On Thu, Feb 22, 2018 at 10:54:05PM +0100, Jakub Sitnicki wrote:
> Use the logical switch/router UUID as hash input, instead of the UUID of
> the Datapath_Binding row, when calculating the hash value for lflows in
> the SBDB.
>
> Otherwise the hash value will never match the one computed from NBDB
In one place, ovn-northd was hashing Logical_Switch or Logical_Router UUIDs
for ovn_lflow, and in another place it was hashing Datapath_Binding UUIDs.
This caused problems. This commit changes ovn-northd to always hash the
Datapath_Binding UUID.
Jacob Sitnicki reported the following performance
Hi Yipeng,
>If we store pointers in DFC, then the memory requirement is large. When
>there are VMs or multiple PMDs running on the same platform, they will
>compete the shared cache. So we want DFC to be as memory efficient as
>possible.
>
>Indirect table is a simple hash table that map the
On Thu, Feb 22, 2018 at 10:30:37AM +, Markos Chandras wrote:
> Hello,
>
> This patchset is a backport of the following python-3 fixes
> https://mail.openvswitch.org/pipermail/ovs-dev/2017-December/342460.html
> for branch 2.8
>
> Markos Chandras (4):
> ovsdb: Use items() instead of
This is great! I will re-run the test with this patch and send over the
results soon!
On Fri, Feb 23, 2018 at 3:25 PM, Han Zhou wrote:
> On Fri, Feb 23, 2018 at 2:03 PM, Ben Pfaff wrote:
> >
> > ofproto_port_open_type() was surprisingly slow because it called
For the patchset:
Acked-By: Mark Michelson
On 02/20/2018 11:39 AM, Lorenzo Bianconi wrote:
Changes since v1:
- removed 1/3 of v1 since it has been already merged
- added ovn automatic tests for icmp4 action and for ACL reject rules
- used tos of incoming IP frames for
On Wed, Feb 21, 2018 at 03:09:00PM +0300, Ilya Maximets wrote:
> 'upcall_xlate()' makes a decision to compose slow path actions
> by checking the 'upcall->fitness', which is not initialized in
> case of calling from the 'upcall_cb()'.
>
> 'upcall_cb()' receives the real flow, so the fitness
On Fri, Feb 23, 2018 at 2:03 PM, Ben Pfaff wrote:
>
> ofproto_port_open_type() was surprisingly slow because it called the
> function ofproto_class_find__(), which itself was surprisingly slow
because
> it actually creates a set of strings and enumerates all of the available
>
On Tue, Feb 20, 2018 at 02:04:02PM +0100, Zoltan Balogh wrote:
> Signed-off-by: Zoltan Balogh
> CC: Zoltan Balogh
Thanks, I applied this to master.
Any given person is only supposed to be in one list or the other, so I
changed the patch
On Thu, Feb 15, 2018 at 04:49:57PM -0600, Mark Michelson wrote:
> The ovn-nbctl, ovn-sbctl, and ovs-vsctl manpages are inconsistent in
> their "Database Commands" section when it comes to referring to what
> database tables exist. This commit amends this by making each *ctl
> manpage reference the
On Fri, Feb 23, 2018 at 01:15:38PM -0600, Mark Michelson wrote:
> The ofport_is_internal_or_patch() function called
> ofproto_port_open_type() twice. This resulted in the ofproto_class being
> looked up twice, when we only needed to look it up once.
>
> This patch alters
On Wed, Feb 21, 2018 at 03:10:54PM -0800, Gregory Rose wrote:
> On 2/21/2018 2:54 PM, William Tu wrote:
> >Trivial grammer fix.
> >
> >Signed-off-by: William Tu
> >---
> > datapath/linux/compat/include/net/ip_tunnels.h | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
On Sun, Feb 18, 2018 at 08:51:16PM +0800, Guoshuai Li wrote:
> It can only receive 4096 bytes of data each time in jsonrpc,
> when there are similar and Chinese characters occupy multiple bytes,
> it may receive half a character, this time the decoding will be abnormal.
> We need to receive the
From: zhangliping
If the format specifier does not have the 0 flag, we should pad with
blanks instead of zeroes.
Signed-off-by: zhangliping
---
lib/vlog.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/vlog.c
On 19 February 2018 at 07:14, Aaron Conole wrote:
> The selinux policy currently builds manually, as a process that either
> the user or distribution maintainer undertakes. That process consists
> of:
>
> 1. Convert the intermediary files into their file form through
>
On 19 February 2018 at 06:55, Aaron Conole wrote:
> Newer libvirt and openstack versions will now label the unix socket as
> an `svirt_tmpfs_t` object. This means that in order to support
> deploying with the recommended configuration (using a
> dpdkvhostuserclient socket),
I Am Mr Pedro, Powerball Lotter Winner. I have a Donation Award for you, Email
me back on my personal email below for more info: peddroquez...@gmail.com
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
Jakub and I discussed this patch offline and I am now on board with this
patch.
Acked-By: Mark Michelson
Tested-By: Mark Michelson
On 02/23/2018 08:42 AM, Mark Michelson wrote:
Hi Jakub,
This patch fixes the issue. However, I think the resulting
On Thu, Feb 22, 2018 at 08:29:11PM +0530, Satyavalli Rama wrote:
> Jan and Ben we requesting you both kindly provide your valuable inputs with
> respect new enhancement i.e adding --oxs-stats option to the existing
> commands.
> If everything mentioned above is fine then we will share the
The ofport_is_internal_or_patch() function called
ofproto_port_open_type() twice. This resulted in the ofproto_class being
looked up twice, when we only needed to look it up once.
This patch alters ofport_is_internal_or_patch() to look up the
ofproto_class in-line, and then use the looked up
Hi Han,
First off, great work! This is fantastic to see, and I'm super excited
to see such a drop in CPU usage with this patch.
Since this is an RFC patch, I won't go too deep into nitty-gritty
details on this. Rather, I'll look at it from a high level.
The API design idea is great. I like
Hi Mark, thank you very much for the feedback!!
On Fri, Feb 23, 2018 at 10:56 AM, Mark Michelson
wrote:
>
> Hi Han,
>
> First off, great work! This is fantastic to see, and I'm super excited to
see such a drop in CPU usage with this patch.
>
> Since this is an RFC patch, I
Hi Yipeng,
Thanks for the patch. Some high level questions/comments.
(1) Am I right in understanding that this patch *only* introduces a new cache
approach in to DFC to reduce the collisions?
(2) Why the number of entries per Bucket is set to '8'? With this each
dfc_bucket size is 80
On Fri, Feb 23, 2018 at 3:24 AM, Jakub Sitnicki wrote:
> Use the logical switch/router UUID as hash input, instead of the UUID of
> the Datapath_Binding row, when calculating the hash value for lflows in
> the SBDB.
>
> Otherwise the hash value will never match the one computed
On Thu, 22 Feb 2018 18:07:55 +0100
Jakub Sitnicki wrote:
> On Wed, Feb 14, 2018 at 09:54 PM GMT, Ben Pfaff wrote:
[...]
> > +uint32_t
> > +sbrec_logical_flow_hash(const struct sbrec_logical_flow *lf)
> > +{
> > +const struct sbrec_datapath_binding *ld =
Hi Jakub,
This patch fixes the issue. However, I think the resulting code can lead
to some confusion. The name of the first parameter given to
ovn_logical_flow_hash() is called "logical_datapath". The result of your
patch is that the actual first parameter passed to that function is
either a
26 matches
Mail list logo