On Thu, Aug 10, 2017 at 01:47:26PM -0700, Ben Pfaff wrote:
> On Thu, Aug 10, 2017 at 04:41:19PM -0400, Lance Richardson wrote:
> > Some recent travis builds have failed due to having exceeded the
> > per-job time limit of 50 minutes. This change enables parallel
> > builds and parallel test
The first time after boot, when openvswitch is started by using
--delete-transient-ports-on-boot ovs-ctl option, all transient ports will be
removed.
Signed-off-by: Timothy Redaelli
---
utilities/ovs-ctl.8 | 13 -
utilities/ovs-ctl.in | 6 ++
2 files
Timothy Redaelli writes:
> Use ovs-ctl --delete-transient-ports-on-boot to start ovsdb-server.
>
> This feature can be disabled by appending --no-delete-transient-ports-on-boot
> to OPTIONS in /etc/sysconfig/openvswitch
>
> Signed-off-by: Timothy Redaelli
On Fri, Aug 11, 2017 at 11:06:45AM -0700, Joe Stringer wrote:
> In C++, 'mutable' is a keyword. If this is used as the name for a field,
> then C++ compilers can get confused about the context and fail to
> compile references to such fields. Rename the field to 'is_mutable' to
> avoid this issue.
On Fri, Aug 11, 2017 at 11:06:44AM -0700, Joe Stringer wrote:
> In C++, 'class' is a keyword. If this is used as the name for a field,
> then C++ compilers can get confused about the context and fail to
> compile references to such fields. Rename the field to 'class_' to
> avoid this issue.
>
>
On 08/11/2017 09:27 AM, Jan Scheurich wrote:
> Hi Ilya,
>
> Tuning the per EMC insertion probability per port based on detailed knowledge
> about the nature of traffic patterns seems a micro-optimization to me, which
> might be helpful in very controlled setups e.g. in synthetic benchmarks, but
In the OVSDB IDL, we use C++ keywords such as "new", "mutable", "class"
for variable and field names. This series updates these names so that
they don't conflict with the C++ keywords, which should improve the
ability to use the IDL from C++ code. This series focuses primarily on
code that exists
-Original Message-
From: Ilya Maximets
Date: Friday, August 11, 2017 at 6:01 AM
To: "ovs-dev@openvswitch.org"
Cc: Heetae Ahn , Antonio Fischetti
, Darrell Ball ,
Use ovs-ctl --delete-transient-ports-on-boot to start ovsdb-server.
This feature can be disabled by appending --no-delete-transient-ports-on-boot
to OPTIONS in /etc/sysconfig/openvswitch
Signed-off-by: Timothy Redaelli
---
rhel/usr_lib_systemd_system_ovsdb-server.service
On 11 August 2017 at 04:26, Joe Perches wrote:
> OVS_NLERR already adds a newline so these just add blank
> lines to the logging.
>
> Signed-off-by: Joe Perches
Acked-by: Joe Stringer
___
dev mailing
In C++, 'new' is a keyword. If this is used as the name for a field,
then C++ compilers can get confused about the context and fail to
compile references to such fields. Rename the field to 'new_datum' to
avoid this issue.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
On Fri, Aug 11, 2017 at 7:41 AM, Ilya Maximets wrote:
>
> Could you suggest another dp_packet_batch_XXX() name (which can be exposed
> to the end user) for the function that doesn't check the boundaries instead
> of 'add_unsafe', if you think it is not accurate?
How about
Now that the 'new' datum is named 'new_datum', be more consistent by
renaming 'old' to 'old_datum' to match.
Signed-off-by: Joe Stringer
Acked-by: Ben Pfaff
---
v3: Add ack.
v2: New patch.
---
lib/ovsdb-data.h | 4 +--
lib/ovsdb-idl-provider.h | 22
Hi Bhanu
Thanks for your efforts.
The reason I sent this e-mail regarding patch 5 is that as is, it could not be
merged.
Since, I have not seen the incremental changes (used in some referred customer
testing) beyond this, I wanted to
give a reasonable chance to evaluate the full changes.
On 08/11/2017 06:20 PM, Darrell Ball wrote:
> The DPDK introductory documentation has various references to
> pmd-cpu-mask, including a section devoted to it. These parts of
> the documentation seemed to have been written at different times
> and look like they were individually ported from other
The DPDK introductory documentation has various references to
pmd-cpu-mask, including a section devoted to it. These parts of
the documentation seemed to have been written at different times
and look like they were individually ported from other sources.
They all include an example command which
Add --delete-transient-ports-on-boot flag to ovs-ctl in order to remove
all transient ports (--delete-transient-ports) only the first time after a
boot.
This is important on certain environments where some ports are going to be
recreated after reboot, but other ports need to be persisted in the
I don't know. This patch seems to say that it fixes a problem if we
revert commit 8c2c225e481d ("netdev: Fix netdev_open() to track and
recreate classless interfaces"). That patch hasn't been reverted.
Given that, can you explain the value of this patch?
Is removing route_table_reset() related
On 10 August 2017 at 13:30, Ben Pfaff wrote:
> On Wed, Aug 09, 2017 at 03:27:41PM -0700, Joe Stringer wrote:
>> In C++, 'new' is a keyword. If this is used as the name for a field,
>> then C++ compilers can get confused about the context and fail to
>> compile references to such
I just wanted to specially call out this patch for testing. My hope is that we
can merge this into 2.8, but wanted to give people a heads up that they may
want to try it with their applications. The most visible change is that a call
to ct() will clear the ct_state for any actions that follow
Hi Ilya,
Tuning the per EMC insertion probability per port based on detailed knowledge
about the nature of traffic patterns seems a micro-optimization to me, which
might be helpful in very controlled setups e.g. in synthetic benchmarks, but
very hard to apply in more general use cases, such as
Hi Ilya,
I fully agree with storing 'now' as part of the pmd struct instead of passing
it around as function arguments.
For me struct dp_netdev_pmd_thread is *the* the PMD thread context in
dpif-netdev. I don't really see the benefit of creating a sub-struct
dp_netdev_pmd_thread_ctx and
On Thu, 10 Aug 2017 21:21:15 +0800, Yi Yang wrote:
> OVS master and 2.8 branch has merged NSH userspace
> patch series, this patch is to enable NSH support
> in kernel data path in order that OVS can support
> NSH in 2.8 release in compat mode by porting this.
Please include changelog when
So far, we're not clear how we can support MD type 2 better, as I explained
before, we need to reuse tun_metadata in struct flow_tnl which is the thing
Geneve is using. Geneve predefined 64 keys for this from tun_metadata0 to
tun_metadata63, we will reuse it for MD type 2. But you know NSH is
On Fri, 11 Aug 2017 09:24:25 +, Yang, Yi Y wrote:
> So far, we're not clear how we can support MD type 2 better, as I
> explained before, we need to reuse tun_metadata in struct flow_tnl
> which is the thing Geneve is using. Geneve predefined 64 keys for
> this from tun_metadata0 to
On Fri, 11 Aug 2017 16:47:23 +0800, Yang, Yi wrote:
> is "__be32 context[4]" ok?
Yes, that looks better.
> So define three new netlink attributes
>
> OVS_ACTION_ATTR_NSH_BASE_HEADER
> OVS_ACTION_ATTR_NSH_MD1_DATA
> OVS_ACTION_ATTR_NSH_MD2_DATA
>
> OVS_ACTION_ATTR_PUSH_NSH is nested netlink
Packet and Connection state is only available to the processing path
that follows the "recirc_table" argument of the ct() action. The
previous behavior made these states available until the end of the
pipeline. This commit changes the behavior so that the Packet and
Connection state are cleared
On Fri, Aug 11, 2017 at 10:24:18AM +0200, Jiri Benc wrote:
> On Thu, 10 Aug 2017 21:21:15 +0800, Yi Yang wrote:
> > OVS master and 2.8 branch has merged NSH userspace
> > patch series, this patch is to enable NSH support
> > in kernel data path in order that OVS can support
> > NSH in 2.8 release
> > and a series we would like to get into 2.8
> >
> > netdev-dpdk: Use intermediate queue during packet transmission. Bhanu Jun
> > 29/V3
> > netdev: Add netdev_txq_flush function.
> > netdev-dpdk: Add netdev_dpdk_txq_flush function.
> > netdev-dpdk: Add netdev_dpdk_vhost_txq_flush function.
>
Hello All,
Adding all the people here who had either reviewed or provided their feedback
on the batching patches at some stage.
you are already aware that there are 2 different series on ML to implement tx
batching (netdev layer vs dpif layer) that improves DPDK datapath performance.
Our
>On 09.08.2017 15:35, Bodireddy, Bhanuprakash wrote:
+static int
+netdev_dpdk_vhost_tx_burst(struct netdev_dpdk *dev, int qid) {
+struct dpdk_tx_queue *txq = >tx_q[qid];
+struct rte_mbuf **cur_pkts = (struct rte_mbuf
+**)txq->vhost_burst_pkts;
+
+
On 11.08.2017 11:13, Jan Scheurich wrote:
> Hi Ilya,
>
> I fully agree with storing 'now' as part of the pmd struct instead of passing
> it around as function arguments.
>
> For me struct dp_netdev_pmd_thread is *the* the PMD thread context in
> dpif-netdev. I don't really see the benefit of
Raw Message Printable View
ATTENTION FUND BENEFICIARY:
Information reaching us from our head office Money Gram Corporate
Headquarters now, States that you only have 48hours to effect payment
for the activation of your transfer payment $900,000.00 US dollars (Nine
Hundred Thousand United
On 10.08.2017 21:52, Bodireddy, Bhanuprakash wrote:
>>>
> } else {
> +/* If the queue is disabled in the guest, the
> corresponding qid
> + * map shall be set to OVS_VHOST_QUEUE_DISABLED(-2).
> + *
> +
Thanks for fixing this, Ilya!
It was correct at the time it was committed but very likely to break afterwards.
I should have removed the loop from the start and change the way you propose.
Acked-by: Jan Scheurich
> -Original Message-
> From: Ilya Maximets
On 11.08.2017 16:11, Bodireddy, Bhanuprakash wrote:
>> On 09.08.2017 15:35, Bodireddy, Bhanuprakash wrote:
>
> +static int
> +netdev_dpdk_vhost_tx_burst(struct netdev_dpdk *dev, int qid) {
> +struct dpdk_tx_queue *txq = >tx_q[qid];
> +struct rte_mbuf **cur_pkts =
REF NR: .576061825
STAPEL NR: .6485769789/646
LOTTO GEWINNER
BETR.: ENDGULTIGE NOTIFIKATION
GLUCKWUNSCHE!
Vereinigten Staaten von Amerika Lotterie Inc hiermit teilt Ihnen die
If this is just about ovs-ofctl action syntax, I guess that is a matter of
taste. In my eyes there is a value in keeping a closer correspondence between
the OpenFlow actions and the action syntax to avoid confusion for users, even
if the resulting syntax looks slightly less elegant. ovs-ofctl
DP_STAT_LOOKUP_HIT statistics used mistakenly for calculation
of total number of packets. This leads to completely wrong
per packet cycles statistics.
For example:
emc hits:0
megaflow hits:253702308
avg. subtable lookups per hit:1.50
miss:0
lost:0
On Fri, Aug 11, 2017 at 11:44:49AM +0200, Jiri Benc wrote:
> On Fri, 11 Aug 2017 09:24:25 +, Yang, Yi Y wrote:
> > So far, we're not clear how we can support MD type 2 better, as I
> > explained before, we need to reuse tun_metadata in struct flow_tnl
> > which is the thing Geneve is using.
Fri, Aug 11, 2017 at 12:09:36PM CEST, jan.scheur...@ericsson.com wrote:
>> -Original Message-
>> From: Jiri Benc [mailto:jb...@redhat.com]
>> Sent: Friday, 11 August, 2017 11:45
>>
>> The context field does not apply to MD type 2. It looks wrong for the
>> context field to be included in
REF NR: .576061825
STAPEL NR: .6485769789/646
LOTTO GEWINNER
BETR.: ENDGULTIGE NOTIFIKATION
GLUCKWUNSCHE!
Vereinigten Staaten von Amerika Lotterie Inc hiermit teilt Ihnen die
Fix comments for emc_processing and dp_netdev_input__
regarding md_is_valid.
Signed-off-by: Antonio Fischetti
---
lib/dpif-netdev.c | 13 +++--
1 file changed, 7 insertions(+), 6 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index
Hi Ben, this patch is ok ?
On Fri, Aug 4, 2017 at 10:45 AM, Tonghao Zhang wrote:
> We can avoid the deadlock via removing the route_table_reset() from
> the route_table_init()
>
> the call trace is below
> dp_initialize (ovsthread_once)
> route_table_init
>
On Fri, 11 Aug 2017 10:09:36 +, Jan Scheurich wrote:
> Unless someone can explain to me why the datapath should understand the
> internal structure/format of metadata in push_nsh, I would strongly
> vote to keep the metadata as variable length octet sequence in the
> non-structured
> -Original Message-
> From: Jiri Benc [mailto:jb...@redhat.com]
> Sent: Friday, 11 August, 2017 11:45
>
> The context field does not apply to MD type 2. It looks wrong for the
> context field to be included in netlink attribute for anything other
> than MD type 1. Perhaps it needs to be
OVS_NLERR already adds a newline so these just add blank
lines to the logging.
Signed-off-by: Joe Perches
---
net/openvswitch/conntrack.c| 14 +-
net/openvswitch/flow_netlink.c | 2 +-
2 files changed, 6 insertions(+), 10 deletions(-)
diff --git
> On Aug 11, 2017, at 8:54 AM, Stokes, Ian wrote:
>
> Below are the features that Intel are planning to submit for the OVS 2.9
> release.
>
> I've also created a google spreadsheet so that people can contribute their
> own planned features. This can be found at:
>
>
Acked-by: Sairam Venugopal
On 8/11/17, 1:41 PM, "ovs-dev-boun...@openvswitch.org on behalf of Anand Kumar"
wrote:
> - Update the CT comparison function to compare individual fields instead of
Acked-by: Sairam Venugopal
On 8/10/17, 8:59 PM, "ovs-dev-boun...@openvswitch.org on behalf of Anand Kumar"
wrote:
>During SNAT/DNAT, we should not be updating the port field of ct_endpoint
>struct, as
- Update the CT comparison function to compare individual fields instead of
NdisEqualMemory.
- Add in some padding for the ct_endpoint's union.
- Update the Orig Tuple to use ICMP Type and Code instead of Port for ICMP
Co-authored-by: Sairam Venugopal
Signed-off-by: Anand
From: Joe Perches
Date: Fri, 11 Aug 2017 04:26:26 -0700
> OVS_NLERR already adds a newline so these just add blank
> lines to the logging.
>
> Signed-off-by: Joe Perches
Applied to net-next.
___
dev mailing list
On 9 August 2017 at 17:18, Joe Stringer wrote:
> Previously, netdev_ports_insert() would allocate and insert an
> ifindex->odp_port mapping, but netdev_ports_remove() would never remove
> the mapping or free the mapping structure. This patch fixes these up.
>
> Fixes:
On 9 August 2017 at 20:54, Gao Zhenyu wrote:
> How about mutable --> is_mutable ?
Thanks, I'll apply this change and respin.
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
In C++, 'class' is a keyword. If this is used as the name for a field,
then C++ compilers can get confused about the context and fail to
compile references to such fields. Rename the field to 'class_' to
avoid this issue.
Signed-off-by: Joe Stringer
---
v3: Also update
In C++, 'mutable' is a keyword. If this is used as the name for a field,
then C++ compilers can get confused about the context and fail to
compile references to such fields. Rename the field to 'is_mutable' to
avoid this issue.
Signed-off-by: Joe Stringer
---
v3: Rename to
On 8 August 2017 at 01:03, Simon Horman wrote:
> On Mon, Aug 07, 2017 at 06:19:04PM +0300, Roi Dayan wrote:
>> Hi,
>>
>> This series adds support for offloading ip ttl and tcp flags
>> using tc interface.
>
> This looks nice, thanks.
>
> Acked-by: Simon Horman
-Original Message-
From: Jan Scheurich
Date: Friday, August 11, 2017 at 2:16 AM
To: Ilya Maximets , "ovs-dev@openvswitch.org"
, Darrell Ball
Cc: "Bodireddy, Bhanuprakash
When OVS is configured as a firewall, with thousands of active
concurrent connections, the EMC gets quicly saturated and may
come under heavy thrashing for the reason that original and
recirculated packets keep overwriting the existing active EMC
entries due to its limited size (8k).
This
This patchset comes from an attempt to improve the ConnTracker
performance in OvS-DPDK when used as a firewall.
We focused on some aspects like the management of recirculated packets
and some function tuning.
Some of this changes may be a benefit also for cases other than the
connection tracker
When EMC is disabled the reading of RSS hash is skipped.
Also, for packets that are not recirculated it retrieves the hash
value without considering the recirc id.
Signed-off-by: Antonio Fischetti
---
This patch depends on previous patch in this series.
Port-to-Port
memcpy replaces the several single copies inside
dp_packet_clone_with_headroom().
Signed-off-by: Antonio Fischetti
---
I tested this change by comparing the CPU Time over a 60 sec analysis
with VTune.
In original ovs:
dp_packet_clone_with_headroom4.530s
+ this
Current time is passed to conntrack_execute so it doesn't have
to recompute it again.
Signed-off-by: Antonio Fischetti
Acked by: Sugesh Chandran
---
In a firewall testbench set up with
table=0, priority=1 actions=drop
table=0,
Below are the features that Intel are planning to submit for the OVS 2.9
release.
I've also created a google spreadsheet so that people can contribute their own
planned features. This can be found at:
On 09.08.2017 17:19, Andy Zhou wrote:
>> Maybe I don't fully understand what you're trying to say, but I want to use
>> unsafe function in dpif-netdev for per-flow packet batching (see the patch)
>> and it should not be internal for that case.
>> (It's safe to use unsafe function there because
Hi Bhanu,
Thanks a lot for all the work you put into making the original hackfest patch
good enough for upstreaming in an increasingly complex and ever-evolving DPDK
datapath.
I tend to agree that the complexity to achieve thread safety is becoming
overwhelming. Ilya's v3 patch set submitted
66 matches
Mail list logo