On Tue, Mar 15, 2011 at 06:37:12PM -0700, Ethan Jackson wrote:
> >ofp-parse: Factor out parse_ofp_add_flow_file() into new function.
>
> This subject line doesn't make sense to me. Shouldn't it read:
> "Factor out ds_get_preprocessed_line() into new function"?
You're right, the subject line does
On Tue, Mar 15, 2011 at 04:03:16PM -0700, Ethan Jackson wrote:
> The monitoring logic and bonding code are unrelated. This commit
> pulls the monitoring logic out. As a result all interfaces, not
> just those participating in bonds, are monitored. This will be
> required to run LACP on non-bonde
On Tue, Mar 15, 2011 at 04:03:17PM -0700, Ethan Jackson wrote:
> Creates new compose_lacp_pdu() from the old compose_lacp_packet()
> function. This will allow a LACP PDU to be created without
> necessarily knowing the Ethernet Source Address required for
> generating the packet.
Will there be som
On Tue, Mar 15, 2011 at 04:03:18PM -0700, Ethan Jackson wrote:
> This seems more stylistically consistent with the rest of the
> code base and the sys_priority and port_priority in the same
> struct.
Looks good, thanks.
___
dev mailing list
dev@openvswit
Is this because LACP influences bonding but not vice versa? Makes
sense, but you could add that to the commit message.
Thanks,
Ben.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Artifact of monitoring being done as part of bonding. I added the
monitors neer port_update_bonding. No particular reason other than
history. I will go ahead and change it.
Ethan
On Wed, Mar 16, 2011 at 9:24 AM, Ben Pfaff wrote:
> On Tue, Mar 15, 2011 at 04:03:16PM -0700, Ethan Jackson wrote:
On Tue, Mar 15, 2011 at 04:03:20PM -0700, Ethan Jackson wrote:
> This commit pulls LACP code out of the bridge into its own LACP
> module. Currently this module is only used by the existing bonding
> infrastructure.
Thanks. Making this more modular makes the code easier to understand.
I didn't
On Tue, Mar 15, 2011 at 04:03:21PM -0700, Ethan Jackson wrote:
> Only transmit when the LACP partner needs to be updated instead of
> whenever it might need to be updated.
Thanks, this looks good to me.
Now I understand the slave->tx = 0 assignments :-)
In the second loop in lacp_run(), I think
Looks good to me.
Should we document that "slow" is the default mode?
Presumably there is some disadvantage to "fast" mode, since it is not
the default--I guess that a little latency somewhere can cause
spurious port-down detection--and so maybe we should document that
disadvantage.
Thanks,
Ben
On Tue, Mar 15, 2011 at 04:03:23PM -0700, Ethan Jackson wrote:
> When a slave may no longer participate in a LACP bond, it is polite
> to transmit a notification indicating so. This will allow the
> remote partner to adapt more quickly. The specification does not
> indicate what to do in this sit
Also: This commit leaves LACP_ACTIVE and a few other macros in
vswitchd/bridge.c, but nothing uses them anymore, so I guess that you
can delete them.
Thanks,
Ben.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Sorry about that. Intended to clean these patches up a bit before
sending them out. I'll get rid of them.
Ethan
On Wed, Mar 16, 2011 at 12:38 PM, Ben Pfaff wrote:
> Also: This commit leaves LACP_ACTIVE and a few other macros in
> vswitchd/bridge.c, but nothing uses them anymore, so I guess tha
Also: as with the lacp_active member of struct port, I suspect that
the lacp_fast member can be implemented entirely in struct lacp. It
could initially be set as a parameter to lacp_create(), for example.
___
dev mailing list
dev@openvswitch.org
http://o
Twitter connects you with everything you want to know, right now. Short bursts
of information are readily available from news organizations, corporate
entities, politicians, celebrities, local businesses - even your close friends
and family. Also, if you have something to share with the world, T
Twitter connects you with everything you want to know, right now. Short bursts
of information are readily available from news organizations, corporate
entities, politicians, celebrities, local businesses - even your close friends
and family. Also, if you have something to share with the world, T
Twitter connects you with everything you want to know, right now. Short bursts
of information are readily available from news organizations, corporate
entities, politicians, celebrities, local businesses - even your close friends
and family. Also, if you have something to share with the world, T
On Tue, Mar 15, 2011 at 11:36 AM, Ben Pfaff wrote:
> ---
> ofproto/ofproto.c | 3 ++-
> 1 files changed, 2 insertions(+), 1 deletions(-)
Looks good to me, thanks.
-Andrew
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listi
On Tue, Mar 15, 2011 at 11:36 AM, Ben Pfaff wrote:
> "secchan" is long gone.
Looks good to me, thanks.
-Andrew
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Tue, Mar 15, 2011 at 12:18 PM, Ben Pfaff wrote:
> Some Linux network drivers support a feature called "VLAN acceleration",
> associated with a data structure called a "vlan_group". A vlan_group is,
> abstractly, a dictionary that maps from a VLAN ID (in the range 0...4095)
> to a VLAN device,
On 3/15/11 11:36 AM, Ben Pfaff wrote:
> Back in 2008 or so, I introduced this extension as a way to provide
> information about switch status to the new "switch UI" program. Since
> then, the switch UI program has been removed and the important information
> that was provided by the switch status
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
This set of patches moves the CFM implementation from ovs-vswitchd
into ofproto. This is one step toward my goal of getting rid of the
hooks from ofproto back into vswitchd.
Ben Pfaff (3):
cfm: Immediately signal a fault upon receiving an unexpected MPID.
ofproto: Specialize ofproto_send_pack
An unexpected MPID is always a fault, but the CFM code didn't signal the
fault until the next time cfm_run() was called. In one experiment I
saw a visible lag in the database (although I wasn't able to reproduce it
again within a few tries).
---
lib/cfm.c |3 ++-
1 files changed, 2 insertions
The callers of ofproto_send_packet() actually just want to send a packet
out on a port, possibly tagged with a VLAN, but the interface forced them
to compose a set of OpenFlow actions, which made it harder to use than
necessary. This commit specializes the interface for the purposes that
the calle
This takes one step toward the larger goal of getting rid of all of the
hooks from ofproto back into vswitchd, by eliminating one of the reasons
that they are required.
---
ofproto/ofproto.c | 132 +---
ofproto/ofproto.h | 11 -
vswitchd/bridg
On Wed, Mar 16, 2011 at 02:18:55PM -0700, Jesse Gross wrote:
> On Tue, Mar 15, 2011 at 12:18 PM, Ben Pfaff wrote:
> > Some Linux network drivers support a feature called "VLAN acceleration",
> > associated with a data structure called a "vlan_group". ?A vlan_group is,
> > abstractly, a dictionary
On 3/15/11 11:36 AM, Ben Pfaff wrote:
> I've never heard of anyone actually using controller discovery.
> It adds a great deal of code to the source tree, and a little
> bit of complication to ofproto, so this commit removes it.
Looks good to me, thanks.
-Andrew
__
I pushed this series up to this point.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
On Wed, Mar 16, 2011 at 02:45:43PM -0700, Andrew Evans wrote:
> On 3/15/11 11:36 AM, Ben Pfaff wrote:
> > I've never heard of anyone actually using controller discovery.
> > It adds a great deal of code to the source tree, and a little
> > bit of complication to ofproto, so this commit removes it.
I forgot to update the test when I added the VLAN bug workaround.
---
I already pushed this test fix, but I'm emailing it to the list
anyway for completeness, and in case anyone sees anything that could
be done better.
diff --git a/tests/interface-reconfigure.at b/tests/interface-reconfigure.at
in
I'll review this series as I wrote the CFM code originally.
Ethan
On Wed, Mar 16, 2011 at 2:37 PM, Ben Pfaff wrote:
> This set of patches moves the CFM implementation from ovs-vswitchd
> into ofproto. This is one step toward my goal of getting rid of the
> hooks from ofproto back into vswitchd.
Looks Good to me. The fault won't get registered until the database
dumps again but this is an improvement.
Ethan
On Wed, Mar 16, 2011 at 2:37 PM, Ben Pfaff wrote:
> An unexpected MPID is always a fault, but the CFM code didn't signal the
> fault until the next time cfm_run() was called. In on
Garbage collection introduced in
d3643fc5989f88b1eda701d8aefca0ed3ff9dcdc changed ovs-vsctl so that
it would allow the garbage collector to reclaim unused tables
instead of manually deleting them itself. Since garbage collection
runs at transaction completion, undeleted tables would hang around
an
Note. I did not actually test that this fixes the xenserver issue
we've been seeing. It passes make check. I'll test xenserver before
merging.
Ethan
On Wed, Mar 16, 2011 at 3:24 PM, Ethan Jackson wrote:
> Garbage collection introduced in
> d3643fc5989f88b1eda701d8aefca0ed3ff9dcdc changed ovs-
On 3/15/11 11:36 AM, Ben Pfaff wrote:
> This removes some code from ofproto.c.
Thanks, looks good to me.
-Andrew
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
This does in fact fix the xenserver issue we were seeing.
On Wed, Mar 16, 2011 at 3:25 PM, Ethan Jackson wrote:
> Note. I did not actually test that this fixes the xenserver issue
> we've been seeing. It passes make check. I'll test xenserver before
> merging.
>
> Ethan
>
> On Wed, Mar 16, 201
On Wed, Mar 16, 2011 at 03:24:22PM -0700, Ethan Jackson wrote:
> Garbage collection introduced in
> d3643fc5989f88b1eda701d8aefca0ed3ff9dcdc changed ovs-vsctl so that
> it would allow the garbage collector to reclaim unused tables
> instead of manually deleting them itself. Since garbage collectio
Oops about the commit.
I'll go ahead and add a test case and resubmit.
Ethan
On Wed, Mar 16, 2011 at 3:34 PM, Ben Pfaff wrote:
> On Wed, Mar 16, 2011 at 03:24:22PM -0700, Ethan Jackson wrote:
>> Garbage collection introduced in
>> d3643fc5989f88b1eda701d8aefca0ed3ff9dcdc changed ovs-vsctl so th
I tested this against our autobuilder and it passed.
---
utilities/automake.mk |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/utilities/automake.mk b/utilities/automake.mk
index 1233762..147e650 100644
--- a/utilities/automake.mk
+++ b/utilities/automake.mk
@@ -63,7 +63,
Looks Good.
On Wed, Mar 16, 2011 at 3:45 PM, Ben Pfaff wrote:
> I tested this against our autobuilder and it passed.
> ---
> utilities/automake.mk | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/utilities/automake.mk b/utilities/automake.mk
> index 1233762..147e650
Thanks, I pushed this.
On Wed, Mar 16, 2011 at 03:47:15PM -0700, Ethan Jackson wrote:
> Looks Good.
>
> On Wed, Mar 16, 2011 at 3:45 PM, Ben Pfaff wrote:
> > I tested this against our autobuilder and it passed.
> > ---
> > ?utilities/automake.mk | ? ?2 +-
> > ?1 files changed, 1 insertions(+), 1
I noticed one more thing. After this patch, lacp_slave_enable() is
reduced to just:
void
lacp_slave_enable(struct lacp *lacp, void *slave_, bool enabled)
{
struct slave *slave = slave_lookup(lacp, slave_);
if (slave->enabled != enabled) {
slave->enabled = enabled;
}
}
which
Haha this is why I didn't want to send the series out yet =).
Ethan
On Wed, Mar 16, 2011 at 4:22 PM, Ben Pfaff wrote:
> I noticed one more thing. After this patch, lacp_slave_enable() is
> reduced to just:
>
> void
> lacp_slave_enable(struct lacp *lacp, void *slave_, bool enabled)
> {
> stru
Well it's been very helpful to look at it so thanks for humoring me.
On Wed, Mar 16, 2011 at 04:47:45PM -0700, Ethan Jackson wrote:
> Haha this is why I didn't want to send the series out yet =).
>
> Ethan
>
> On Wed, Mar 16, 2011 at 4:22 PM, Ben Pfaff wrote:
> > I noticed one more thing. ?Afte
No problem =).
On Wed, Mar 16, 2011 at 4:49 PM, Ben Pfaff wrote:
> Well it's been very helpful to look at it so thanks for humoring me.
>
> On Wed, Mar 16, 2011 at 04:47:45PM -0700, Ethan Jackson wrote:
>> Haha this is why I didn't want to send the series out yet =).
>>
>> Ethan
>>
>> On Wed, Ma
Garbage collection introduced in
c5f341ab193b9126dffef8c77bf8ed35e91290fd changed ovs-vsctl so that
it would allow the garbage collector to reclaim unused tables
instead of manually deleting them itself. Since garbage collection
runs at transaction completion, undeleted tables would hang around
an
David wants this in as soon as possible. Could someone please have a
look at the lastest version of the patch. Simply changed the commit
message and added tests.
Ethan
On Wed, Mar 16, 2011 at 4:52 PM, Ethan Jackson wrote:
> Garbage collection introduced in
> c5f341ab193b9126dffef8c77bf8ed35e91
On Wed, Mar 16, 2011 at 5:23 PM, Ethan Jackson wrote:
> David wants this in as soon as possible. Could someone please have a
> look at the lastest version of the patch. Simply changed the commit
> message and added tests.
It's fine. Thanks for the updates.
_
Thanks. I pushed this.
On Wednesday, March 16, 2011, Ben Pfaff wrote:
> On Wed, Mar 16, 2011 at 5:23 PM, Ethan Jackson wrote:
>> David wants this in as soon as possible. Could someone please have a
>> look at the lastest version of the patch. Simply changed the commit
>> message and added test
49 matches
Mail list logo
