On Fri, Oct 10, 2014 at 5:47 PM, Andy Zhou wrote:
> O.K. Coming back to this patch. Would you please explain why this
> patch needs to affect how GRE, or iptunnel API is used?
>
OVS has compat code for GRE, VXLAN and ip-tunnel APIs. this is
controlled by GRE_USE_KERNEL_GRE_HANDLE_OFFLOADS. But v
O.K. Coming back to this patch. Would you please explain why this
patch needs to affect how GRE, or iptunnel API is used?
On Fri, Oct 10, 2014 at 4:18 PM, Pravin Shelar wrote:
> On Fri, Oct 10, 2014 at 3:29 PM, Andy Zhou wrote:
>> Would it be better to use kernel's CONFIG_VXLAN setting to deter
On Fri, Oct 10, 2014 at 10:15:26PM +, Daniele Di Proietto wrote:
> On 10/10/14, 9:47 AM, "Ben Pfaff" wrote:
>
> >On Wed, Oct 08, 2014 at 02:09:54PM -0700, Daniele Di Proietto wrote:
> >> Signed-off-by: Daniele Di Proietto
> >
> >Do you have any ideas on how to make the new thread stats code
On Fri, Oct 10, 2014 at 09:59:55PM +, Daniele Di Proietto wrote:
>
>
> On 10/10/14, 9:42 AM, "Ben Pfaff" wrote:
> >
> >I thought of another question. I think that there's a hard limit, with
> >this mechanism, on the maximum size of the client's statistics data
> >structure. Without looking
On Fri, Oct 10, 2014 at 09:56:45PM +, Daniele Di Proietto wrote:
>
> On 10/10/14, 9:35 AM, "Ben Pfaff" wrote:
> >I think that you could get atomic clearing by replacing the "clear" bool
> >by a uint32_t that is a sequence number. Calling "clear" would just
> >increment the thread_stats's seq
On Fri, Oct 10, 2014 at 3:29 PM, Andy Zhou wrote:
> Would it be better to use kernel's CONFIG_VXLAN setting to determine
> if vxlan compat code should
> be used?
>
tunnel compat code do not check for CONFIG option. User is expected to
have these module options turned on to use OVS tunneling to get
classifier_find_rule_exactly() only needs the mutex to protect the
list traversal. Subtables are already RCU protected.
Locking here can be eliminated completely when RCU list is available.
Signed-off-by: Jarno Rajahalme
---
lib/classifier.c | 20
1 file changed, 8 inser
>
> > Found by running ovs on RHEL7.
> >
> > Signed-off-by: Alex Wang
>
> It's not obvious to me where the flow escapes from the loop, but the
> fix looks fine.
>
Sure, in upcall_receive(), the &flow is assigned to upcall->flow,
upcall->flow is referred several times in handle_upcalls(),
And t
On Fri, Oct 10, 2014 at 02:29:07PM -0700, Andy Zhou wrote:
> According to RFC 7047, 'ephemeral' annotation does not guarantee
> the values to be durable. This fix Removes this annotation.
>
> VMware-BZ: 1332235
>
> Signed-off-by: Andy Zhou
s/salve/slave/ in the subject line.
s/Removes/removes
On Fri, Oct 10, 2014 at 03:02:09PM -0700, Alex Wang wrote:
> Commit cc377352d (ofproto: Reorganize in preparation for direct
> dpdk upcalls.) introduced the bug that keeps reference to 'struct
> flow' defined on the stack inside while loop when running out of
> the scope. This causes strange bug l
Remove locking that is not needed.
Signed-off-by: Jarno Rajahalme
---
lib/tnl-ports.c | 102 ++-
1 file changed, 41 insertions(+), 61 deletions(-)
diff --git a/lib/tnl-ports.c b/lib/tnl-ports.c
index 4dbda7d..e46dc58 100644
--- a/lib/tnl-port
Signed-off-by: Jarno Rajahalme
---
lib/tnl-router.c | 217 +-
1 file changed, 133 insertions(+), 84 deletions(-)
diff --git a/lib/tnl-router.c b/lib/tnl-router.c
index 4397509..aae457a 100644
--- a/lib/tnl-router.c
+++ b/lib/tnl-router.c
@@ -2
I wanted to see that the recent classifier improvement was enough to
remove the need to add locking when an classifier instance is used as
a container. To that end I improved the relevant code in Pravin's
userspace tunneling patch. These can only be applied after that
patch, or preferably Pravin
On Fri, Oct 10, 2014 at 10:06:29AM -0400, Dave Benson wrote:
> Previous behavior forced the use of the 802.1Q Ethertype for all VLAN headers.
> In order to open the door for future support of VLAN stacking or Provider
> Based
> Bridging, support of the 802.1AD Ethertype is a logical first step.
>
On Oct 10, 2014, at 8:29 AM, Ben Pfaff wrote:
> On Thu, Oct 09, 2014 at 04:21:51PM -0700, Jarno Rajahalme wrote:
>> classifier already provides lockless lookups, and protected
>> modifications. When user wants to remove a flow, we currently require
>> the flow to exist in the classifier. To be
Acked-by: Alex Wang
On Fri, Oct 10, 2014 at 2:29 PM, Andy Zhou wrote:
> According to RFC 7047, 'ephemeral' annotation does not guarantee
> the values to be durable. This fix Removes this annotation.
>
> VMware-BZ: 1332235
>
> Signed-off-by: Andy Zhou
> ---
> vswitchd/vswitch.ovsschema | 7 ++
Would it be better to use kernel's CONFIG_VXLAN setting to determine
if vxlan compat code should
be used?
On Fri, Oct 10, 2014 at 8:21 AM, Pravin B Shelar wrote:
> Kernel 3.11 is only kernel where GRE APIs are available but
> not vxlan. Add check for vxlan xmit to detect this case.
>
> Reported-b
On 10/10/14, 9:47 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:54PM -0700, Daniele Di Proietto wrote:
>> Signed-off-by: Daniele Di Proietto
>
>Do you have any ideas on how to make the new thread stats code a little
>more flexible, so that the awkward special case for clearing only som
On 10/10/14, 9:40 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:53PM -0700, Daniele Di Proietto wrote:
>> Signed-off-by: Daniele Di Proietto
>
>Can you estimate the additional memory overhead per-datapath flow that
>this adds? Our recent review of customer data (for the kernel datapat
a. OvsGetUpcallMsgSize
Should not we do alignment after getting the final size?
b. /* XXX no need as the length is alreadt in the NL attrib */
Spelling typo.
c. We can use NlFillOvsMsg for creating upcall message.
Acked-by: Ankur Sharma
From: dev on b
OvsTunKeyAttrSize & OvsFlowKeyAttrSize, i think the function names are swapped.
Acked-by: Ankur Sharma
From: dev on behalf of Eitan Eliahu
Sent: Friday, October 10, 2014 11:08 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 2/4] Upcall NL packet fo
Acked-by: Ankur Sharma
From: dev on behalf of Eitan Eliahu
Sent: Friday, October 10, 2014 11:08 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 1/4] Upcall NL packet format: Add NL utilities
Add NlAttrSize and NlAttrTotalSize
Signed-off-by: Eitan
On 10/10/14, 9:37 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:52PM -0700, Daniele Di Proietto wrote:
>> Signed-off-by: Daniele Di Proietto
>
>I would fold this into the previous patch. It makes me more confident
>about the module's correctness. Thanks for writing tests.
Sure, no p
On 10/10/14, 9:42 AM, "Ben Pfaff" wrote:
>
>I thought of another question. I think that there's a hard limit, with
>this mechanism, on the maximum size of the client's statistics data
>structure. Without looking at the patch again, I think it's 64 - 8 = 56
>bytes. Is there anything in the cod
Commit cc377352d (ofproto: Reorganize in preparation for direct
dpdk upcalls.) introduced the bug that keeps reference to 'struct
flow' defined on the stack inside while loop when running out of
the scope. This causes strange bug like wrong mask extraction
when the part of memory is corrupted, and
Acked-by: Ankur Sharma
From: dev on behalf of Eitan Eliahu
Sent: Friday, October 10, 2014 11:08 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 4/4] Upcall NL packet format: Call NL missed packet
function
Remove hard coded queue is, pass the key
On 10/10/14, 9:35 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:51PM -0700, Daniele Di Proietto wrote:
>> Signed-off-by: Daniele Di Proietto
>
>This looks very useful! Thank you.
>
>I am not sure that we need the "ovs" prefix on all the names. I
>included it in some ovsthread_*() name
According to RFC 7047, 'ephemeral' annotation does not guarantee
the values to be durable. This fix Removes this annotation.
VMware-BZ: 1332235
Signed-off-by: Andy Zhou
---
vswitchd/vswitch.ovsschema | 7 +++
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/vswitchd/vswitch.ov
On 10/10/14, 9:13 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:50PM -0700, Daniele Di Proietto wrote:
>> An rcu-slist can be read and written without explicit locking: it uses
>>RCU
>> protected pointers and compare-exchange operations.
>>
>> Signed-off-by: Daniele Di Proietto
>
>I t
On 10/10/14, 8:53 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:49PM -0700, Daniele Di Proietto wrote:
>> OVSRCU_TYPE_INITIALIZER now accepts a parameter (the initial value of
>>the RCU
>> protected pointer).
>>
>> Signed-off-by: Daniele Di Proietto
>
>Good idea.
>
>It looks like the
On 10/10/14, 8:47 AM, "Ben Pfaff" wrote:
>On Wed, Oct 08, 2014 at 02:09:47PM -0700, Daniele Di Proietto wrote:
>> This also fixes a memory leak if an upcall fails
>>
>> Signed-off-by: Daniele Di Proietto
Thanks for the all the reviews, Ben!
>
>I generally prefer to separate bug fixes from ot
On Fri, Oct 10, 2014 at 07:20:14PM +0530, Rishi Bamba wrote:
> Hi Ben,
>
> # In reference to the [Patch 1/4] This patch includes the Design file changes
> as suggested and test cases for add-flow and dump-flow
> # CLI in order to perform testing of the importance parameter in a rule for
> differ
On Fri, Oct 10, 2014 at 07:11:45PM +0530, Rishi Bamba wrote:
> Hi Ben,
>
> # Thank you for reviewing the patch.As per the comments received ,all the
> changes are made and incorporated for this patch i.e Clang
> # compiler doesn't report any error now and all the test cases are successful
> incl
On Fri, Oct 10, 2014 at 11:36:39AM +0530, Shashwat Srivastava wrote:
> We are working on the implementation on eviction mechanism on basis of
> importance as per openflow specs-1.4. We have added importance parameter
> in the flow entry. And are going through various commands (replace-flows,
> d
Kernel 3.11 is only kernel where GRE APIs are available but
not vxlan. Add check for vxlan xmit to detect this case.
Reported-by: Dave Benson
Signed-off-by: Pravin B Shelar
---
acinclude.m4 |1 +
datapath/linux/compat/gre.c|2 +-
dat
Acked-by: Ankur Sharma
From: dev on behalf of Eitan Eliahu
Sent: Tuesday, October 7, 2014 8:01 PM
To: Nithin Raju; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v2] lib/netlink-socket.c: always pass the output
buffer in a transaction
Hi Nithin,
Th
On Wed, Oct 08, 2014 at 02:09:56PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Oct 08, 2014 at 02:09:55PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Oct 08, 2014 at 02:09:54PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
Do you have any ideas on how to make the new thread stats code a little
more flexible, so that the awkward special case for clearing only some
stats in this patch can be avoided? One idea wo
On Fri, Oct 10, 2014 at 09:35:52AM -0700, Ben Pfaff wrote:
> On Wed, Oct 08, 2014 at 02:09:51PM -0700, Daniele Di Proietto wrote:
> > Signed-off-by: Daniele Di Proietto
>
> This looks very useful! Thank you.
>
> I am not sure that we need the "ovs" prefix on all the names. I
> included it in s
On Wed, Oct 08, 2014 at 02:09:53PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
Can you estimate the additional memory overhead per-datapath flow that
this adds? Our recent review of customer data (for the kernel datapath)
showed, if I recall correctly, a 99th percentil
FYI: The status of the openvswitch source package
in Debian's testing distribution has changed.
Previous version: 2.1.0+git20140411-3
Current version: 2.3.0+git20140819-2
--
This email is automatically generated once a day. As the installation of
new packages into testing happens multiple
On Wed, Oct 08, 2014 at 02:09:52PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
I would fold this into the previous patch. It makes me more confident
about the module's correctness. Thanks for writing tests.
___
dev mailing l
On Wed, Oct 08, 2014 at 02:09:51PM -0700, Daniele Di Proietto wrote:
> Signed-off-by: Daniele Di Proietto
This looks very useful! Thank you.
I am not sure that we need the "ovs" prefix on all the names. I
included it in some ovsthread_*() names to provide a contrast and
analogy to pthread_*().
On Wed, Oct 08, 2014 at 02:09:50PM -0700, Daniele Di Proietto wrote:
> An rcu-slist can be read and written without explicit locking: it uses RCU
> protected pointers and compare-exchange operations.
>
> Signed-off-by: Daniele Di Proietto
I think that there are a couple of key properties of this
On Wed, Oct 08, 2014 at 02:09:49PM -0700, Daniele Di Proietto wrote:
> OVSRCU_TYPE_INITIALIZER now accepts a parameter (the initial value of the RCU
> protected pointer).
>
> Signed-off-by: Daniele Di Proietto
Good idea.
It looks like the main value in OVSRCU_TYPE_INIT_NULL is that the name
of
On Wed, Oct 08, 2014 at 02:09:48PM -0700, Daniele Di Proietto wrote:
> ovsrcu_compare_exchange() can be used to write to RCU pointers without
> explicit
> locking. It maybe useful to implement lists and other simple data structures.
>
> Following commits make use of this
>
> Signed-off-by: Danie
On Wed, Oct 08, 2014 at 02:09:47PM -0700, Daniele Di Proietto wrote:
> This also fixes a memory leak if an upcall fails
>
> Signed-off-by: Daniele Di Proietto
I generally prefer to separate bug fixes from other changes, when it is
not difficult. Can the memory leak fix be a separate patch?
In
On Thu, Oct 09, 2014 at 10:18:32AM +0900, Simon Horman wrote:
> On Fri, Sep 26, 2014 at 04:55:42PM -0700, Ben Pfaff wrote:
> > On Thu, Sep 18, 2014 at 10:55:05AM +0900, Simon Horman wrote:
> > > Add a multiple field to struct nl_policy which if set suppresses
> > > warning of duplicate attributes i
On Thu, Oct 09, 2014 at 04:21:51PM -0700, Jarno Rajahalme wrote:
> classifier already provides lockless lookups, and protected
> modifications. When user wants to remove a flow, we currently require
> the flow to exist in the classifier. To be thread safe, this requires
> the caller to introduce
On Fri, Oct 10, 2014 at 01:39:56PM +0800, Liuyongqiang(A) wrote:
> From: liuyongqiang
>
> this patch work for the ovs serporting the IEEE 802.1ah
> protocol(Mac in Mac or PBB)
>
> IEEE 802.1ah (PBB or MAC-in-MAC):
> PBB (also known as MAC-in-MAC) is used by SPs to resolve these problems.
> PBB i
Hi Ben,
# This is the last patch which includes test case for replace-flows command to
validate addition of the importance parameter in a
# rule as per OF14.Also attaching the patch file which includes all the four
patches[1-4] for you reference.
---
This patch adds a test case for testing the r
Previous behavior forced the use of the 802.1Q Ethertype for all VLAN headers.
In order to open the door for future support of VLAN stacking or Provider Based
Bridging, support of the 802.1AD Ethertype is a logical first step.
This patch adds support for OpenFlow flows to specify push_vlan actions
Hi Ben,
---
This patch enhances the diff-flows & replace-flows CLI that takes into
account the addition of importance parameter in a rule as per the
OpenFlow Specs 1.4.
Signed-off-by: Rishi Bamba
---
utilities/ovs-ofctl.c | 8
1 file changed, 8 insertions(+)
diff --git a/utilities/ovs
Hi Ben,
# In reference to the [Patch 1/4] This patch includes the Design file changes
as suggested and test cases for add-flow and dump-flow
# CLI in order to perform testing of the importance parameter in a rule for
different protocols of OpenFlow.
---
This patch enables a user to set importan
Hi Ben,
# Thank you for reviewing the patch.As per the comments received ,all the
changes are made and incorporated for this patch i.e Clang
# compiler doesn't report any error now and all the test cases are successful
including the one added by us for the same.Also along
# with this patch sendi
[1] Allocate a queue element and space to hold the packet, key, tunnel key
and user data in NL format.
[2] Format the NL header
[3] Store packet, key, tunnel key and user data in NL format
[4] Calculates and insert checksum if offloaded.
signed-off-by: Eitan Eliahu
---
datapath-windows/include/O
Add NlAttrSize and NlAttrTotalSize
Signed-off-by: Eitan Eliahu
---
datapath-windows/ovsext/Netlink/Netlink.h | 22 ++
1 file changed, 22 insertions(+)
diff --git a/datapath-windows/ovsext/Netlink/Netlink.h
b/datapath-windows/ovsext/Netlink/Netlink.h
index cd55647..cd69f3a 1
Extend the key and tunnel key conversion to nested NL format functions use the
NL attribute as a parameter so we can use them for missed packet formatting.
Add functions for calculating the space needed for storing the key and the
tunnel key in NL format
Signed-off-by: Eitan Eliahu
---
datapath
Remove hard coded queue is, pass the key rather the tunnel key,
remove the none NL implementation
Signed-off-by: Eitan Eliahu
---
datapath-windows/ovsext/Actions.c | 29 ++---
datapath-windows/ovsext/PacketIO.c | 7 +-
datapath-windows/ovsext/Tunnel.c | 4 +-
datapath-windows/ovsext/User
On Oct 9, 2014, at 11:02 PM, Ankur Sharma wrote:
> During vswitchd boot up kernel is receiving FLOW_ADD commands
> without ETHERTYPE attribute. Added additional check for the same
> in this patch.
>
> Signed-off-by: Ankur Sharma
> +/* TODO: Ideally ETHERTYPE should not be optional.
s/TOD
61 matches
Mail list logo
