Sure that makes sense. I guess what's a little weird is that the load
balancer lives on the logical switch instead of the logical router.
Typically, when I think of a load balancer in the real world, it *is*
a router that just happens to rewrite IP addresses. It wouldn't be a
router behind a
.com>
>
> --Justin
>
>
> > On Sep 1, 2015, at 10:38 AM, Ethan J. Jackson <et...@nicira.com> wrote:
> >
> > From: Ethan Jackson <e...@eecs.berkeley.edu>
> >
> > Signed-off-by: Ethan J. Jackson <e...@eecs.berkeley.edu>
> > ---
> > AU
Yeah, there's another Ethan Jackson at MSR unfortunately.
On Tue, Sep 1, 2015 at 10:42 AM, Ben Pfaff <b...@nicira.com> wrote:
> On Tue, Sep 01, 2015 at 10:38:26AM -0700, Ethan J. Jackson wrote:
>> From: Ethan Jackson <e...@eecs.berkeley.edu>
>>
>> S
agreed, sent a new version.
Ethan
On Wed, Aug 12, 2015 at 5:01 PM, Jarno Rajahalme jrajaha...@nicira.com wrote:
On Aug 12, 2015, at 4:13 PM, Ethan J. Jackson et...@nicira.com wrote:
From: Ethan Jackson et...@nicira.com
Future patches will need to modify ukey actions in some instances
wrote:
On Aug 10, 2015, at 6:46 PM, Ethan J. Jackson et...@nicira.com wrote:
From: Ethan Jackson et...@nicira.com
Since revalidator_sweep() doesn't hold the ukey mutex for each full
loop iteration, it's theoretically possible that two threads may
call ukey_delete() on the same ukey
in
the case we currently know suffers from the lack of modify support?
Also, there are some minor comments below you could address at the same time.
Jarno
On Aug 5, 2015, at 4:07 PM, Ethan Jackson et...@nicira.com wrote:
There are certain use cases (such as bond rebalancing) where a
datapath flow's
Great, I'll merge both today.
Thanks a lot,
Ethan
On Thu, Aug 6, 2015 at 9:00 AM, Stokes, Ian ian.sto...@intel.com wrote:
No problem,
I've sent a patch to the mailing list with these changes for INSTALL.DPDK.md.
Thanks
Ian
-Original Message-
From: Ethan Jackson [mailto:et
with
the MODIFY flag set.
Signed-off-by: Ethan Jackson et...@nicira.com
---
ofproto/ofproto-dpif-upcall.c | 118 +++---
1 file changed, 87 insertions(+), 31 deletions(-)
diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
index 59010c2..7abc97d
the issue.
Thanks
Ian
-Original Message-
From: Ben Pfaff [mailto:b...@nicira.com]
Sent: Monday, August 03, 2015 6:11 PM
To: Ethan Jackson
Cc: Stokes, Ian; Justin Pettit; Pravin Shelar; Traynor, Kevin; Daniele
Di Proietto; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH 2/2] netdev-dpdk
Hey,
Really sorry this has taken so long to merge. I dropped the ball. It's in now.
Ethan
On Tue, Aug 4, 2015 at 5:26 AM, Ilya Maximets i.maxim...@samsung.com wrote:
Will anyone plan to apply this patch?
Best regards, Ilya Maximets.
On 28.07.2015 23:48, Flavio Leitner wrote:
On Tue, Jul
.
Not until we have more explicit configuration details available for the HW
device from DPDK.
Thanks
Ian
-Original Message-
From: Ethan Jackson [mailto:et...@nicira.com]
Sent: Wednesday, July 29, 2015 10:13 PM
To: Stokes, Ian
Cc: Traynor, Kevin; Daniele Di Proietto; dev@openvswitch.org
Sorry for taking so long to get to this. The one question I have is:
Is OVS the right layer to be fixing this? Isn't this really an issue
of DPDK reporting a number of available queues that for practical
purposes is wrong? I.E. Shouldn't this be fixed by the DPDK driver of
this system? This
.
However, maybe for your use-case this isn’t an issue?
Yeah for our (Melvin + Me) specific use case, we'll just do (a) and
not free or allocate packets, shouldn't be too hard for us.
Ethan
Signed-off-by: Melvin Walls mwall...@gmail.com
Signed-off-by: Ethan Jackson et...@nicira.com
---
lib
Ah my bad sorry.
Acked-by: Ethan Jackson et...@nicira.com
On Tue, Jul 28, 2015 at 6:53 PM, Ben Pfaff b...@nicira.com wrote:
Fixes the following error:
The following files are in git but not the distribution:
ovn/OVN-GW-HA.md
CC: Ethan Jackson et...@nicira.com
Signed-off-by: Ben
Ben, Justin, should this be backported? I'm not up on the policy at the moment.
Etha
On Thu, Jul 23, 2015 at 3:42 AM, Traynor, Kevin kevin.tray...@intel.com wrote:
-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Daniele Di
Proietto
Sent: Thursday,
and we think that the cause is this patch.
When we tried branch-2.4, we ran into no errors.
Today I am working on confirming that this is the patch.
Luis E. P.
Sent from my iPhone
On Jul 23, 2015, at 15:38, Ethan Jackson et...@nicira.com wrote:
Ben, Justin, should this be backported
-0700, Ethan Jackson wrote:
Sort of. Helped with the creation of the patch a bit so I figured I
should add a sign off. Someone other than me should review it for
this reason.
Ethan
On Wed, Jul 22, 2015 at 3:07 PM, Ben Pfaff b...@nicira.com wrote:
On Wed, Jul 22, 2015 at 01:51:46PM -0700
: netdev_dpdk_ring_rxq_recv() and netdev_dpdk_ring_send__() to
handle
tx/rx on dpdk rings in secondary processes.
Signed-off-by: Melvin Walls mwall...@gmail.com
Signed-off-by: Ethan Jackson et...@nicira.com
These sign-offs are weird. Ethan, is this Signed-off-by meant to be an
Acked-by from
, Jul 22, 2015 at 04:34:39PM -0700, Ethan Jackson wrote:
Sure, Melvin could you add that to the next version you send out
(after this version is reviewed).
Ethan
On Wed, Jul 22, 2015 at 4:31 PM, Ben Pfaff b...@nicira.com wrote:
I understand now. I recommend adding a Co-authored
Thanks for the review, I've addressed the comments and will send out
another version of the patch shortly. One comment on the feedback
below.
Under Controller Independent Active-backup, I am not sure that I buy
the argument here, because currently ovn-northd doesn't care about the
layout of
No worries. I sent a new version (actually ran spell check on this one ;) )
On Tue, Jul 21, 2015 at 11:35 AM, Ben Pfaff b...@nicira.com wrote:
On Tue, Jul 21, 2015 at 11:32:21AM -0700, Ethan Jackson wrote:
Thanks for the review, I've addressed the comments and will send out
another version
similar systems. The hope is that it can be used as
a starting point for design discussions and an eventual
implementation.
Signed-off-by: Ethan Jackson et...@nicira.com
---
ovn/OVN-GW-HA.md | 375 +++
1 file changed, 375 insertions(+)
create mode
similar systems. The hope is that it can be used as
a starting point for design discussions and an eventual
implementation.
Signed-off-by: Ethan Jackson et...@nicira.com
---
OVN-GW-HA.md | 374 +++
1 file changed, 374 insertions(+)
create mode
Acked-by: Ethan Jackson et...@nicira.com
Merged. Changed some tabs to spaces
Ethan
On Thu, Jul 2, 2015 at 12:34 PM, Luis E. P lp...@vmware.com wrote:
Signed-off-by: Luis E. P lp...@vmware.com
---
AUTHORS| 1 +
datapath/README.md | 7 ---
2 files changed, 5 insertions(+), 3
The patch introduces some trailing whitespace.
- - If userspace's notion of the flow key for the packet matches the
+ - If the userspace's notion of the flow key for the packet matches the
I don't agree with this change. It's pretty common to talk about
userspace without a the as the README
Typically adding yourself to the authors file wouldn't be a separate
patch, I'd just merge this into the previous one.
Also, Looking at the file now the first list of people are those who
have authored or signed off on commits in the Open vSwitch source
code. I think the patch you suggested
Also, there is a typo in the subject.
Doh, I should have caught that.
Ethan
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
a v3 in the next patch.
--Justin
On Jul 2, 2015, at 12:01 PM, Luis E Pena lp...@vmware.com wrote:
Should I redo this patch with my name in the first list?
On Jul 2, 2015, at 11:15, Ben Pfaff b...@nicira.com wrote:
On Thu, Jul 02, 2015 at 11:13:54AM -0700, Ethan Jackson wrote:
Also
Acked-by: Ethan Jackson et...@nicira.com
On Tue, Jun 16, 2015 at 10:53 AM, Flavio Leitner f...@sysclose.org wrote:
On Tue, Jun 16, 2015 at 08:47:34AM -0700, Ben Pfaff wrote:
The packed annotation doesn't do anything here because all of the members
in the structure are naturally aligned
If you tried it and it didn't improve, then I think what we have now is great.
Ethan
On Tue, Jun 16, 2015 at 9:08 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
On 16/06/2015 09:55, Gray, Mark D mark.d.g...@intel.com wrote:
Acked-by: Ethan Jackson et...@nicira.com
One question I
Acked-by: Ethan Jackson et...@nicira.com
One question I had for a future patch. Have you considered
prefetching more packets at once? I.E. 4 at a time or something?
That's how these things are typically written, though I don't know if
it would actually affect things.
Merged.
Ethan
On Mon, Jun
I have some general questions about the vhost user feature which
aren't totally clear to me. I would greatly appreciate some
clarification, or pointers to documentation.
1) Since vhost-user is using standard Qemu interfaces, will it support
all of the features of the standard vhost
Great thanks a lot for the detailed response, this is super helpful.
I'll take a look at the presentation and read the documentation.
Ethan
On Wed, Jun 3, 2015 at 3:09 PM, Gray, Mark D mark.d.g...@intel.com wrote:
Hi Ethan
I have some general questions about the vhost user feature which
Acked-by: Ethan Jackson et...@nicira.com
Merged.
Ethan
On Wed, Jun 3, 2015 at 10:45 AM, Flavio Leitner f...@sysclose.org wrote:
On Wed, Jun 03, 2015 at 03:55:16PM +0100, Daniele Di Proietto wrote:
Non pmd threads have a core_id == UINT32_MAX, while queue ids used by
netdevs range from 0
This is useful and I think we would like to use it here. I have a few
questions.
Great! I really hope this moves the community in the direction of
more realistic test cases.
How come the pipeline is set up with metadata instead of a chain of tables
as I presume that would also work?
I
-by: Daniele Di Proietto diproiet...@vmware.com
On 19/05/2015 22:34, Ethan Jackson et...@nicira.com wrote:
I think you're running into the issue that sparse doesn't understand
the gnu_inline macro which is used extensively in the DPDK header
files. If you run tip of master sparse
(git
Pravin has been handling this on our end, so he'll do it.
Btw how does this perform out of curiosity?
Ethan
On Fri, May 22, 2015 at 1:33 PM, Flavio Leitner f...@sysclose.org wrote:
On Fri, May 22, 2015 at 04:40:00PM +0100, Ciara Loftus wrote:
This patch adds support for a new port type to the
Acked-by: Ethan Jackson et...@nicira.com
On Fri, May 22, 2015 at 9:14 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
DPDK lcore_id is unsigned. We need to support big values like
LCORE_ID_ANY (=UINT32_MAX). Therefore I am changing the type everywhere
in OVS.
Signed-off-by: Daniele
Acked-by: Ethan Jackson et...@nicira.com
On Fri, May 22, 2015 at 9:14 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
We used to reserve DPDK lcore 0 for non pmd operations, making it
difficult to use core 0 for packet processing.
DPDK 2.0 properly support non EAL threads with lcore
Acked-by: Ethan Jackson et...@nicira.com
On Fri, May 22, 2015 at 9:14 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
Right now ethernet and ring devices use a mutex, while vhost devices use
a mutex or a spinlock to protect statistics. This commit introduces a
single spinlock that's
In netdev_dpdk_send__() I folded in an OVS_UNLIKELY around the
txq_needs_locking check as I'd like to optimize for the standard case
where we don't. Otherwise looks good to me, I'll merge this series
shortly.
Acked-by: Ethan Jackson et...@nicira.com
On Fri, May 22, 2015 at 9:14 AM, Daniele Di
Thanks, I've merged.
On Thu, May 21, 2015 at 1:44 PM, Pravin Shelar pshe...@nicira.com wrote:
On Wed, May 20, 2015 at 5:06 PM, Ethan Jackson et...@nicira.com wrote:
When executing actions, it's possible a recirculation will occur
causing dp_netdev_input() to be called multiple times
Daniele/Vasmi,
You were the last two on our end to stand up ivshmem. Would you mind
resubmitting this patch with whatever tweaks you needed to make it
work?
Ethan
On Thu, May 21, 2015 at 12:41 PM, Flavio Leitner f...@redhat.com wrote:
On Wed, May 20, 2015 at 01:22:18PM -0700, Ethan Jackson
Applied, thanks!
On Thu, May 21, 2015 at 9:26 AM, Kevin Traynor kevin.tray...@intel.com wrote:
This patch simplifies Rx/Tx NIC configuration by removing
custom values and using the defaults provided by the DPDK
PMDs. This also enables Rx vectorisation which improves
performance.
v2: rebase
Applied, thanks!
Ethan
On Thu, May 21, 2015 at 8:38 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
I totally agree with the suggested change. Should the need arise,
we could make it configurable later.
Acked-by: Daniele Di Proietto diproiet...@vmware.com
On 13/05/2015 14:54, Ciara
Hey guys,
I know this is a super old thread, but is there any interest in
reviving it? On our end we've had a terrible time getting ivshmem
working until we found this documentation. I expect future users will
have the same problem.
I'd resubmit the patch myself, but I have none of the
lookup. This could be very bad.
This patch fixes the problem by zeroing out flow batch pointers before
calling packet_batch_execute(). This probably has a slightly negative
performance impact, though I haven't tried it.
Signed-off-by: Ethan Jackson et...@nicira.com
---
lib/dpif-netdev.c | 5 -
1
Cool thanks for trying it out. I'll merge soon.
Ethan
On Mon, May 18, 2015 at 1:22 PM, Traynor, Kevin kevin.tray...@intel.com wrote:
-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ethan Jackson
Sent: Monday, May 18, 2015 5:08 PM
To: dev
Cool thanks for trying it out. I'll merge soon.
Ethan
On Mon, May 18, 2015 at 1:22 PM, Traynor, Kevin kevin.tray...@intel.com wrote:
-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ethan Jackson
Sent: Monday, May 18, 2015 5:08 PM
To: dev
Thanks I'll merge soon
On Mon, May 18, 2015 at 1:26 PM, Daniele Di Proietto
diproiet...@vmware.com wrote:
Acked-by: Daniele Di Proietto diproiet...@vmware.com
On 18/05/2015 16:53, Ethan Jackson et...@nicira.com wrote:
These are all minor style issues.
Signed-off-by: Ethan Jackson et
This version of the patch breaks sparse, I sent out another.
Ethan
On Sat, May 16, 2015 at 11:24 AM, Ethan Jackson et...@nicira.com wrote:
The MAX_PKT_BURST and NETDEV_MAX_RX_BATCH macros had a confusing
relationship. They basically purport to do the same thing, making it
unclear which
These are all minor style issues.
Signed-off-by: Ethan Jackson et...@nicira.com
---
lib/netdev-dpdk.c | 22 ++
1 file changed, 14 insertions(+), 8 deletions(-)
diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
index 505ab75..5f8c60f 100644
--- a/lib/netdev-dpdk.c
+++ b/lib
Sparse doesn't like several of the DPDK header files. This patch
works around it so we can get analysis when compiling DPDK.
Signed-off-by: Ethan Jackson et...@nicira.com
---
include/sparse/automake.mk | 4
include/sparse/emmintrin.h | 21 +
include/sparse
-by: Ethan Jackson et...@nicira.com
---
lib/dpif-netdev.c | 16
lib/netdev-dpdk.c | 9 +++--
lib/netdev.h | 2 +-
3 files changed, 12 insertions(+), 15 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index f1d65f5..96c71ac 100644
--- a/lib/dpif-netdev.c
-by: Ethan Jackson et...@nicira.com
---
lib/dpif-netdev.c | 10 +-
lib/netdev-dpdk.c | 7 ++-
lib/netdev.h | 2 +-
3 files changed, 8 insertions(+), 11 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index f1d65f5..4216865 100644
--- a/lib/dpif-netdev.c
+++ b/lib
Hi Ian,
This email is in response to your QoS patch here:
http://openvswitch.org/pipermail/dev/2015-January/050642.html
I don't seem to have received it, which makes me thing the email was
sent during that brief window when the dev list was down. Apologies
for getting to it so late.
Anyways,
Would we care about making it run also with python3? Only 3 changes
are required (see inline).
I don't think we care. At least none of the other python code in the
OVS tree does, this seems like an arbitrary place to start worrying
about it.
I think it would be useful to add an option to
Is it useful without accompanying code to generate appropriate packets
to exercise the flows?
Yes.
Daniele is using it with at the moment by running packet traces
directly through the system. Plus a lot of synthetic workloads we use
regularly would be more characteristic if run on something
virtualization workloads.
Signed-off-by: Ethan Jackson et...@nicira.com
---
utilities/automake.mk| 5 +-
utilities/ovs-pipegen.py | 122 +++
2 files changed, 125 insertions(+), 2 deletions(-)
create mode 100755 utilities/ovs-pipegen.py
diff --git
I really don't think this should be configurable, I can't imagine a
situation in which a user would have enough knowledge to tweak the
knob correctly. I think the only reason we'd want to configure it is
the current EMC size is too small at the moment. My preference would
be to figure out how to
I'll merge this shortly.
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 15, 2015 at 11:11 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
As stated by the comment above the structure, the 'action' pointer does not
change during the 'dp_netdev_actions' lifetime: we might as well
Looks like a big win.
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 15, 2015 at 11:11 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
Now that we have per packet metadata, there's no need to split packet
batches when recirculating.
Signed-off-by: Daniele Di Proietto diproiet
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 15, 2015 at 11:11 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
We already have the 'dp_hash' embedded in the metadata. This caused
confusion in the code. With this commit it should be clear that
'rss_hash' is the packet hash used
value and account for the changes in the
packet or in the metadata.
Requested-by: Ethan Jackson et...@nicira.com
Signed-off-by: Daniele Di Proietto diproiet...@vmware.com
---
lib/dpif-netdev.c | 6 ++
1 file changed, 6 insertions(+)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
Acked-by: Ethan Jackson et...@nicira.com
Any reason not to just add some sampling points by default? Wouldn't
it be a useful debugging mechanism in the field?
Also btw, why is this better than perf?
Ethan
On Sat, Mar 21, 2015 at 12:00 AM, Andy Zhou az...@nicira.com wrote:
This integration
There's some trailing whitespace and minor gramatical errors which I've fixed
in the following patch. If you're willing to apply it, then go ahead and merge.
Acked-by: Ethan Jackson et...@nicira.com
---
lib/perf-counter.c | 6 +++---
lib/perf-counter.h | 10 +-
2 files changed, 8
an ovs_mutex, if
you think it’s better
On 10 Apr 2015, at 02:00, Ethan Jackson et...@nicira.com wrote:
Could you explain a little more clearly why you want to use an
rte_spinlock over an ovs_mutex? It's not entirely clear from the
commit message.
Ethan
On Thu, Apr 2, 2015 at 9:00 AM
Acked-by: Ethan Jackson et...@nicira.com
On Tue, Apr 7, 2015 at 12:02 PM, Daniele Di Proietto
diproiet...@vmware.com wrote:
A read operation from a non atomic shared value (without external
locking) can return incorrect values. Using the atomic semantics
prevents this from happening
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 1, 2015 at 9:21 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
This init function is called when the dpif class is registered. It will
be used by following commits
Signed-off-by: Daniele Di Proietto diproiet...@vmware.com
---
lib
I don't really like the names of the two new functions, they're
verbose and don't convey much information (one or the other is fine,
both less so). I'm not sure what would be better, but off the top of
my head let me suggest something like:
cycle_reset()
cycle_tally()
Think about it, there's
I think this is fine. Though I agree with Justin the script
documentation should be update. I can take care of that tomorrow.
Acked-by: Ethan Jackson et...@nicira.com
On Sun, Apr 5, 2015 at 1:40 PM, Ben Pfaff b...@nicira.com wrote:
This is hardly urgent so I'll leave this one for Ethan
Yeah I'd definitely like to see a re-spin of this with a relaxed
increment helper.
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 1, 2015 at 9:21 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
We used to count exact match cache hits and masked classifier hits
together
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 1, 2015 at 9:20 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
Since statistics updates might require locking (in future commits)
grouping them will reduce the locking overhead.
Signed-off-by: Daniele Di Proietto diproiet
, so just writing the value should be
fine.
There should be a comment in dp_netdev_flow_used() explaining why we
do updates the way we do.
Alternatively, the patch might benefit from a relaxed add helper
local to dpif_netdev which does a relaxed read, add, relaxed write.
Acked-by: Ethan Jackson
Acked-by: Ethan Jackson et...@nicira.com
On Wed, Apr 1, 2015 at 9:20 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
Since flow statistics are thread local and updated without any lock,
it is not correct to do a memset from another thread.
This commit simply removes the support
Oops nvm, looks like they ended up in my spam folder.
Ethan
On Thu, Apr 2, 2015 at 6:03 PM, Ethan Jackson et...@nicira.com wrote:
Could you please resend the emails for these patches? They showed up
on the mailing list, but for some reason they didn't end up in my
inbox. Wondering
Since statistics updates might require locking (in future commits)
grouping them will reduce the locking overhead.
Do they actually require locking in future commits or is this speculative?
Adding locks for these stats seems like a pretty significant step
backwards, have you measured the
Acked-by: Ethan Jackson et...@nicira.com
Merged, thanks.
On Fri, Mar 27, 2015 at 9:29 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
Packets for which an upcall has failed (lost packets) must be deleted.
We also need to count them as MISS and LOST.
Signed-off-by: Daniele Di Proietto
You're not going to like my review of this patch =)
So the code is very high quality, though I haven't read it super
closely yet. My main question, is if this is really the right
approach to handle stats in the DPDK fast path in general.
So here's my question: why not just use atomic variables
it should block the
patch.
Also itd be nice to switch bonds to using this same framework.
Acked-by: Ethan Jackson et...@nicira.com
On Tue, Mar 24, 2015 at 11:08 AM, Jarno Rajahalme jrajaha...@nicira.com wrote:
Thank you for the review. Ethan promised to review this as well, so I’ll wait
for his
Patch no longer applies unfortunately, daniele could you please rebase
and resend it when you have a chance?
Ethan
On Thu, Mar 12, 2015 at 11:04 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
This commit changes the semantics of 'netdev_set_multiq()' to allow OVS
DPDK to run on device
I suppose my only question with this patch is why we can't just
allocate MIN_NB_MBUF sized pools in the first place.
Acked-by: Ethan Jackson et...@nicira.com
On Thu, Mar 12, 2015 at 11:04 AM, Daniele Di Proietto
diproiet...@vmware.com wrote:
If rte_mempool_create() fails with ENOMEM, try
Solution seems clean, I'm happy with this as well.
Acked-by: Ethan Jackson et...@nicira.com
On Thu, Feb 12, 2015 at 11:02 AM, Alex Wang al...@nicira.com wrote:
Looks good to me,
/* A MAC learning table entry.
- * Guarded by owning 'mac_learning''s rwlock */
+ * Guarded by owning
before the cmap, all the
values are guaranteed to be initialized during flow dumps. With this
patch, I can no longer reproduce the crash.
Signed-off-by: Ethan Jackson et...@nicira.com
---
lib/dpif-netdev.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/dpif-netdev.c
and you can tell me what you think.
Ethan
On Sat, Jan 3, 2015 at 2:53 PM, Ethan Jackson et...@nicira.com wrote:
Before this patch, dp_netdev_flow_add() inserted newly minted flows in
the flow_table cmap before inserting them into the per core dpcls
classifier. Since dpcls_insert() initializes
The current implementation is not standards compliant and inferior to
BFD in every way. This patch marks it as deprecated so it may be
removed in the future.
Signed-off-by: Ethan Jackson et...@nicira.com
---
lib/cfm.c| 7 +++
vswitchd/vswitch.xml | 4
2 files changed, 11
I'm doing some performance testing which heavily stresses the OVS DPDK slow
path, and notice we spend a ridiculous amount of CPU time on the rule_dpif stats
mutex. For my test case, by commenting out the mutex I literally doubled the
performance. I'm in a bit of a rush, so as a short term hack,
In our experience, most people actually expect this number to be
fairly high. For example, our traffic generator runs a little bit of
traffic to populate caches, pauses for a couple of seconds, before
finally starting the main test. On top of that, we've had some minor
complaints from people
.
Signed-off-by: Ethan Jackson et...@nicira.com
---
ofproto/ofproto.c | 8
1 file changed, 8 insertions(+)
diff --git a/ofproto/ofproto.c b/ofproto/ofproto.c
index 7b1d478..754d290 100644
--- a/ofproto/ofproto.c
+++ b/ofproto/ofproto.c
@@ -1522,6 +1522,14 @@ ofproto_run(struct ofproto *p
FYI All I've dropped this patch.
Ethan
On Mon, Sep 8, 2014 at 2:42 PM, Ben Pfaff b...@nicira.com wrote:
On Mon, Sep 08, 2014 at 12:51:11PM -0700, Ethan Jackson wrote:
Is there any point to testing OVS in a configuration that we won't
distribute?
This is a subject in which there could
On Wed, Sep 17, 2014 at 1:42 PM, Ben Pfaff b...@nicira.com wrote:
On Wed, Sep 17, 2014 at 01:25:19PM -0700, Ethan Jackson wrote:
Frequently we've run into controller bugs which result in hundreds of
thousands, or even millions of rules being installed in an OpenFlow
table. This isn't something
Acked-by: Ethan Jackson et...@nicira.com
On Sun, Sep 14, 2014 at 8:06 PM, Joe Stringer joestrin...@nicira.com wrote:
Although I reviewed and applied #1, I'll leave this one up to Ethan as it's
more relevant to his other proposed changes.
On 13 September 2014 05:35, Daniele Di Proietto
Acked-by: Ethan Jackson et...@nicira.com
On Thu, Sep 11, 2014 at 3:03 PM, Joe Stringer joestrin...@nicira.com wrote:
The datapath max_idle value determines how long to wait before deleting
an idle datapath flow when operating below the flow_limit. This patch
increases the max_idle to 10
Is there any point to testing OVS in a configuration that we won't
distribute?
This is a subject in which there could legitimately be many points of
view, but I'll take a stab at explaining how I think about it.
To answer your question: In general no, but with DPDK yes.
If you look at how
These options don't make sense when building portable code, but when
using the dev script, OVS is built on the same system it's run on.
They make a small difference in the OVS DPDK testing, hence their
addition.
Signed-off-by: Ethan Jackson et...@nicira.com
---
utilities/ovs-dev.py | 6 +++---
1
They may or may not make a difference, but there's no reason not to
support passing them.
Signed-off-by: Ethan Jackson et...@nicira.com
---
utilities/ovs-dev.py | 11 +++
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/utilities/ovs-dev.py b/utilities/ovs-dev.py
index
They have slightly different support characteristics, so it's nice to
easily switch between them for testing.
Signed-off-by: Ethan Jackson et...@nicira.com
---
utilities/ovs-dev.py | 18 ++
1 file changed, 14 insertions(+), 4 deletions(-)
diff --git a/utilities/ovs-dev.py b
warning: incorrect type in argument 1 (different base types)
expected restricted ovs_be16 [usertype] old_csum
got unsigned short [unsigned] [usertype] icmp6_cksum
Signed-off-by: Ethan Jackson et...@nicira.com
---
lib/packets.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/lib
oops, I missed it. I'll drop this then
On Tue, Sep 2, 2014 at 6:44 PM, Jesse Gross je...@nicira.com wrote:
On Tue, Sep 2, 2014 at 6:36 PM, Ethan Jackson et...@nicira.com wrote:
warning: incorrect type in argument 1 (different base types)
expected restricted ovs_be16 [usertype] old_csum
Acked-by: Ethan Jackson et...@nicira.com
Thanks for taking care of this.
Ethan
On Fri, Aug 15, 2014 at 1:15 AM, Alex Wang al...@nicira.com wrote:
Commit cc377352d (ofproto: Reorganize in preparation for direct
dpdk upcalls.) introduced the bug that uses variable defined on
the stack inside
1 - 100 of 2861 matches
Mail list logo