On 04/06/15 at 02:00pm, Gurucharan Shetty wrote:
> > I think lxc deployment inside the VM definitely makes sense from a
> > hardware isolation perspective but it may not be feasible for
> > everyone. A simple example is resource allocation across multiple
> > tenants on process level which becomes
On FreeBSD using AC_SEARCH_LIBS is not sufficient for finding the
linker flags for linking with a pthread library.
Add macro ax_pthread.m4 from GNU autoconf macro archive and use
macro AX_PTHREAD.
Signed-off-by: Kevin Lo
---
diff --git a/configure.ac b/configure.ac
index 8d47eb9..531a1f0 100644
We have to specify -w to set tunable sysctls on NetBSD.
Signed-off-by: Kevin Lo
---
diff --git a/INSTALL.userspace.md b/INSTALL.userspace.md
index 300d6c1..04c02cd 100644
--- a/INSTALL.userspace.md
+++ b/INSTALL.userspace.md
@@ -76,7 +76,7 @@ Other settings
On NetBSD, depending on your network
> We have to specify -w to set tunable sysctls on NetBSD.
>
> Signed-off-by: Kevin Lo
Acked-by: YAMAMOTO Takashi
> ---
>
> diff --git a/INSTALL.userspace.md b/INSTALL.userspace.md
> index 300d6c1..04c02cd 100644
> --- a/INSTALL.userspace.md
> +++ b/INSTALL.userspace.md
> @@ -76,7 +76,7 @@ Oth
I used to live in the gym Not anymore
I just want you to feel as good as I do.
http://x.co/8jbUo
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
You might start by looking at the Rocker switch implementation which creates
multiple netdev devices for each port of the switch. >> I will look at this
implementation.
In your case you'd
likely create these dummy netdev definitions within you PF driver. My
assumption is that ALL packets woul
Thomas Graf (3):
ovs-ctl: Unload & reload vport modules on force-reload-kmod
ovs-dev: Hint that vport modules may need to be loaded manually
NEWS: Leave a comment that vports are now separate kernel modules
NEWS | 3 +++
utilities/ovs-ctl.in | 22 --
uti
We manually rmmod the loaded vports as using modprobe -r
only works if the modules are available through modules.dep
We do not treat failures to load vports as a fatal error in case
the vport module has been renamed. Bringing the bridge back up is
considered more important. The error is still repo
In case the openvswitch.ko is loaded manually without dependencies
built, the vport modules need to be loaded manually as well.
Signed-off-by: Thomas Graf
---
utilities/ovs-dev.py | 3 +++
1 file changed, 3 insertions(+)
diff --git a/utilities/ovs-dev.py b/utilities/ovs-dev.py
index 9467df5..98
Signed-off-by: Thomas Graf
---
NEWS | 3 +++
1 file changed, 3 insertions(+)
diff --git a/NEWS b/NEWS
index 9f9dc4c..9714831 100644
--- a/NEWS
+++ b/NEWS
@@ -77,6 +77,9 @@ Post-v2.3.0
numbers. OpenFlow is 6653 and OVSDB is 6640.
- Support for DPDK vHost.
- Support for outer UDP ch
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Tue, Apr 07, 2015 at 04:18:47PM +0800, Kevin Lo wrote:
> We have to specify -w to set tunable sysctls on NetBSD.
>
> Signed-off-by: Kevin Lo
Applied, thanks!
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Hi,all
I know in normal vxlan support,vteps can be in different subnets.
But when I config the vxlan tunnel of openvswitch with the local ip and the
remote ip in different subnets,
I find the flow can not be matched.
tap1 of vm1(192.168.10.2)--br-tun in Host1(eth3.1124
10.10.24.69)==
This variable is already defined at the top of the function, so just
reuse that instead of defining it again.
Signed-off-by: Russell Bryant
---
ovn/ovn-nbd.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/ovn/ovn-nbd.c b/ovn/ovn-nbd.c
index 343be1d..5eb1ee7 100644
--- a/ov
v1->v2:
- Drop first patch that was merged.
- Update 1/3 to fix unsafe HMAP_FOR_EACH_WITH_HASH usage.
- Fix issues in pre-existing code to address the same
unsafe HMAP_FOR_EACH_WITH_HASH usage.
[PATCH 1/3] ovn-nbd: Match bindings on lport name.
[PATCH 2/3] ovn-nbd: Don't shadow hash_node var
When ovn-nbd is looking through bindings to figure out if any logical
ports in the northbound DB need to have their 'up' state changed, it
should be using the logical port's name instead of UUID for matching
up to rows in the OVN db.
The switch to using the name instead of UUID means we don't have
The previous code assumed that hash_node would be NULL when the loop
terminated without a match. That's not the case, so track the match a
little differently.
Signed-off-by: Russell Bryant
---
ovn/ovn-nbd.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/ovn/ovn-nbd.c b
On 04/07/2015 03:29 AM, Thomas Graf wrote:
> On 04/06/15 at 02:00pm, Gurucharan Shetty wrote:
>>> I think lxc deployment inside the VM definitely makes sense from a
>>> hardware isolation perspective but it may not be feasible for
>>> everyone. A simple example is resource allocation across multipl
> -Original Message-
> From: Pravin Shelar [mailto:pshe...@nicira.com]
> Sent: Friday, April 3, 2015 6:32 AM
> To: Traynor, Kevin
> Cc: Flavio Leitner; Rogers, Gerald; dev@openvswitch.org
> Subject: Re: [ovs-dev] [PATCH RFC 1/1] netdev-dpdk: add dpdk vhost-user ports
>
> On Thu, Apr 2, 20
> -Original Message-
> From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of O Mahony,
> Billy
> Sent: Thursday, April 2, 2015 11:13 PM
> To: Ben Pfaff
> Cc: dev@openvswitch.org
> Subject: Re: [ovs-dev] [PATCH RFC 1/1] docs: Clarify bonding of DPDK
> enabled interfaces.
>
>
> > --
On 04/06/2015 06:53 PM, Justin Pettit wrote:
>
>> On Apr 6, 2015, at 12:12 PM, Russell Bryant wrote:
>>
>> +if (ctx->argc == 5) {
>> +/* Validate tag. */
>> +if (sscanf(ctx->argv[4], "%"SCNd64, &tag) != 1 || tag < 0 || tag >
>> 4095) {
>
> We have a wrapped version called ov
On 04/06/2015 08:50 PM, Justin Pettit wrote:
>
>> On Apr 6, 2015, at 12:12 PM, Russell Bryant wrote:
>>
>> ovn-nbd updates the Bindings table of the OVN database when there are
>> updates to the OVN_Northbound database. This update makes ovn-nbd
>> copy over the container integration related fie
On 04/07/15 at 10:20am, Russell Bryant wrote:
> IIRC, the proposal was actually quite explicit that the tag is a VLAN
> ID. It's not a hidden implementation detail because something (not OVN)
> has to set up ovs inside the VM with all of the containers attached and
> have it tag traffic from each
Hi Sorin,
This is much better (it is hard to protect on the object using the object
itself).
Thank you!
Eitan
Acked-by: Eitan Eliahu Thanks, Eitan
-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, April 03, 2015 8:41 AM
To: dev
On Tue, Apr 7, 2015 at 6:56 AM, 马啸 wrote:
> Hi,all
>
>
> I know in normal vxlan support,vteps can be in different subnets.
> But when I config the vxlan tunnel of openvswitch with the local ip and the
> remote ip in different subnets,
> I find the flow can not be matched.
Yes, it should work.
I
ovn-nbd updates the Bindings table of the OVN database when there are
updates to the OVN_Northbound database. This update makes ovn-nbd
copy over the container integration related fields from OVN_Northbound
into the Bindings table.
Signed-off-by: Russell Bryant
---
ovn/ovn-nbd.c | 44 ++
v1->v2
- Tweaks to ovn-nbctl as suggested by Juston
- Update ovn-nbd to handle changes to parent and tag on a logical port
after it is first created
[PATCH 1/2] ovn-nbctl: Updates for container integration.
[PATCH 2/2] ovn-nbd: Update Bindings for container integration.
ovn-nbctl.8.xml |
Thanks for your review Eitan.
-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com]
Sent: Tuesday, 7 April, 2015 18:39
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [PATCH v2] datapath-windows: Solved BSOD when uninstalling the
driver (race condition)
Hi Sorin,
This
The BSOD occurred because the FilterAttach routine released the switch
context, while there were IRPs in processing.
The solution was to add a reference count to prevent premature deallocation of
the
global switch context structure, gOvsSwitchContext.
Signed-off-by: Sorin Vinturis
Reported-by:
Add support for specifying a parent port name and tag when creating
logical ports. Also add commands for getting the parent_name or tag
set on a logical port. These are necessary for dealing with container
interfaces that sit behind normal interfaces.
Signed-off-by: Russell Bryant
---
ovn/ovn-
On Mon, Apr 6, 2015 at 10:19 PM, Ben Pfaff wrote:
> On Mon, Apr 06, 2015 at 09:18:58PM -0700, Ansis Atteka wrote:
>> This patch fixes a deadlock introduced by commit 6b59b543 (ovs-thread:
>> Use fair (but nonrecursive) rwlocks on glibc.)
>>
>> If STP is enabled, then a handler thread could have al
Yamamoto,
ofconn_send(), which will eventually be called for these, already does
ofpmsg_update_length() before passing the message to rconn_send(). Given this,
it seems unnecessary to make this change.
Jarno
> On Apr 6, 2015, at 11:43 PM, YAMAMOTO Takashi wrote:
>
> Signed-off-by: YAMAMOTO
On Tue, 7 Apr 2015 12:43:41 +0200
Thomas Graf wrote:
> We manually rmmod the loaded vports as using modprobe -r
> only works if the modules are available through modules.dep
>
> We do not treat failures to load vports as a fatal error in case
> the vport module has been renamed. Bringing the br
On Tue, 7 Apr 2015 12:43:42 +0200
Thomas Graf wrote:
> In case the openvswitch.ko is loaded manually without dependencies
> built, the vport modules need to be loaded manually as well.
>
> Signed-off-by: Thomas Graf
> ---
> utilities/ovs-dev.py | 3 +++
> 1 file changed, 3 insertions(+)
>
>
On Tue, Apr 7, 2015 at 8:29 AM, Thomas Graf wrote:
> On 04/07/15 at 10:20am, Russell Bryant wrote:
>> IIRC, the proposal was actually quite explicit that the tag is a VLAN
>> ID. It's not a hidden implementation detail because something (not OVN)
>> has to set up ovs inside the VM with all of the
On 04/07/15 at 10:49am, Jesse Gross wrote:
> On Tue, Apr 7, 2015 at 8:29 AM, Thomas Graf wrote:
> > I remember this discussion. One alternative that comes to mind is to
> > simply push a Geneve header in front of it. It would provide a lot
> > more flexibility down the road and we could transmit a
On 04/07/15 at 02:47pm, Flavio Leitner wrote:
> On Tue, 7 Apr 2015 12:43:42 +0200
> Thomas Graf wrote:
>
> > In case the openvswitch.ko is loaded manually without dependencies
> > built, the vport modules need to be loaded manually as well.
> >
> > Signed-off-by: Thomas Graf
> > ---
> > utili
On Tue, Apr 7, 2015 at 11:14 AM, Thomas Graf wrote:
> On 04/07/15 at 10:49am, Jesse Gross wrote:
>> On Tue, Apr 7, 2015 at 8:29 AM, Thomas Graf wrote:
>> > I remember this discussion. One alternative that comes to mind is to
>> > simply push a Geneve header in front of it. It would provide a lot
> On 3 Apr 2015, at 02:19, Ethan Jackson wrote:
>
> Do we have to use the ATOMIC(long long) syntax, or can we use
> atomic_ullong for most of the struct variable definitions? If we have
> to use the former, there should be a comment explaining why.
>
I’ve changed it to atomic_ullong
> In dp_
I totally see your point. I guess it’s not important to keep the
OTHER counter, so I’ve removed it.
Using the new mutex-like semantics the code is much easier to
understand. I’ve changed the names accordingly
I’m sending a v6
Thanks for the review
Daniele
> On 6 Apr 2015, at 22:49, Ethan Jac
These commands can be used to get packets and cycles counters on a pmd
thread basis. They're useful to get a clearer picture about the
performance of the userspace datapath.
They export these pieces of information:
- A (per-thread) view of the caches hit rate. Hits in the exact match
cache are
The counters use x86 TSC if available (currently only with DPDK). They
will be exposed by subsequents commits
Signed-off-by: Daniele Di Proietto
---
lib/dpif-netdev.c | 55 +++
1 file changed, 55 insertions(+)
diff --git a/lib/dpif-netdev.c b/
Since statistics updates might require locking (in future commits)
grouping them will reduce the locking overhead.
Signed-off-by: Daniele Di Proietto
Acked-by: Ethan Jackson
---
lib/dpif-netdev.c | 11 ++-
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/lib/dpif-netdev.c b
A read operation from a non atomic shared value (without external
locking) can return incorrect values. Using the atomic semantics
prevents this from happening.
However:
* No memory barriers are used. We don't need that kind of consistency
for statistics (we use relaxed operations).
* The upda
The goal of this series is to add an appctl command to retrieve detailed
performance statistics from the pmd thread execution. These statistics
include exact match cache and masked classifier hits and rough cycles counters.
This series also fixes some concurrency issues with statistics in the
user
This init function is called when the dpif class is registered. It will
be used by following commits
Signed-off-by: Daniele Di Proietto
Acked-by: Ethan Jackson
---
lib/dpif-netdev.c | 1 +
lib/dpif-netlink.c | 1 +
lib/dpif-provider.h | 8
lib/dpif.c | 8
4 files c
We used to count exact match cache hits and masked classifier hits
together. This commit splits the DP_STAT_HIT counter into two.
This change will be used by future commits.
Signed-off-by: Daniele Di Proietto
Acked-by: Ethan Jackson
---
lib/dpif-netdev.c | 16 ++--
1 file changed, 1
Nice to meet you
After 4 days I noticed the change.
http://catcut.net/UH3
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
assume av=1001, am = 1000, bv=1000, bm=1001.
Should dv's last bit be zero? if not, then my interpretation of what
this function does
is wrong.
On Mon, Apr 6, 2015 at 9:31 PM, Ben Pfaff wrote:
> On Mon, Apr 06, 2015 at 03:28:37PM -0700, Andy Zhou wrote:
>> On Tue, Mar 31, 2015 at 9:52 PM, Ben Pfa
Eat more and gain less in my moto
This is the new way to shape your body.
http://x.co/8jbrG
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On 04/07/15 at 11:34am, Jesse Gross wrote:
> I just meant the software versions of offloading, not that we would
> need actual hardware support for this. We'd want to make sure that
> GSO, virtio, etc. support whatever we do in order to ensure that we
> can efficiently communicate between the host
On Tue, Apr 7, 2015 at 9:35 AM, Sorin Vinturis
wrote:
> The BSOD occurred because the FilterAttach routine released the switch
> context, while there were IRPs in processing.
>
> The solution was to add a reference count to prevent premature deallocation
> of the
> global switch context structure
I tried everything to get in shape
Trust me when I tell you that it's the best for your waistline.
http://x.co/8jbiY
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> The IP TTL is currently omitted in the extracted tunnel information
> that is stored in the flow for userspace tunneling. This includes it
> so that the same logic used by the kernel also applies.
>
> Signed-off-by: Jesse Gross
Acked-by: Pravi
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> On receive, the userspace GRE code doesn't check the protocol
> field. Since OVS only understands Ethernet packets, this adds a
> check that the inner protocol is Ethernet and discards other types
> of packets.
>
> Signed-off-by: Jesse Gross
A
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> The GRE checksum is a 16 bit field stored in a 32 bit option (the
> rest is reserved). The current code treats the checksum as a 32-bit
> field and places it in the right place for little endian systems but
> not big endian. This fixes the prob
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> The indication to calculate the GRE checksum is currently the port
> config rather than the tunnel flow. Currently there is a one to one
> mapping between the two so there is no difference. However, the
> kernel datapath must use the flow and i
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> Currently when printing a userspace tunnel action for VXLAN, the
> VNI is treated as a 32 bit field rather than 24 bit. Even if this
> is the representation that we use internally, we should still show
> the right VNI to avoid confusing people.
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> Currently, the userspace VXLAN implementation contains the code
> for generating and parsing both the UDP and VXLAN headers. This
> pulls out the UDP portion for better layering and to make it
> easier to support additional UDP based tunnels an
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> This adds basic userspace dataplane support for the Geneve
> tunneling protocol. The rest of userspace only has the ability
> to handle Geneve without options and this follows that pattern
> for the time being. However, when the rest of userspa
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> Kernel based OVS recently added the ability to support checksums
> for UDP based tunnels (Geneve and VXLAN). This adds similar support
> for the userspace datapath to bring feature parity.
>
> Signed-off-by: Jesse Gross
Thanks for adding Gen
On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
> As OVS adds userspace support for being the endpoint in protocols
> like tunnels, it will need to be able to calculate pseudoheaders
> as part of the checksum calculation.
>
> Signed-off-by: Jesse Gross
Acked-by: Pravin B Shelar
Thank you for your reply.
The topology as:
tap1(192.168.10.2)--br-int--br-tun(vxlan-0a0b1840 19/5 local_ip=10.10.24.69,
remote_ip=10.11.24.64)==(vxlan-0a0a1845 3/8 local_ip=10.11.24.64,
remote_ip=10.10.24.69)br-tun--br-int--tap2 (192.168.10.1)
I ping from 192.168.10.1 to 192.168.10.2
On Tue, Apr 7, 2015 at 4:22 PM, Pravin Shelar wrote:
> On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
>> diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c
>> index 0c9f5a4..ef96862 100644
>> --- a/lib/netdev-vport.c
>> +++ b/lib/netdev-vport.c
>> @@ -61,6 +61,11 @@ static struct vlog_rate
The following series breaks windows build. Zero length arrays are not
supported in the middle of structures.
source='lib/bfd.c' object='lib/bfd.lo' libtool=yes \
DEPDIR=.deps depmode=none /bin/sh ./build-aux/depcomp \
/bin/sh ./libtool --tag=CC --mode=compile build-aux/cccl
-DHAVE_CONFIG_H -I. -I
It was confusing having a database called "ovn", since it's also the
name of the project. Since we have an "ovn-nb", rename "ovn" to
"ovn-sb".
Signed-off-by: Justin Pettit
---
ovn/.gitignore | 12 +-
ovn/TODO | 22 +-
ovn/automake.mk| 66 +++---
o
Thanks, I pushed a fix to avoid this - nothing was using the member at
the moment, so I just replaced it with a comment.
On Tue, Apr 7, 2015 at 5:31 PM, Gurucharan Shetty wrote:
> The following series breaks windows build. Zero length arrays are not
> supported in the middle of structures.
>
> so
A very popular prescription ED drug may work well for many people, but it is
not the only solution.
http://x.co/8npfE
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Tue, Apr 7, 2015 at 6:01 PM, Jesse Gross wrote:
> Thanks, I pushed a fix to avoid this - nothing was using the member at
> the moment, so I just replaced it with a comment.
Thank you for the quick fix!
>
> On Tue, Apr 7, 2015 at 5:31 PM, Gurucharan Shetty wrote:
>> The following series break
On Tue, Apr 7, 2015 at 4:59 PM, Jesse Gross wrote:
> On Tue, Apr 7, 2015 at 4:22 PM, Pravin Shelar wrote:
>> On Mon, Mar 30, 2015 at 3:14 PM, Jesse Gross wrote:
>>> diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c
>>> index 0c9f5a4..ef96862 100644
>>> --- a/lib/netdev-vport.c
>>> +++ b/lib/n
It looks like this series also breaks the unit tests on Linux. Do you see it?
tnl_push(tnl_port(6),header(size=50,type=5,eth(dst=f8:bc:12:44:34:b6,src=f8:bc:12:46:58:e0,dl_type=0x0800),ipv4(src=1.1.2.88,dst=1.1.2.92,proto=17,tos=0,ttl=64,frag=0x40),udp(src=0,dst=6081,csum=0x0),geneve(oam,vni=0x1c7)
It doesn't trigger on my system (since it is due to uninitialized
memory) but I saw the problem from the autobuilder. I just pushed a
patch to fix it:
https://github.com/openvswitch/ovs/commit/46e7137c77d845c488e17b718eac7c3fb97cedcc
Hopefully that's everything...
On Tue, Apr 7, 2015 at 6:57 PM,
Windows uses pthreads-win32 library to provide the Linux pthread
functionality. It is observed that when the main thread calls
a pthread destructor after it exits, undefined behavior is seen
(e.g., junk values in data).
To avoid this, this commit de-registers the thread destructor
when the main th
It has been observed that sometimes Windows unit tests hang.
This happens when a daemon is started but does not get terminated
when the test ends.
In one particular case, OVS_VSWITCHD_STOP is called which inturn
calls 'ovs-appctl exit'. This causes ovs-vswitchd's atexit handler
to cleanup the pidf
I found it hard to drop belt sizes, until I tried for myself
Your health with improve once you have this.
http://zzb.bz/bGPKE
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
> On Apr 7, 2015, at 8:03 AM, Russell Bryant wrote:
>
> On 04/06/2015 06:53 PM, Justin Pettit wrote:
>
>> It might be nice to print the parent and tag in "lport-list".
>
> Any thoughts on formatting? I don't care much. I was trying to keep it
> simple so it can be parsed without too much pai
76 matches
Mail list logo
