Yes, I could reproduce it on my VM setup,
It is the timing issue again, if the p1 starts sending ccm first, it will
have flap count = 0.
If p1 receives ccm from p0 first, it will go to [rdi] and then []. so it
will have flap count = 2.
I think the safest way is to drop the "CFM_VSCTL_LIST_IFACE
On Thu, Dec 12, 2013 at 08:24:54PM -0800, Ethan Jackson wrote:
> This patch adds a new function flow_unildcard_tp_ports() which doesn't
> unwildcard the upper half of tp_src and tp_dst with ICMP packets.
> Unfortunately, this matters in future patches when we compare masks
> carefully to determine
On Thu, Dec 12, 2013 at 08:24:55PM -0800, Ethan Jackson wrote:
> In accordance with the style guidelines.
>
> Signed-off-by: Ethan Jackson
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Thu, Dec 12, 2013 at 08:24:56PM -0800, Ethan Jackson wrote:
> There's no particular reason for the function controlling the number
> of threads to be bound up with dpif_recv_set(). This patch breaks
> them up, but as a side effect means threads will run doing nothing
> when datapath upcall rece
In accordance with the style guidelines.
Signed-off-by: Ethan Jackson
---
utilities/ovs-ofctl.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/utilities/ovs-ofctl.c b/utilities/ovs-ofctl.c
index 5eb8cf4..c481210 100644
--- a/utilities/ovs-ofctl.c
+++ b/utilities/ovs-ofctl.c
This patch adds a new function flow_unildcard_tp_ports() which doesn't
unwildcard the upper half of tp_src and tp_dst with ICMP packets.
Unfortunately, this matters in future patches when we compare masks
carefully to determine if flows should be evicted from the datapath.
Signed-off-by: Ethan Jac
There's no particular reason for the function controlling the number
of threads to be bound up with dpif_recv_set(). This patch breaks
them up, but as a side effect means threads will run doing nothing
when datapath upcall receiving is disabled. By doing this, the udpif
thread creation API become
This patch occasionally fails the cfm flap count test for me. Can
either of you reproduce it?
Ethan
On Thu, Dec 12, 2013 at 5:52 PM, Joe Stringer wrote:
> Currently, every time a monitoring port is added or reconfigured, the
> main thread notifies the monitoring thread to wake up immediately us
;>>> +
>>>> # since p0 has been in decay, now the RX will show 3000ms.
>>>> BFD_CHECK_TX([p0], [500ms], [300ms], [500ms])
>>>> BFD_CHECK_RX([p0], [3000ms], [3000ms], [500ms])
>>>> @@ -629,8 +629,6 @@ done
>>>>
>>>> # rec
> Might be nice to mention why.
Good point, I added a sentence explaining that facets and subfacets
are going away.
Ethan
>
> Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
> This patch does not remove avg_n_subfacet. It removes
> avg_subfacet_life.
oops
>
> The change to ofproto/ofproto-dpif-upcall.c seems unrelated and I think
> it could be dropped or folded into the previous patch.
Rebasing error.
Thanks
>
> Acked-by: Ben Pfaff
__
> Perhaps ofproto_type_get_memory_usage() should verify that the class it
> gets is nonnull before it tries to use it?
Agreed, changed thanks.
Ethan
>
> Acked-by: Ben Pfaff
> ___
> dev mailing list
> dev@openvswitch.org
> http://openvswitch.org/mailma
Good catch thanks. I ditched the first of the two.
Ethan
On Thu, Dec 12, 2013 at 4:23 PM, Ben Pfaff wrote:
> On Sun, Dec 08, 2013 at 06:45:22PM -0800, Ethan Jackson wrote:
>> They don't really make sense in a multithreaded architecture. Once
>> flow miss batches are dispatched with, they will
OK, thanks for the background.
On Dec 12, 2013 6:57 PM, "Ethan Jackson" wrote:
> > I wonder how much time it really saves?
>
> Jarno said offline that disabling the decision tree stuff he wrote
> increases flow setup performance by roughly 5%. In the new multi
> threaded revalidation setup, shav
> I wonder how much time it really saves?
Jarno said offline that disabling the decision tree stuff he wrote
increases flow setup performance by roughly 5%. In the new multi
threaded revalidation setup, shaving a bit off the cost of
xlate_actions() will directly lead to larger datapath flow table
On Thu, Dec 12, 2013 at 06:12:32PM -0800, Ben Pfaff wrote:
> On Fri, Dec 13, 2013 at 09:44:06AM +0900, Simon Horman wrote:
> > On Thu, Dec 12, 2013 at 09:56:18AM -0800, Ben Pfaff wrote:
> > > On Thu, Dec 12, 2013 at 05:38:59PM +0900, Simon Horman wrote:
> > > > If VLAN acceleration is used when the
On Thu, Dec 12, 2013 at 06:01:26PM -0800, Ben Pfaff wrote:
> On Fri, Dec 13, 2013 at 09:45:43AM +0900, Simon Horman wrote:
> > On Thu, Dec 12, 2013 at 08:11:47AM -0800, Jarno Rajahalme wrote:
> > > IMO this patch could go a bit further and have the rx_recv also
> > > update the but size to reflect
On Wed, Dec 11, 2013 at 10:23:36AM +0900, Simon Horman wrote:
> * The ethertype of the packet is always set to
> the ethertype of the push_mpls action
> * Drop references to unicast and multicast MPLS ethertypes,
> in RFC5332 0x8847 and 0x8848 do not always have
> those meanings.
>
> Signed-
On Fri, Dec 13, 2013 at 09:44:06AM +0900, Simon Horman wrote:
> On Thu, Dec 12, 2013 at 09:56:18AM -0800, Ben Pfaff wrote:
> > On Thu, Dec 12, 2013 at 05:38:59PM +0900, Simon Horman wrote:
> > > If VLAN acceleration is used when the kernel receives a packet
> > > then the outer-most VLAN tag will n
On Fri, Dec 13, 2013 at 09:45:43AM +0900, Simon Horman wrote:
> On Thu, Dec 12, 2013 at 08:11:47AM -0800, Jarno Rajahalme wrote:
> > IMO this patch could go a bit further and have the rx_recv also
> > update the but size to reflect the number of bytes placed in the
> > buffer and maybe change the r
Yeah, if you want clang to catch that then you would need to add a
redundant prototype near the top of the C file (not actually a bad idea).
On Dec 12, 2013 5:12 PM, "Ethan Jackson" wrote:
> Annoying that clang didn't catch that . . . I suppose it's cause
> 'mutex' doesn't exist where we define
Annoying that clang didn't catch that . . . I suppose it's cause
'mutex' doesn't exist where we define netflow_run()'s protoype and we
have to annotate in both places.
At any rate. I'll make a netflow_run__() function which doesn't
require the mutex.
On Thu, Dec 12, 2013 at 2:56 PM, Ben Pfaff
> I think that the flow parameter to netflow_expire(),
> netflow_flow_clear(), and netflow_flow_update() could be const.
Actually none of them can because they either much with the flows hmap
or the packet. I could const cast them, but it doesn't seem worth it.
> Acked-by: Ben Pfaff
I've folde
I really just want an update to the commit message.
On Dec 12, 2013 4:51 PM, "Ethan Jackson" wrote:
> > It seems harmless but the reason for it isn't really obvious. Maybe
> > this is to support some later change in the series? If so then it would
> > be nice to hint at that in the commit messa
Currently, every time a monitoring port is added or reconfigured, the
main thread notifies the monitoring thread to wake up immediately using
monitor_seq. When adding a large number of ports at once, this causes
contention as the threads fight over access to the monitor heap---one
thread adding new
> It seems harmless but the reason for it isn't really obvious. Maybe
> this is to support some later change in the series? If so then it would
> be nice to hint at that in the commit message.
The main reason is that the API became a bit unwieldy once there were
two types of thread to deal with.
Sure.
On 12 December 2013 16:44, Ethan Jackson wrote:
> Joe, could you rebase it and resend it please?
>
> Ethan
>
> On Thu, Dec 12, 2013 at 5:17 PM, Alex Wang wrote:
>> Hey Ben, Ethan,
>>
>> Could any of you apply this patch?
>>
>> Thanks,
>> Alex Wang,
>>
>>
>> On Thu, Dec 12, 2013 at 4:15 PM,
On Thu, Dec 12, 2013 at 08:11:47AM -0800, Jarno Rajahalme wrote:
> IMO this patch could go a bit further and have the rx_recv also update the
> but size to reflect the number of bytes placed in the buffer and maybe change
> the return value to a normal error code.
Hi,
I would not be opposed to
Joe, could you rebase it and resend it please?
Ethan
On Thu, Dec 12, 2013 at 5:17 PM, Alex Wang wrote:
> Hey Ben, Ethan,
>
> Could any of you apply this patch?
>
> Thanks,
> Alex Wang,
>
>
> On Thu, Dec 12, 2013 at 4:15 PM, Alex Wang wrote:
>>
>> This patch looks good to me, thx for doing this.
On Thu, Dec 12, 2013 at 09:29:12AM -0800, Ben Pfaff wrote:
> On Thu, Dec 12, 2013 at 05:38:58PM +0900, Simon Horman wrote:
> > Update the netdev_class so that struct ofpbuf * is passed to rx_recv()
> > to provide both the data and size of the data to read a packet into.
> >
> > This patch should n
, [300ms], [500ms])
>>>
>>> # wait for 5000ms to decay.
>>> for i in `seq 0 9`; do ovs-appctl time/warp 500; done
>>> diff --git a/tests/cfm.at b/tests/cfm.at
>>> index 9e351d0..620e3e0 100644
>>> --- a/tests/cfm.at
>>> +++ b/tests/cfm.a
On Thu, Dec 12, 2013 at 09:56:18AM -0800, Ben Pfaff wrote:
> On Thu, Dec 12, 2013 at 05:38:59PM +0900, Simon Horman wrote:
> > If VLAN acceleration is used when the kernel receives a packet
> > then the outer-most VLAN tag will not be present in the packet
> > when it is received by netdev-linux. R
Looks good to me. Just one minor comment. You probably also want to include the
following?
#include
#pragma comment(lib, "advapi32.lib")
Btw, printing the error message will require a small change which maps
strerror_r to strerror_s. Not sure, if we should just add it to this commit.
Saurabh
Use WaitForMultipleObjects for polling on windows. This works on all kinds
of objects, e.g. sockets, files, especially ioctl calls to the kernel.
One additional paramater is passed down in poll_fd_wait() to help
WaitForMultipleObjects.
latch is signaled with event, to be waited/polled by WaitFo
Hey Ben, Ethan,
Could any of you apply this patch?
Thanks,
Alex Wang,
On Thu, Dec 12, 2013 at 4:15 PM, Alex Wang wrote:
> This patch looks good to me, thx for doing this.
>
>
> On Thu, Dec 12, 2013 at 1:03 PM, Joe Stringer wrote:
>
>> Currently, every time a monitoring port is added or reconf
This patch looks good to me, thx for doing this.
On Thu, Dec 12, 2013 at 1:03 PM, Joe Stringer wrote:
> Currently, every time a monitoring port is added or reconfigured, the
> main thread notifies the monitoring thread to wake up immediately using
> monitor_seq. When adding a large number of por
On 11 December 2013 18:11, Ethan Jackson wrote:
> The first line of the commit message needs a period.
Woops, missed this. I'll try to remember in future.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Sun, Dec 08, 2013 at 06:45:27PM -0800, Ethan Jackson wrote:
> From: Joe Stringer
>
> Previously we used facets for ovs-appctl dpif/dump-flows commands. This
> switches to fetching flows directly from the dpif.
>
> Signed-off-by: Joe Stringer
> Signed-off-by: Ethan Jackson
> Acked-by: Ethan
On Sun, Dec 08, 2013 at 06:45:26PM -0800, Ethan Jackson wrote:
> It's not entirely clear why someone would want to delete the flows
> associated with only one bridge and not be fine with simply using
> ovs-dpctl del-flows to get all of them. Additionally, this will be
> hard to implement cleanly i
This valgrind leak checker isn't really useful without this.
Signed-off-by: Ethan Jackson
---
utilities/ovs-dev.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/utilities/ovs-dev.py b/utilities/ovs-dev.py
index c5445e2..7b69094 100755
--- a/utilities/ovs-dev.py
+++ b/utilit
On Sun, Dec 08, 2013 at 06:45:25PM -0800, Ethan Jackson wrote:
> This is going to be hard to compute efficiently in the new
> architecture introduced in future patches.
>
> Signed-off-by: Ethan Jackson
This patch does not remove avg_n_subfacet. It removes
avg_subfacet_life.
The change to ofpro
On Sun, Dec 08, 2013 at 06:45:24PM -0800, Ethan Jackson wrote:
> Shows debugging information related to upcall handling.
>
> Signed-off-by: Ethan Jackson
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listi
On Sun, Dec 08, 2013 at 06:45:23PM -0800, Ethan Jackson wrote:
> Signed-off-by: Ethan Jackson
Perhaps ofproto_type_get_memory_usage() should verify that the class it
gets is nonnull before it tries to use it?
Acked-by: Ben Pfaff
___
dev mailing list
d
On Sun, Dec 08, 2013 at 06:45:22PM -0800, Ethan Jackson wrote:
> They don't really make sense in a multithreaded architecture. Once
> flow miss batches are dispatched with, they will be extra useless.
>
> Signed-off-by: Ethan Jackson
I had to resolve a few conflicts due to changes on master. I
On Sun, Dec 08, 2013 at 06:45:21PM -0800, Ethan Jackson wrote:
> Future patches will need to poll_block() on it.
>
> Signed-off-by: Ethan Jackson
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Thanks Joe, I've merged the series.
Ethan
On Thu, Dec 12, 2013 at 11:33 AM, Joe Stringer wrote:
> Previously, we iterated through all interfaces in instant_stats_run(),
> grabbing up-to-date information about device and port status. After
> assembling all of this information for all interfaces,
On Sun, Dec 08, 2013 at 06:45:20PM -0800, Ethan Jackson wrote:
> As time goes on and the classifier becomes more complicated, calculate
> the wildcard mask will get more and more expensive. This patch adds a
> mechanism to xlate_actions() allowing callers to disable wildcard
> calculation when it
Looks good.
On 12 December 2013 14:40, Ethan Jackson wrote:
> Here's an incremental. I folded in some tweaks which make it slightly easier
> to
> build without clang or sparse.
>
> Still good?
> Ethan
>
> ---
> .gitignore | 6 ++
> utilities/ovs-dev.py | 19 +++
>
Whatever you like, I don't feel like I have a stake in this, I was just
explaining my own reasons.
On Dec 12, 2013 2:10 PM, "Ethan Jackson" wrote:
> so what do you think ben/joe? _build-clang? _build_clang?
> build-clang? I don't care ...
>
> Ethan
>
> On Thu, Dec 12, 2013 at 2:08 PM, Ben Pfaf
Here's an incremental. I folded in some tweaks which make it slightly easier to
build without clang or sparse.
Still good?
Ethan
---
.gitignore | 6 ++
utilities/ovs-dev.py | 19 +++
2 files changed, 17 insertions(+), 8 deletions(-)
diff --git a/.gitignore b/.git
On Sun, Dec 08, 2013 at 06:45:19PM -0800, Ethan Jackson wrote:
> This patch moves statistics accounting for netflow, bonding, netdev,
> and mirroring inside xlate_actions(). Moving all statistics into one
> place makes it very difficult to mess up.
>
> Signed-off-by: Ethan Jackson
Acked-by: Ben
I applied your patch
(http://openvswitch.org/pipermail/dev/2013-November/034001.html) and added the
includes things started to compile :).
But again it is up too you how the file structure will be in end.
Kind Regards,
Alin.
From: Gurucharan Shetty [shet
On Sun, Dec 08, 2013 at 06:45:17PM -0800, Ethan Jackson wrote:
> This patch changes flow_mask_hash_fields() so it doesn't unwildcard
> the upper half of tp_src and tp_dst with ICMP packets. Unfortunately,
> this matters in future patches when we compare masks carefully to
> determine if flows shou
I see. I like the prefix idea.
Didn't mean to bikeshed, I say just pick something and go with it ;-)
On 12 December 2013 14:08, Ben Pfaff wrote:
> On Thu, Dec 12, 2013 at 01:24:53PM -0800, Joe Stringer wrote:
>> On 8 December 2013 18:45, Ethan Jackson wrote:
>> > +/build_gcc
>> > +/build_clang
On Sun, Dec 08, 2013 at 06:45:18PM -0800, Ethan Jackson wrote:
> This seems cleaner than having separate accessors for them.
>
> Signed-off-by: Ethan Jackson
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/l
On Sun, Dec 08, 2013 at 06:45:16PM -0800, Ethan Jackson wrote:
> This patch changes netflow_mask_wc() so it doesn't unwildcard the
> upper half of tp_src and tp_dst with ICMP packets. Unfortunately,
> this matters in future patches when we compare masks carefully to
> determine if flows should be
On Sun, Dec 08, 2013 at 06:45:15PM -0800, Ethan Jackson wrote:
> This patch changes how ofproto-dpif-xlate handles netflow so that it's
> more similar to how ipfix and sflow are handled. This not only
> simplifies the code, but it will allow xlate to call netflow in future
> if necessary.
>
> Sig
so what do you think ben/joe? _build-clang? _build_clang?
build-clang? I don't care ...
Ethan
On Thu, Dec 12, 2013 at 2:08 PM, Ben Pfaff wrote:
> On Thu, Dec 12, 2013 at 01:24:53PM -0800, Joe Stringer wrote:
>> On 8 December 2013 18:45, Ethan Jackson wrote:
>> > +/build_gcc
>> > +/build_clan
On Thu, Dec 12, 2013 at 01:39:04PM -0800, Ethan Jackson wrote:
> > Hmm. Using too many could easily cause a crash, but at least it would
> > recover after restarting when the number was lowered again.
>
> Interesting, why would we crash? Memory consumption? At any rate,
> this is the kind of th
On Thu, Dec 12, 2013 at 01:36:08PM -0800, Ethan Jackson wrote:
> > I'd rather not modify
> > xenserver/opt_xensource_libexec_InterfaceReconfigure.py since it's
> > maintained in conjunction with Citrix.
>
> Meh, I don't like that we have code in the tree that we can't change.
Not "can't change" b
On Thu, Dec 12, 2013 at 01:24:53PM -0800, Joe Stringer wrote:
> On 8 December 2013 18:45, Ethan Jackson wrote:
> > +/build_gcc
> > +/build_clang
> > +/_build
> > +/build
>
> Most directories in ovs seem to use - instead of _ in the names (ie,
> 'build-clang' rather than 'build_clang').
I tend to
On Sun, Dec 08, 2013 at 06:45:14PM -0800, Ethan Jackson wrote:
> Useful in future patches.
>
> Signed-off-by: Ethan Jackson
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Updated. Thanks for the feedback :-).
Final version would look something like:
lib/entropy.c | 13 +
1 files changed, 13 insertions(+), 0 deletions(-)
---
diff --git a/lib/entropy.c b/lib/entropy.c
index 02f56e0..45e83ec 100644
--- a/lib/entropy.c
+++ b/lib/entropy.c
@@ -33,6 +33,
On Sun, Dec 08, 2013 at 06:45:13PM -0800, Ethan Jackson wrote:
> In future patches upcall handler threads will need to update netflow.
>
> Signed-off-by: Ethan Jackson
I think I see a deadlock. gen_netflow_rec() requires 'mutex',
netflow_run() excludes it, but the former calls the latter.
Mayb
On Sun, Dec 08, 2013 at 06:45:12PM -0800, Ethan Jackson wrote:
> The netflow code has its tentacles all over the ofproto-dpif module.
> This is fine today, but in future facets, which correspond roughly to
> netflow_flows, will be retired. In preparation, this patch hides as
> much implementation
> Hmm. Using too many could easily cause a crash, but at least it would
> recover after restarting when the number was lowered again.
Interesting, why would we crash? Memory consumption? At any rate,
this is the kind of thing where you have to go out of your way to get
it wrong, so I'm fine wit
> I'd rather not modify
> xenserver/opt_xensource_libexec_InterfaceReconfigure.py since it's
> maintained in conjunction with Citrix.
Meh, I don't like that we have code in the tree that we can't change.
But this doesn't really matter *at all*, so I'll just change ofctl.
Ethan
___
On 8 December 2013 18:45, Ethan Jackson wrote:
> +/build_gcc
> +/build_clang
> +/_build
> +/build
Most directories in ovs seem to use - instead of _ in the names (ie,
'build-clang' rather than 'build_clang').
I didn't have any 'build' or '_build' directory after running this script.
Otherwise,
On Sun, Dec 08, 2013 at 06:45:11PM -0800, Ethan Jackson wrote:
> There's no particular reason for the function controlling the number
> of threads to be bound up with dpif_recv_set(). This patch breaks
> them up, but as a side effect means threads will run doing nothing
> when datapath upcall rece
From: Francesco Fusco
Date: Thu, 12 Dec 2013 16:09:04 +0100
> We are introducing a fast hash function (see patch1) that can be
> used in the context of OpenVSwitch to reduce the hashing footprint
> (patch2). For details, please see individual patches!
No objections from my side, I'll let the x86
Currently, every time a monitoring port is added or reconfigured, the
main thread notifies the monitoring thread to wake up immediately using
monitor_seq. When adding a large number of ports at once, this causes
contention as the threads fight over access to the monitor heap---one
thread adding new
Just confirm, with the initialization, the intermittent test failures
(relate to ofproto-dpif megaflow tests) are gone.
On Wed, Dec 11, 2013 at 6:27 PM, Ethan Jackson wrote:
> Btw to whoever reviews this, this version of the patch fails to
> initialize skip_wildcards in xlate_in_init().
>
> Eth
On Sun, Dec 08, 2013 at 06:45:10PM -0800, Ethan Jackson wrote:
> There's no particular reason to limit the number of handler threads to
> 64. If someone wants to try more, let them.
>
> Signed-off-by: Ethan Jackson
Hmm. Using too many could easily cause a crash, but at least it would
recover a
On Sun, Dec 08, 2013 at 06:45:09PM -0800, Ethan Jackson wrote:
> Signed-off-by: Ethan Jackson
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Sun, Dec 08, 2013 at 06:45:08PM -0800, Ethan Jackson wrote:
> In accordance with the style guidelines.
>
> Signed-off-by: Ethan Jackson
I'd rather not modify
xenserver/opt_xensource_libexec_InterfaceReconfigure.py since it's
maintained in conjunction with Citrix.
The ovs-ofctl change seems f
On Sun, Dec 08, 2013 at 06:45:07PM -0800, Ethan Jackson wrote:
> Many deployments prefer to use gcc in production and it supports
> sparse, while clang supports thread safety analysis and has
> more/better warnings and error messages. This patch changes
> ovs-dev.py to build with both so that deve
I applied the patches and everything seems in order.
The same remark about adding string.h to lib/.gitignore
Thanks,
Alin.
From: dev-boun...@openvswitch.org [dev-boun...@openvswitch.org] on behalf of
Gurucharan Shetty [shet...@nicira.com]
Sent: Thursday,
On Thu, Dec 12, 2013 at 7:09 AM, Francesco Fusco wrote:
> Currently OVS uses jhash2() for calculating flow hashes in its
> internal flow_hash() function. The performance of the flow_hash()
> function is critical, as the input data can be hundreds of bytes
> long.
>
> OVS is largely deployed in x86
Previously, we tracked status changes for ofports on a per-device basis.
Each time in the main thread's loop, we would inspect every ofport
to determine whether the status had changed for corresponding devices.
This patch replaces the per-netdev change_seq with a global 'struct seq'
which tracks s
Previously, we iterated through all interfaces in instant_stats_run(),
grabbing up-to-date information about device and port status. After
assembling all of this information for all interfaces, we would
determine whether anything changed and only send an update to
ovsdb-server if something changed.
Currently, configuring a large number of ports in ovs-vswitchd causes the main
thread to use an unreasonably high amount of CPU. This is largely due to cases
in the main loop where all ports are iterated through, even when this is
unnecessary.
This patchset shifts the netdev change_seq out from pe
Previously, as part of ofproto-dpif run() processing, we would loop
through all ports and poll for changes to carrier, bfd, cfm and lacp
status. This information is used to determine whether bundles may be
enabled, and to perform revalidation when needed.
This patch makes the bfd, cfm, lacp and st
When dealing with a large number of ports, bundle_run() and
bundle_wait() add significant unnecessary processing to the main run
loop, even when there are no bonds and lacp is not configured. This
patch skips such execution if it is unneeded, reducing average CPU usage
of the main thread from about
Between, what is the difference between NOTICE and COPYING?
On Thu, Dec 12, 2013 at 11:27 AM, Gurucharan Shetty wrote:
> On Wed, Dec 11, 2013 at 3:04 PM, Ben Pfaff wrote:
>> This will be used for the #include_next in string.h in the following
>> commit.
>>
>> Signed-off-by: Ben Pfaff
>> ---
>>
On Wed, Dec 11, 2013 at 3:04 PM, Ben Pfaff wrote:
> Signed-off-by: Ben Pfaff
> ---
> configure.ac |3 +++
> lib/automake.mk |2 +-
> lib/{string.h => string.h.in} |4 ++--
> 3 files changed, 6 insertions(+), 3 deletions(-)
> rename lib/{string.h => str
On Wed, Dec 11, 2013 at 3:04 PM, Ben Pfaff wrote:
> This will be used for the #include_next in string.h in the following
> commit.
>
> Signed-off-by: Ben Pfaff
> ---
> NOTICE|5 +-
> build-aux/.gitignore |2 +
> debian/copyright.in | 10 ++-
> m4/absolute-header.m4 |
On 12/12/2013 06:54 PM, Nicolas Dichtel wrote:
...
Is it possible to trace that this comes from the DPDK?
At least in the commit log, like it was done in the v1.
Hm, that got lost; but we can put that back into the commit log.
In any case, we properly included the header comment, of course.
+
On Thu, Dec 12, 2013 at 05:38:59PM +0900, Simon Horman wrote:
> If VLAN acceleration is used when the kernel receives a packet
> then the outer-most VLAN tag will not be present in the packet
> when it is received by netdev-linux. Rather, it will be present
> in auxdata.
>
> This patch uses recvms
Title: EXPERTO EN COACHING EJECUTIVO
EXPERTO EN COACHING EJECUTIVO Pulse AQUÍ si no visualiza correctamente las imágenes.
Le 12/12/2013 16:09, Francesco Fusco a écrit :
We introduce a new hashing library that is meant to be used in
the contexts where speed is more important than uniformity of the
hashed values. The hash library leverages architecture specific
implementation to achieve high performance and fall backs
On Thu, Dec 12, 2013 at 05:38:58PM +0900, Simon Horman wrote:
> Update the netdev_class so that struct ofpbuf * is passed to rx_recv()
> to provide both the data and size of the data to read a packet into.
>
> This patch should not have any behavioural changes.
>
> This patch is in preparation fo
Thanks, pushed!
Jarno
On Dec 11, 2013, at 9:25 PM, Simon Horman wrote:
> On Wed, Dec 11, 2013 at 03:41:19PM -0800, Jarno Rajahalme wrote:
>> - Use the GCC predefined macro __POPCNT__ to detect the availability
>> of fast __builtin_popcnt function.
>> - Use portable preprocessor macros
IMO this patch could go a bit further and have the rx_recv also update the but
size to reflect the number of bytes placed in the buffer and maybe change the
return value to a normal error code.
Jarno
On Dec 12, 2013, at 12:38 AM, Simon Horman wrote:
> Update the netdev_class so that struct
We introduce a new hashing library that is meant to be used in
the contexts where speed is more important than uniformity of the
hashed values. The hash library leverages architecture specific
implementation to achieve high performance and fall backs to
jhash() for the generic case.
On Intel-based
From: Daniel Borkmann
We are introducing a fast hash function (see patch1) that can be
used in the context of OpenVSwitch to reduce the hashing footprint
(patch2). For details, please see individual patches!
Thanks !
v1->v2:
- Make hash generic and place it under lib
Francesco Fusco (2):
li
Currently OVS uses jhash2() for calculating flow hashes in its
internal flow_hash() function. The performance of the flow_hash()
function is critical, as the input data can be hundreds of bytes
long.
OVS is largely deployed in x86_64 based datacenters. Therefore,
we argue that the performance cri
From: Daniel Borkmann
We are introducing a fast hash function (see patch1) that can be
used in the context of OpenVSwitch to reduce the hashing footprint
(patch2). For details, please see individual patches!
Thanks !
v1->v2:
- Make hash generic and place it under lib
Francesco Fusco (2):
li
On Wed, Dec 11, 2013 at 10:00 PM, Ben Pfaff wrote:
> On Tue, Dec 10, 2013 at 02:49:29PM +0100, Helmut Schaa wrote:
>> Use a global array that gets initialized by constructor functions
>> per coverage definition.
>>
>> Signed-off-by: Helmut Schaa
>> ---
>>
>> The realloc is not the nicest thing to
If VLAN acceleration is used when the kernel receives a packet
then the outer-most VLAN tag will not be present in the packet
when it is received by netdev-linux. Rather, it will be present
in auxdata.
This patch uses recvmsg() instead of recv() to read auxdata for
each packet and if the vlan_tid
If VLAN acceleration is used when the kernel receives a packet
then the outer-most VLAN tag will not be present in the packet
when it is received by netdev-linux. Rather, it will be present
in auxdata.
The purpose of this series is to obtain the vlan_tid from auxdata
as appropriate and push it ont
1 - 100 of 101 matches
Mail list logo
