Split ovs_dp_packet_flow_lookup() into its own API. In preparation for
the next patch.
Signed-off-by: Andy Zhou
---
datapath/datapath.c | 27 ---
datapath/datapath.h | 1 +
2 files changed, 21 insertions(+), 7 deletions(-)
diff --git a/datapath/datapath.c b/datapath/dat
Future patches will change the recirc action implementation to not
using recursion. The stack depth detection is no longer necessary.
Signed-off-by: Andy Zhou
---
datapath/actions.c | 58 +++--
datapath/datapath.c | 6 +++---
datapath/datapath.h
Since kernel stack is limited in size, it is not wise to using
recursive function with large stack frames.
This patch provides an alternative implementation of recirc action with
out using recursion.
A recirc action can be embedded in the middle of an actions list.
Instead of immediately (and rec
Future pathces will make use of those functions.
Signed-off-by: Andy Zhou
---
datapath/Modules.mk | 1 +
datapath/actions.c | 57 +
datapath/actions.h | 54 ++
datapath/datapath.c | 4
da
On Mon, Aug 18, 2014 at 1:29 PM, Daniele Di Proietto
wrote:
> Since lookups in the classifier can be pretty expensive, we introduce this
> (thread local) cache which simply compares the miniflows of the packets
>
Patch looks good, I have few comment, nothing major.
> Signed-off-by: Daniele Di Pro
On Mon, Aug 18, 2014 at 1:29 PM, Daniele Di Proietto
wrote:
> ofpbuf_rss() is introduced to get the RSS hash from the NIC. It works only for
> packets received by DPDK (otherwise it returns 0, which tells the caller to
> compute the hash in another way)
>
> This commit also configure DPDK devices
Hi Ben,
>On Thu, Aug 14, 2014 at 08:35:36PM -0700, Nithin Raju wrote:
>> The datapath interface defined in odp-netlink.h needs some extensions
>> that are platform dependent. Some examples are the name of the
>>communication
>> device on Windows and a set of commands that are specific to Windows.
On Aug 18, 2014, at 5:35 PM, Alin Serdean
wrote:
> I think you can work with Sam together this week to get the Netlink part in
> the kernel up and running.
Alin,
Just one question here. From my understanding, VMware folks are working on the
kernel patch to add netlink support, and Alin (and p
On Aug 18, 2014, at 5:35 PM, Alin Serdean
wrote:
> I worked mostly on the userspace last week and Sam worked on multiple NB to
> NBL.
>
> I will redo my patches with the build and userspace somewhere tomorrow and
> will repost Sam's patch (we spent some time in the weekend testing it).
Sound
Hi Guru,
I tested the same thing with /FS
(http://msdn.microsoft.com/en-us/library/dn502518.aspx ) earlier today but it
was still hanging. I will retry on a fresh environment tomorrow.
$ uname -r
1.0.18(0.48/3/2)
Thanks,
Alin.
-Mesaj original-
De la: Gurucharan Shetty [mailto:shet...@n
I worked mostly on the userspace last week and Sam worked on multiple NB to NBL.
I will redo my patches with the build and userspace somewhere tomorrow and will
repost Sam's patch (we spent some time in the weekend testing it).
I think you can work with Sam together this week to get the Netlink
On Aug 18, 2014, at 5:14 PM, Alin Serdean
wrote:
> I'm fine with this patch with the idea in mind that OvsDpInterfaceExt.h can
> be added to odp-netlink.h. Not sure about the name though :).
Alin,
As the name indicates, the idea is to add "extensions" to the DP interface. We
could add the co
On Aug 18, 2014, at 5:17 PM, Alin Serdean
wrote:
> Good place to start in my opinion but I will leave the final review to Sam
> because he is working on the Netlink integration part aswell.
Alin,
Thanks for taking a look late into the night :) So, I'll take this as an ACK
from you.
It is fin
The validation part I think needs more checking in some places, but it is a
good place to start.
Again I will leave the final review to Sam because he is also working on this
matter as well.
Thanks,
Alin.
-Mesaj original-
De la: dev [mailto:dev-boun...@openvswitch.org] În numele Nithin
Good place to start in my opinion but I will leave the final review to Sam
because he is working on the Netlink integration part aswell.
Alin.
-Mesaj original-
De la: dev [mailto:dev-boun...@openvswitch.org] În numele Nithin Raju
Trimis: Friday, August 15, 2014 6:36 AM
Către: dev@openvsw
I'm fine with this patch with the idea in mind that OvsDpInterfaceExt.h can be
added to odp-netlink.h. Not sure about the name though :).
Also might be useful to define the fourth ioctl for the PID.
Alin.
-Mesaj original-
De la: dev [mailto:dev-boun...@openvswitch.org] În numele Nithin
Hi Nithin,
Just a few thoughts:
Since get PID is a special case would it not be better to treat it as a
DeviceIoControl;dwIoControlCode like OVS_IOCTL_TRANSACT (i.e.
OVS_IOCTL_GET_PID)?
We can drop the extra headers and the whole Netlink message building/parsing
since the goal is just to get
On Mon, Aug 18, 2014 at 1:44 PM, Andy Zhou wrote:
> On Sun, Aug 17, 2014 at 7:43 PM, Pravin Shelar wrote:
>> On Fri, Aug 15, 2014 at 3:12 AM, Andy Zhou wrote:
>>> Since kernel stack is limited in size, it is not wise to using
>>> recursive function with large stack frames.
>>>
>>> This patch pro
On Mon, Aug 18, 2014 at 1:32 PM, Andy Zhou wrote:
> This is what I should do if I understand your intention correctly:
>
> 1. Change this patch to remove all loop detection mechanisms.
> 2. Submit a separate patch to the compact code that check for a lower
> xmit_recursion limit in rpl_dev_queue_x
On Fri, Aug 15, 2014 at 9:58 AM, Alin Serdean
wrote:
> Hey Guru,
>
> I can add a follow up on the after the version 2 to propose a solution on how
> to make the compilation faster :).
It looks like the "hangs" that I was seeing was because of this bug:
http://sourceforge.net/p/mingw/bugs/1950/?li
On Thu, Aug 14, 2014 at 08:35:39PM -0700, Nithin Raju wrote:
> To verify if the netlink support in the kernel works, I updated
> the netlink-socket.c code to get the PID for a given device
> descriptor.
>
> In the existing code, userspace sets the PID, which will not be
> unique across different p
On Sun, Aug 17, 2014 at 7:43 PM, Pravin Shelar wrote:
> On Fri, Aug 15, 2014 at 3:12 AM, Andy Zhou wrote:
>> Since kernel stack is limited in size, it is not wise to using
>> recursive function with large stack frames.
>>
>> This patch provides an alternative implementation of recirc action with
ofpbuf_rss() is introduced to get the RSS hash from the NIC. It works only for
packets received by DPDK (otherwise it returns 0, which tells the caller to
compute the hash in another way)
This commit also configure DPDK devices to compute RSS hash for UDP and IPv6
packets
Signed-off-by: Daniele D
Since lookups in the classifier can be pretty expensive, we introduce this
(thread local) cache which simply compares the miniflows of the packets
Signed-off-by: Daniele Di Proietto
---
lib/dpif-netdev.c | 422 +-
1 file changed, 357 insertions
This series introduces in dpif-netdev an exact match cache for frequently
used flows. It uses the RSS hash (currently implemented only for netdev-dpdk)
to search the miniflow in an hash table.
While there might be still some tuning left to do, these patch provides
significant performance improveme
This new function uses ofpbuf_rss(), or hashes the packet 5-tuple, to get the
hash of a packet. Used by OVS_ACTION_ATTR_HASH and future commits.
Signed-off-by: Daniele Di Proietto
---
lib/dpif-netdev.c | 29 ++---
1 file changed, 22 insertions(+), 7 deletions(-)
diff --g
On Sun, Aug 17, 2014 at 7:43 PM, Pravin Shelar wrote:
> On Fri, Aug 15, 2014 at 3:12 AM, Andy Zhou wrote:
>> Future pathces will make use of those functions.
>>
>> Signed-off-by: Andy Zhou
>> ---
>> datapath/Modules.mk | 1 +
>> datapath/actions.c | 27
>> datapath/actions.h
This is what I should do if I understand your intention correctly:
1. Change this patch to remove all loop detection mechanisms.
2. Submit a separate patch to the compact code that check for a lower
xmit_recursion limit in rpl_dev_queue_xmit().
If this is not what you had in mind, please let me k
Thanks!
On 8/18/14, 1:23 PM, "Ben Pfaff" wrote:
>On Mon, Aug 18, 2014 at 12:57:42PM -0700, Daniele Di Proietto wrote:
>> fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could
>>block
>> in the "slow path" case. This commit adds fat_rwlock_try_get_slot__()
>>which
>> does not bl
Sorry about those,
I will send an updated version soon.
Thanks,
Daniele
On 8/18/14, 9:45 AM, "Ben Pfaff" wrote:
>On Fri, Aug 15, 2014 at 12:43:35PM -0700, Daniele Di Proietto wrote:
>> ofpbuf_rss() is introduced to get the RSS hash from the NIC. It works
>>only for
>> packets received by DPDK
On Mon, Aug 18, 2014 at 12:57:42PM -0700, Daniele Di Proietto wrote:
> fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could block
> in the "slow path" case. This commit adds fat_rwlock_try_get_slot__() which
> does not block, even in the "slow path" case.
>
> This fixes a minor
On Mon, Aug 18, 2014 at 07:47:53PM +, Daniele Di Proietto wrote:
> On 8/18/14, 11:47 AM, "Ben Pfaff" wrote:
>
> >On Mon, Aug 18, 2014 at 11:26:22AM -0700, Daniele Di Proietto wrote:
> >> fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could
> >>block
> >> in the "slow path"
fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could block
in the "slow path" case. This commit adds fat_rwlock_try_get_slot__() which
does not block, even in the "slow path" case.
This fixes a minor issue in dpif-netdev: when the datapath has no registered
upcall handler (e.g.
On 8/18/14, 11:47 AM, "Ben Pfaff" wrote:
>On Mon, Aug 18, 2014 at 11:26:22AM -0700, Daniele Di Proietto wrote:
>> fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could
>>block
>> in the "slow path" case. This commit adds fat_rwlock_try_get_slot__()
>>which
>> does not block, eve
On Mon, Aug 18, 2014 at 10:37 AM, Ben Pfaff wrote:
> On Mon, Aug 18, 2014 at 10:23:55AM -0700, Daniele Di Proietto wrote:
>> Commit d731058395cb used vswitchd_ovs_vswitchd_LDFLAGS to link with DPDK.
>> This
>> made automake ignore AM_LDFLAGS while linking the daemon.
>>
>> This commit explicitly
Thank you Jesse! I'll ask our SREs for a place I can test this safely.
Just to try, I attempted to apply it against the 1.10 branch, and was met
with the following on compilation:
./lib/packets.c: In function ‘packet_update_csum128’:
../lib/packets.c:577:43: error: ‘l4_size’ undeclared (first us
On Mon, Aug 18, 2014 at 11:26:22AM -0700, Daniele Di Proietto wrote:
> fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could block
> in the "slow path" case. This commit adds fat_rwlock_try_get_slot__() which
> does not block, even in the "slow path" case.
>
> Signed-off-by: Dani
fat_rwlock_tryrdlock() used to call fat_rwlock_get_slot__() which could block
in the "slow path" case. This commit adds fat_rwlock_try_get_slot__() which
does not block, even in the "slow path" case.
Signed-off-by: Daniele Di Proietto
---
lib/fat-rwlock.c | 33 -
I just sent a patch to the list. Please, let me know if this solves all
the issues.
Thanks,
Daniele
On 8/18/14, 10:31 AM, "Ben Pfaff" wrote:
>On Mon, Aug 18, 2014 at 10:28:44AM -0700, Gurucharan Shetty wrote:
>> On Mon, Aug 18, 2014 at 9:51 AM, Ben Pfaff wrote:
>> > On Mon, Aug 18, 2014 at 07
On Mon, Aug 18, 2014 at 10:23:55AM -0700, Daniele Di Proietto wrote:
> Commit d731058395cb used vswitchd_ovs_vswitchd_LDFLAGS to link with DPDK. This
> made automake ignore AM_LDFLAGS while linking the daemon.
>
> This commit explicitly adds AM_LDFLAGS to vswitchd_ovs_vswitchd_LDFLAGS and
> moves
Commit d731058395cb used vswitchd_ovs_vswitchd_LDFLAGS to link with DPDK. This
made automake ignore AM_LDFLAGS while linking the daemon.
This commit explicitly adds AM_LDFLAGS to vswitchd_ovs_vswitchd_LDFLAGS and
moves it to vswitchd/automake.mk
Reported-by: Gurucharan Shetty
Suggested-by: Ben P
On Mon, Aug 18, 2014 at 10:28:44AM -0700, Gurucharan Shetty wrote:
> On Mon, Aug 18, 2014 at 9:51 AM, Ben Pfaff wrote:
> > On Mon, Aug 18, 2014 at 07:48:10AM -0700, Gurucharan Shetty wrote:
> >> Commit d73105839 (netdev-dpdk: Move to DPDK 1.7.0) added
> >> 'vswitchd_ovs_vswitchd_LDFLAGS' in acincl
On Thu, Aug 14, 2014 at 3:23 PM, Mark Maglana wrote:
> Thank you for the quick feedback and the corrections. This looks good.
I applied this to master. Thanks!
>
>
> On Thu, Aug 14, 2014 at 1:05 PM, Gurucharan Shetty
> wrote:
>>
>> From: Mark Maglana
>>
>> The README was not clear on whether th
On Mon, Aug 18, 2014 at 9:51 AM, Ben Pfaff wrote:
> On Mon, Aug 18, 2014 at 07:48:10AM -0700, Gurucharan Shetty wrote:
>> Commit d73105839 (netdev-dpdk: Move to DPDK 1.7.0) added
>> 'vswitchd_ovs_vswitchd_LDFLAGS' in acinclude.m4. This caused the
>> created Makefile to use 'vswitchd_ovs_vswitchd_L
On Mon, Aug 18, 2014 at 9:54 AM, Ben Pfaff wrote:
> On Mon, Aug 18, 2014 at 07:48:11AM -0700, Gurucharan Shetty wrote:
>> Commit b7ea2d480338(Extend OVS IPFIX exporter to export tunnel headers)
>> added a usage for IPROTO_GRE. But that does not look available in any
>> Visual Studio headers. So ad
On Mon, Aug 18, 2014 at 04:44:58PM +1200, Joe Stringer wrote:
> On 13 August 2014 06:24, Ben Pfaff wrote:
> >
> > dp_netdev_flow_add() goes to some trouble to log matches as if they
> > were provided as datapath flows, even though they aren't. I think
> > that match_format() would be simpler and
On Mon, Aug 18, 2014 at 07:48:11AM -0700, Gurucharan Shetty wrote:
> Commit b7ea2d480338(Extend OVS IPFIX exporter to export tunnel headers)
> added a usage for IPROTO_GRE. But that does not look available in any
> Visual Studio headers. So add it.
>
> Signed-off-by: Gurucharan Shetty
I was very
On Mon, Aug 18, 2014 at 07:48:10AM -0700, Gurucharan Shetty wrote:
> Commit d73105839 (netdev-dpdk: Move to DPDK 1.7.0) added
> 'vswitchd_ovs_vswitchd_LDFLAGS' in acinclude.m4. This caused the
> created Makefile to use 'vswitchd_ovs_vswitchd_LINK' without
> including 'AM_LDFLAGS'. This causes failu
On Fri, Aug 15, 2014 at 12:43:35PM -0700, Daniele Di Proietto wrote:
> ofpbuf_rss() is introduced to get the RSS hash from the NIC. It works only for
> packets received by DPDK (otherwise it returns 0, which tells the caller to
> compute the hash in another way)
>
> This commit also configure DPDK
On Fri, Aug 15, 2014 at 09:42:46AM -0700, Srini Seetharaman wrote:
> Current hash uses just the dl_dst field. This patch expands the hash to
> include all L2, L3 and L4 fields, allowing for more balanced selection.
>
> Signed-off-by: Srini Seetharaman
I'm going to let this one sit for a few day
Commit d73105839 (netdev-dpdk: Move to DPDK 1.7.0) added
'vswitchd_ovs_vswitchd_LDFLAGS' in acinclude.m4. This caused the
created Makefile to use 'vswitchd_ovs_vswitchd_LINK' without
including 'AM_LDFLAGS'. This causes failure while building ovs-vswitchd
with external libraries (ex: ovs-vswitchd on
Commit b7ea2d480338(Extend OVS IPFIX exporter to export tunnel headers)
added a usage for IPROTO_GRE. But that does not look available in any
Visual Studio headers. So add it.
Signed-off-by: Gurucharan Shetty
---
include/windows/netinet/in.h | 22 ++
1 file changed, 22 inse
On Mon, Aug 18, 2014 at 06:44:32AM -0400, Thomas F Herbert wrote:
> I have been working on 802.1ad qinq and I have been puzzled by this
> construct in odp-util.c
> where vlan sets are put in the database with a pop, a set of the tci
> into the struct and then a push. I think this is because the pus
On Aug 18, 2014, at 7:19 AM, Nithin Raju
wrote:
> +ovs-dev
>
> hi Sam/Alin,
> I want to propose that I can go ahead get my patches committed, since we am
> working on more patches that are dependent on this. Any review comments that
> I get from you folks, I'll definitely take a look.
>
> Is
Hi Nithin,
Sam is on PTO until Wednesday.
I will try to review your patches until the end of the day today.
You could do a longer patch set with the ones that are depending with the
previous patches.
Thanks,
Alin.
-Mesaj original-
De la: Nithin Raju [mailto:nit...@vmware.com]
Trimis:
+ovs-dev
hi Sam/Alin,
I want to propose that I can go ahead get my patches committed, since we am
working on more patches that are dependent on this. Any review comments that I
get from you folks, I'll definitely take a look.
Is this OK?
thanks,
Nithin
On Aug 15, 2014, at 9:53 AM, Nithin Raj
Just to be clear, I'm planning on implementing it when it's available as a DPDK
library rather than it being a hardened openvswitch plan.
Regards
Maryam
-Original Message-
From: Tahhan, Maryam
Sent: Monday, August 18, 2014 9:23 AM
To: 'loy wolfe'; dev@openvswitch.org
Subject: RE: [ovs-d
All:
I have been working on 802.1ad qinq and I have been puzzled by this
construct in odp-util.c
where vlan sets are put in the database with a pop, a set of the tci
into the struct and then a push. I think this is because the push vlan
is the
only way to pass a vlan_tci through the dpif to th
Hi
The plan is to support vhost-user as soon as it's integrated in DPDK. It's
currently being integrated into DPDK as a library.
Regards
Maryam
-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of loy wolfe
Sent: Monday, August 18, 2014 9:17 AM
To: dev@openvswi
Is there any plan of DPDK-OVS support upstream vhost-user? It is merged in
qemu and libvirt 1.2.7, also with vif-vhostuser BP in openstack. I think
this will accelerate adoption of dpdkovs use in openstack.
___
dev mailing list
dev@openvswitch.org
http://
On Sun, Aug 17, 2014 at 8:19 PM, Wenyu Zhang wrote:
> Extend IPFIX exporter to export tunnel headers when both input and output
> of the port.
> Add three other_config options in IPFIX table: enable-input-sampling,
> enable-output-sampling and enable-tunnel-sampling, to control whether
> sampling
61 matches
Mail list logo
