support Jumbo frames
up to, and including 13K in size. Jumbo frame support is implemented
by increasing the amount of data that each mbuf can encompass, thus
allowing each mbuf segment to accomodate a single jumbo frame.
The series additionally cleans up mbuf in
Add support for Jumbo Frames to DPDK-enabled port types,
using single-segment-mbufs.
Using this approach, the amount of memory allocated for each mbuf
to store frame data is increased to a value greater than 1518B
(typical Ethernet maximum frame length). The increased space
available in the mbuf m
Hi Team,
When I am trying to call rte_eth_dev_start() function for kni interface it is
failing. Can you please let me know what are all the reasons responsible for
failing of rte_eth_dev_start() function.
Thanks & Regards,
Ravali
The information contained in this electronic message and any atta
Signed-Off-by: Numan Siddique
---
tutorial/OVN-Tutorial.md | 77
tutorial/automake.mk | 7 ++-
tutorial/ovn/env7/add-container-ports.sh | 60 +
tutorial/ovn/env7/packet1.sh | 19
tut
Good morning,
Please see the attached invoice and remit payment according to the terms listed
at the bottom of the invoice.
If you have any questions please let us know.
Thank you!
Anastasia Dotson
Accounting Specialist
___
dev mailing list
dev@openvs
All-
Having chatted with Ben on IRC late yesterday afternoon, I sat down and
made a list of events
that would require incremental processing to restart at sequence number 0
because they might
change how a Logical Flow row gets mapped to OF flows without the Logical
Flow changing.
Looking at the
On 02/17/2016 09:12 PM, Justin Pettit wrote:
>
>> On Feb 5, 2016, at 1:30 PM, Russell Bryant wrote:
>>
>>
>> Thank you for the write-up! This approach sounds great to me. Some
>> small questions...
>>
>> 1) If we're only using 1 bit for now, is there any reason to use
>> ct_label over ct_mark?
On 02/18/2016 07:37 PM, Joe Stringer wrote:
> On 17 February 2016 at 18:12, Justin Pettit wrote:
>>
>>> On Feb 5, 2016, at 1:30 PM, Russell Bryant wrote:
>>>
>>>
>>> Thank you for the write-up! This approach sounds great to me. Some
>>> small questions...
>>>
>>> 1) If we're only using 1 bit fo
On Thu, Feb 18, 2016 at 11:59 PM, Simon Horman
wrote:
> On Tue, Feb 16, 2016 at 02:53:39PM -0800, Jesse Gross wrote:
>> On Fri, Feb 12, 2016 at 11:25 AM, Simon Horman
>> wrote:
>> > If an skb was not MPLS initially then it may be GSO and in that case if it
>> > became MPLS then GSO can't be perfo
This patch provides the modifications required in netdev-dpdk.c and
vswitch.xml to allow for a DPDK user space QoS algorithm.
This patch adds a QoS configuration structure for netdev-dpdk and
expected QoS operations 'dpdk_qos_ops'. Various helper functions
are also supplied.
Also included are the
From: RYAN D. MOATS
This is a prerequisite for incremental processing.
Signed-off-by: RYAN D. MOATS
---
ovn/controller/ofctrl.c | 118 +++
ovn/controller/ofctrl.h |2 +
ovn/controller/ovn-controller.c |4 +-
3 files changed, 87 inser
From: RYAN D. MOATS
Currently changes are added to the front of the track list, so
they are looped through in LIFO order. Incremental processing
is more efficient with a FIFO presentation, so add changes to the
back of the track list.
---
lib/ovsdb-idl.c |6 +++---
1 files changed, 3 inserti
From: RYAN D. MOATS
Persist across runs so that a change to this simap can be used
as a trigger for resetting incremental processing.
Signed-off-by: RYAN D. MOATS
---
ovn/controller/lflow.c | 18 --
1 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/ovn/controlle
From: RYAN D. MOATS
Persist logical_datapaths across runs so that a change can be used
as a trigger to reset incremental flow processing.
Signed-off-by: RYAN D. MOATS
---
ovn/controller/ovn-controller.c | 14 +-
1 files changed, 1 insertions(+), 13 deletions(-)
diff --git a/ovn/
From: RYAN D. MOATS
This code changes lflow_run to do incremental process of the
logical flow table rather than processing the full table each run.
Signed-off-by: RYAN D. MOATS
---
ovn/controller/binding.c|2 +
ovn/controller/lflow.c | 53
From: RYAN D. MOATS
Make test 1737 output the OF flows from all three hypervisors to help
in case something goes wrong.
Signed-off-by: RYAN D. MOATS
---
tests/ovn.at | 12
1 files changed, 12 insertions(+), 0 deletions(-)
diff --git a/tests/ovn.at b/tests/ovn.at
index f4117b6..
From: RYAN D. MOATS
This series of patches changes lflow_run from processing the
entire southbound logical flow table to doing incremental
processing during each run.
To make this work, several pre-requisites are necessary, and
these are contained in the initial patches of the series.
RYAN D. M
On 19 February 2016 at 02:35, wrote:
> Hi Joe,
>
> Thanks for pointing out my mistake. Please find below updated one.
>
>
> m 69e63a45e2773c124deb885bbc3d5deb3e032126 Mon Sep 17 00:00:00 2001
> From: soumyadeep chowdhury
> Date: Fri, 19 Feb 2016 07:39:34 -0500
> Subject: [PATCH 4/4] Write the fu
From: David Wragg
Upstream commit:
Allow the MTU of vxlan devices without an underlying device to be set
to larger values (up to a maximum based on IP packet limits and vxlan
overhead).
Previously, their MTUs could not be set to higher than the
conventional ethernet value of
From: David Wragg
Upstream commit:
Prior to 4.3, openvswitch tunnel vports (vxlan, gre and geneve) could
transmit vxlan packets of any size, constrained only by the ability to
send out the resulting packets. 4.3 introduced netdevs corresponding
to tunnel vports. These netdevs ha
From: David Wragg
Upstream commit:
Allow the MTU of geneve devices to be set to large values, in order to
exploit underlying networks with larger frame sizes.
GENEVE does not have a fixed encapsulation overhead (an openvswitch
rule can add variable length options), so there is no
From: David Wragg
Upstream commit:
Calculate the maximum MTU taking into account the size of headers
involved in GENEVE encapsulation, as for other tunnel types.
Changes in v3:
- Correct comment style
Changes in v2:
- Conform more closely to ip_tunnel_change_mtu
- Exc
Since the commit e23775f20e1a ("datapath: Add support for lwtunnel"),
devices underlying each tunnel type have applied quite strict restrictions
on the MTU which can be set for the tunnel device, even if the underlying
physical device has no such restriction. This made it impossible to, for
instanc
Currently, even if the entire path supports jumbo frames, the LISP netdev
limits the path MTU to 1500 bytes, and cannot be configured otherwise.
Relax the constraints on modifying the device MTU, and set it to the
maximum by default.
Signed-off-by: Joe Stringer
---
v2: Fix device registration lea
Currently, even if the entire path supports jumbo frames, the STT netdev
limits the path MTU to 1500 bytes, and cannot be configured otherwise.
Relax the constraints on modifying the device MTU, and set it to the
maximum by default.
Signed-off-by: Joe Stringer
---
v2: Use dev->hard_header_len.
On 19 February 2016 at 10:44, Joe Stringer wrote:
> Currently, even if the entire path supports jumbo frames, the STT netdev
> limits the path MTU to 1500 bytes, and cannot be configured otherwise.
> Relax the constraints on modifying the device MTU, and set it to the
> maximum by default.
>
> Sig
The patch (for references, not fully tested yet) shows the preliminary
results of enabling RTM (Restricted Transactional Memory). A successful
transactional execution elides the lock, i.e., the lock is by-passed, and
exposes concurrency. However, transactions might abort due to several reasons
such
On Fri, Feb 19, 2016 at 10:44 AM, Joe Stringer wrote:
> Since the commit e23775f20e1a ("datapath: Add support for lwtunnel"),
> devices underlying each tunnel type have applied quite strict restrictions
> on the MTU which can be set for the tunnel device, even if the underlying
> physical device h
Add make targets for Fedora and RHEL7 RPMs, update INSTALL.Fedora.md
to document their use
Added distribution tarball and rpm build directory to .gitignore.
Signed-off-by: Lance Richardson
---
.gitignore| 2 +
INSTALL.Fedora.md | 197 +++
> From V3 removed any unnecessary code change and removed cloning of the
> running ovsdb configuration database. We should address SSL in a specific
> patch without depending on the default database.
>
> Signed-off-by: Michael Arnaldi
>
> ---
> NEWS | 4 ++
> ovn/northd/ovn
On Linux jsonrpc server now users poll group by default. It can
be disabled by using an undocumented --disable-epoll command line
options.
For ovsdb-server to maintain 1000 idle connections over TCP with the
default 5s probe interval, the CPU load dropped from 48% to 14%.
Signed-off-by: Andy Zhou
Introduce a new 'poll group' feature to improve the efficiency of poll
loop in dealing with large number of connections, and its first application
to improve ovsdb-server to scale up connections.
Patch 6/6 commit log as some performance numbers. They are copied here
for reference.
For ovsdb-se
---
lib/poll-loop.c | 11 +++
lib/poll-loop.h | 3 +++
2 files changed, 14 insertions(+)
diff --git a/lib/poll-loop.c b/lib/poll-loop.c
index e83d989..63c7a42 100644
--- a/lib/poll-loop.c
+++ b/lib/poll-loop.c
@@ -53,6 +53,7 @@ struct poll_loop {
* wake up immediately, or LLONG_MAX
Signed-off-by: Andy Zhou
---
lib/jsonrpc.c | 74 +++
lib/jsonrpc.h | 7 ++
2 files changed, 77 insertions(+), 4 deletions(-)
diff --git a/lib/jsonrpc.c b/lib/jsonrpc.c
index 35428a6..e8ea8b4 100644
--- a/lib/jsonrpc.c
+++ b/lib/jsonrpc
Poll group is a new poll class that sits between application and
the stream class. Poll group compliments the poll loop facility and the
stream class to make main loop more efficient when dealing with
large number of current connections.
See comments in poll-group-provider.h for more details.
Sig
On 19 February 2016 at 11:05, Jesse Gross wrote:
> On Fri, Feb 19, 2016 at 10:44 AM, Joe Stringer wrote:
>> Since the commit e23775f20e1a ("datapath: Add support for lwtunnel"),
>> devices underlying each tunnel type have applied quite strict restrictions
>> on the MTU which can be set for the tu
Add the first poll group implementation.
Signed-off-by: Andy Zhou
---
lib/automake.mk | 1 +
lib/poll-group-epoll.c| 321 ++
lib/poll-group-provider.h | 4 +
lib/poll-group.c | 9 +-
4 files changed, 332 insertions(+), 3 de
Add new APIs in stream class that works with poll group.
Signed-off-by: Andy Zhou
---
lib/stream-fd.c | 32
lib/stream-provider.h | 17 +++
lib/stream-ssl.c | 40 -
lib/stream-tcp.c | 3 ++
lib/stream-unix.c | 3 ++
lib/s
On Thu, Feb 18, 2016 at 8:20 PM, Andy Zhou wrote:
> This is the new 'poll group' feature I have been working on improve
> the efficiency in dealing with large number of connections, and its
> first application to improve ovsdb-sever.
>
> There is a known limitation:
>
> The probing feature does n
Thanks for the reviews. I applied these to master.
On Thu, Feb 18, 2016 at 10:47:06AM -0800, Jarno Rajahalme wrote:
> Acked-by: Jarno Rajahalme
>
> > On Jan 27, 2016, at 3:50 PM, Ben Pfaff wrote:
> >
> > This code was the only user for OpenFlow header definitions other than
> > struct ofp_hea
The patches are also available at github.
https://github.com/azhou-nicira/ovs-review/tree/poll-group
On Fri, Feb 19, 2016 at 12:40 PM, Andy Zhou wrote:
> Introduce a new 'poll group' feature to improve the efficiency of poll
> loop in dealing with large number of connections, and its first
On Mon, Feb 08, 2016 at 11:14:23AM -0800, Jarno Rajahalme wrote:
>
> > On Feb 5, 2016, at 7:27 PM, Ben Pfaff wrote:
> >
> > On Fri, Feb 05, 2016 at 04:41:29PM -0800, Jarno Rajahalme wrote:
> >>
> >>> On Feb 5, 2016, at 3:30 PM, Ben Pfaff wrote:
> >>>
> >>> Mirroring is supposed to happen at m
Would it be safe to assume that the enabled queues are sequential?
In this case we could just play with 'real_n_txq' instead of keeping
a mapping and the patch would be simpler. I'm not sure if that's a
reasonable assumption though.
Also, on my system, with qemu-2.5.0, vring_state_changed() is cal
warning: ‘OVS_BE128_MAX’ defined but not used [-Wunused-const-variable]
Found using CentOS 6.6 with gcc 6.6.0.
Signed-off-by: William Tu
---
include/openvswitch/types.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/include/openvswitch/types.h b/include/openvswitch/types.h
With a note and a question below:
Acked-by: Jarno Rajahalme
> On Feb 19, 2016, at 12:34 AM, Ben Pfaff wrote:
>
> A number of times I've looked at code and thought that it would be easier
> to understand if I could write an initializer instead of
> ofpbuf_use_const(). This commit adds a functi
On Fri, Feb 19, 2016 at 01:35:55PM -0800, William Tu wrote:
> warning: ‘OVS_BE128_MAX’ defined but not used [-Wunused-const-variable]
> Found using CentOS 6.6 with gcc 6.6.0.
Are you sure about that GCC version number?
___
dev mailing list
dev@openvswitc
Acked-by: Jarno Rajahalme
> On Feb 19, 2016, at 12:34 AM, Ben Pfaff wrote:
>
> It's not const in all cases so it doesn't entirely make sense to mark
> it const here.
>
> Signed-off-by: Ben Pfaff
> ---
> lib/ofp-util.c| 2 +-
> lib/ofp-util.h| 2 +-
> ofproto/connmgr.c | 2 +-
> 3 files c
1 - 100 of 114 matches
Mail list logo
