I think I know whats happening. In case there is an error getting the
datapath flow we should not try logging it since the flow is not valid.
dpif_netlink_operate__
{
...
case DPIF_OP_FLOW_GET:
op->error = dpif_netlink_flow_from_ofpbuf(&reply,
txn->reply);
i
I am able to recreate it. I suspect its some bug at the dpif_netlink layer.
I am trying to narrow it down further. Anything in particular I could look
for ?
Thanks.
(gdb) print *get
$3 = {
key = 0x7fedec092190,
key_len = 0x60,
buffer = 0x7fee162a4b70,
flow = 0x7fee162a4b20
}
(gdb) print *
Pushed, thanks for the review!
Jarno
On Oct 17, 2014, at 4:05 PM, Daniele Di Proietto wrote:
> Thanks for fixing this Jarno!
>
> I've tested the patch and it fixes the problem.
> Would you mind adding a comment on struct netdev_flow_key that 'len'
> accounts for the length of the miniflow on
Signed-off-by: Daniele Di Proietto
---
INSTALL.DPDK | 40 +---
1 file changed, 13 insertions(+), 27 deletions(-)
diff --git a/INSTALL.DPDK b/INSTALL.DPDK
index 7484b4b..0d19012 100644
--- a/INSTALL.DPDK
+++ b/INSTALL.DPDK
@@ -11,6 +11,8 @@ It has not been thor
Thanks for the patch.
Acked-by: Daniele Di Proietto
Pravin, would you mind pushing this? I suggest one minor change, inline.
On 10/13/14, 7:17 AM, "maryam.tahhan" wrote:
>This patch updates the documentation to reflect that DPDK 1.7.1 is
>supported.
>Travis scripts have also been updated to
Forgot to attach the backtrace ::
(gdb) bt
#0 nl_attr_is_valid (maxlen=0xcfea1104, nla=0x0) at lib/netlink.h:139
#1 format_odp_actions (ds=0x7f77c7ffc800, actions=0x0,
actions_len=0x7f77cfea1104) at lib/odp-util.c:608
#2 0x00456749 in log_flow_message (error=0x2,
operation=0x506
You are correct Ben. I confused with the put that has actions in
dpif_flow_put. I guess I was not able to repro the issue then. Let me repro
it and I will resend the fix.
Thanks.
On Fri, Oct 17, 2014 at 4:17 PM, Ben Pfaff wrote:
> On Fri, Oct 17, 2014 at 04:07:25PM -0700, Madhu Challa wrote:
>
On Fri, Oct 17, 2014 at 04:07:25PM -0700, Madhu Challa wrote:
> dpif_flow_get initializes the flow_get part of the union, down the stack
> log_flow_message checks for actions || actions_len that could contain
> garbage leading to the crash.
>
> saw the crash once when running stress tests. can be
dpif_flow_get initializes the flow_get part of the union, down the stack
log_flow_message checks for actions || actions_len that could contain
garbage leading to the crash.
saw the crash once when running stress tests. can be easily recreated
by running ovs-dpctl del-flows in a loop when traffic i
Thanks for fixing this Jarno!
I've tested the patch and it fixes the problem.
Would you mind adding a comment on struct netdev_flow_key that 'len'
accounts for the length of the miniflow only (including the map)?
Other comments inline, otherwise:
Acked-by: Daniele Di Proietto
On 10/17/14, 3:05
On Fri, Oct 17, 2014 at 12:25:41PM -0700, Gurucharan Shetty wrote:
> There was not much difference between the two files after moving
> all of the Windows socket HANDLE polling functionality to poll-loop.c.
> So merge them together.
>
> Signed-off-by: Gurucharan Shetty
Besides reading the code,
On Fri, Oct 17, 2014 at 01:23:04PM -0700, Gurucharan Shetty wrote:
> On Tue, Oct 7, 2014 at 2:25 PM, Ben Pfaff wrote:
> > On Thu, Oct 02, 2014 at 08:16:56AM -0700, Gurucharan Shetty wrote:
> >> On Windows platform, TCP_NODELAY can only be set when TCP is established.
> >> The current code does not
Patch 0de8783a9 (lib/dpif-netdev: Integrate megaflow classifier.)
broke exact match cache lookup, but it went undetected since there are
no separate stats for EMC.
This patch fixes the problem by changing the struct netdev_flow_key
'len' member to cover only the 'mf' member, not the whole
netdev_f
On Thu, Oct 16, 2014 at 11:45:42PM -0700, Eitan Eliahu wrote:
> This change includes the following:
> [1] Handler for subscribe/unsubscribe to a packet queue associated with a
> socket pid.
> [2] Allocation of per socket packet queue on a packet subscription.
> [3] Removal of static allocated q
On Fri, Oct 17, 2014 at 11:12:32AM -0700, Ankur Sharma wrote:
> Added the handlers for update_flags and set_etheraddr.
> These handlers were needed for vswitchd bringup on windows
> platform.
>
> Signed-off-by: Ankur Sharma
> Acked-by: Nithin Raju
Applied, thanks.
__
On Fri, Oct 17, 2014 at 01:19:33PM -0400, Dave Benson wrote:
> On Fri, Oct 10, 2014 at 6:43 PM, Ben Pfaff wrote:
> > With this change, I think that an OpenFlow flow table entry that
> > matches on a VLAN will now match both 0x88a8 and 0x8100 tags for that
> > VLAN. I doubt that that is desirable
Thanks.
On Fri, Oct 17, 2014 at 1:24 PM, Simon Horman
wrote:
>
> 2014/10/17 19:25 "Pravin Shelar" :
>>
>> On Fri, Oct 17, 2014 at 12:16 AM, Simon Horman
>> wrote:
>> > On Thu, Oct 16, 2014 at 04:50:10PM -0700, Andy Zhou wrote:
>> >> Simon, The change makes a lot of sense.
>> >>
>> >> I am just w
On Oct 16, 2014, at 11:09 PM, Alex Wang wrote:
> Thx for the revision,
>
>
> +/* Removes 'rule' from 'cls', also destructing the 'rule'. */
> +static void
> +dpcls_remove(struct dpcls *cls, struct dpcls_rule *rule)
> +{
> +struct dpcls_subtable *subtable;
> +
> +ovs_assert(rule->mask);
2014/10/17 19:25 "Pravin Shelar" :
>
> On Fri, Oct 17, 2014 at 12:16 AM, Simon Horman
> wrote:
> > On Thu, Oct 16, 2014 at 04:50:10PM -0700, Andy Zhou wrote:
> >> Simon, The change makes a lot of sense.
> >>
> >> I am just wondering if we should upstream the netlink.h change first?
> >> To me, it
There was not much difference between the two files after moving
all of the Windows socket HANDLE polling functionality to poll-loop.c.
So merge them together.
Signed-off-by: Gurucharan Shetty
---
lib/automake.mk |5 +-
lib/stream-fd-windows.c | 270 -
On Tue, Oct 7, 2014 at 2:25 PM, Ben Pfaff wrote:
> On Thu, Oct 02, 2014 at 08:16:56AM -0700, Gurucharan Shetty wrote:
>> On Windows platform, TCP_NODELAY can only be set when TCP is established.
>> The current code does not create any problems while running unit tests
>> (because connections get e
Added the handlers for update_flags and set_etheraddr.
These handlers were needed for vswitchd bringup on windows
platform.
Signed-off-by: Ankur Sharma
Acked-by: Nithin Raju
---
lib/netdev-windows.c | 30 +-
1 file changed, 29 insertions(+), 1 deletion(-)
diff --git
I’ll see if we can make the iteration both safe and lockless, so there is no
need to review this patch at this moment.
Jarno
On Oct 16, 2014, at 4:24 PM, Jarno Rajahalme wrote:
> This patch changes the classifier internal mutex to a recursive type.
> This allows simplification of locking dur
On Fri, Oct 17, 2014 at 12:16 AM, Simon Horman
wrote:
> On Thu, Oct 16, 2014 at 04:50:10PM -0700, Andy Zhou wrote:
>> Simon, The change makes a lot of sense.
>>
>> I am just wondering if we should upstream the netlink.h change first?
>> To me, it seems to add a new compat API that does not exist u
Ben,
Thanks for your thoughtful comments. Apologies for my late reply...
Dave.
On Fri, Oct 10, 2014 at 6:43 PM, Ben Pfaff wrote:
> On Fri, Oct 10, 2014 at 10:06:29AM -0400, Dave Benson wrote:
> > Previous behavior forced the use of the 802.1Q Ethertype for all VLAN
> headers.
> > In order to
Hi Nithin,
Please find my comments inline.
Regards,
Ankur
From: dev on behalf of Nithin Raju
Sent: Thursday, October 16, 2014 10:52 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 4/4] lib/netlink-socket.c: fixes in
nl_sock_recv__() on Windows
Hi Nithin,
While reading the ETHERTYPE attribute we already use ntohs.
Regards,
Ankur
From: dev on behalf of Nithin Raju
Sent: Thursday, October 16, 2014 10:52 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 2/4] datapath-windows: fixes in Flow.c i
On Thu, Oct 16, 2014 at 05:54:30PM -0700, Gurucharan Shetty wrote:
> Fixes a rpmbuild failure.
>
> Signed-off-by: Gurucharan Shetty
Acked-by: Ben Pfaff
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
Hi Nithin,
Changes look fine.
Please find a minor comment inline.
Acked-by: Ankur Sharma
From: dev on behalf of Nithin Raju
Sent: Thursday, October 16, 2014 10:52 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 3/4] datapath-windows: Fixes in pack
Acked-by: Ankur Sharma
From: dev on behalf of Nithin Raju
Sent: Thursday, October 16, 2014 10:52 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 1/4] datapath-windows: event read should not fail
when no events
The semantics are read operation
Thanks. I marked Dongdong as the author and retained his Signed-off-by,
and changed your Signed-off-by to an Acked-by to indicate that you
reviewed and approved. I applied this to master. Thanks again!
On Fri, Oct 17, 2014 at 01:43:10AM +, Lichunhe wrote:
> We work together, and talk about
I'm happy to hear that it works for you. Thank you for testing!
On Fri, Oct 17, 2014 at 07:41:38AM +, Niels van Adrichem wrote:
> Sorry for my late reply, I am on a different timezone (Amsterdam time,
> GMT+2).
>
> It is good to bring back the amounts of operations in
> odp_port_is_alive() a
Sorry for my late reply, I am on a different timezone (Amsterdam time,
GMT+2).
It is good to bring back the amounts of operations in
odp_port_is_alive() as it is checked at every packet, hence a single
variable containing a summary of all combinations is the right way.
To answer your question, th
Your message was not delivered due to the following reason:
Your message could not be delivered because the destination server was
unreachable within the allowed queue period. The amount of time
a message is queued before it is returned depends on local configura-
tion parameters.
Most likely the
On Thu, Oct 16, 2014 at 04:50:10PM -0700, Andy Zhou wrote:
> Simon, The change makes a lot of sense.
>
> I am just wondering if we should upstream the netlink.h change first?
> To me, it seems to add a new compat API that does not exist upstream.
Sure, I think that makes sense. Though I'm not sur
Hi Flavio!
I was recently doing some develop job on igmp snooping base on ovs, but found
failed to add route port to mrouter list.
As I looking into this patch, I found that an igmp query must have a source
network address to make the in_port adding to the mrouter
, while the igmp querier, whic
36 matches
Mail list logo
